What this chapter gives you. The aerial photograph before the street map. Chapters 1 through 7 each teach one move, and moves are easier to learn when you can see the board. This chapter shows you the board: what the Swiss legal machine is made of, which codes are its master texts, the two operating modes it runs at once, the difference between the law that defines your rights and the machinery that enforces them, and the two disputes this book keeps returning to. Nothing here needs to be memorized. It needs to be seen once, so that when Chapter 1 starts decomposing articles, you know what building the articles live in.

The claim of this book

Swiss private law behaves like a machine. Not a metaphorical machine, a literal one in the engineering sense: a system with defined inputs, rules that behave predictably, and workings you can inspect. Feed it a dispute and it does not ask who was nicer. It asks: which norm was invoked, are that norm's conditions satisfied, who proved which condition, and what consequence follows.

This is unusual as legal systems go, and it is why the book exists. Codified civil law, the family Switzerland belongs to, was designed — drafted as a coherent system by identifiable people (Eugen Huber wrote the Civil Code essentially alone) rather than accreted case by case like the common law. Design shows. The codes have a clear division of labor, general parts that act like a general rulebook, defaults that can be overridden, and exceptions that are flagged in the text. A person who thinks in systems can read them the way an engineer reads a blueprint: not line by line as prose, but structurally, asking which rule relies on which. (If you come from software: the general parts behave like base classes, and the specialized rules like subclasses that override their defaults.)

Where Switzerland sits on the map

The deep split among legal systems is binary. Common-law systems (England, the US, most of the Commonwealth) have no master code: law accretes case by case, precedent binds (stare decisis), and statutes patch the case law rather than replace it. Asking "which article, which elements" is simply not how you reason there — you reason from decided cases. Civil-law systems — most of continental Europe, Latin America, much of Asia — share Switzerland's big idea: a designed, comprehensive code of norms with conditions and consequences. Within the family, the method of this book transfers; what changes is how the books are laid out:

The design exported well: Turkey adopted the Swiss CC and CO nearly verbatim in 1926, so the machine described here runs there with the same article numbers. The takeaway: Part I's seven moves are Swiss in their addresses but civil-law-generic in their method. Decomposing a norm into elements works in Paris or Berlin with different article numbers; in London it is the wrong tool entirely.

The machine is not fully deterministic — judges evaluate the open terms, and Chapter 11 is honest about how much play that adds. But it is deterministic enough that the untrained party arguing "this is unfair" reliably loses to the trained party asking "which article, which element, whose burden." This book trains the second kind.

The source texts

Swiss federal law is public, versioned, and addressable. Every act has a number in the Recueil systématique (RS), and the current consolidated text lives on Fedlex, in three equally authentic languages (French, German, Italian — pick one, cite precisely). The texts you will actually open:

Two structural facts worth noting now. First, the CO is formally the fifth book of the CC — obligations are not a separate universe but a chapter of the same system, which is why general CC provisions like art. 2 CC (good faith) and art. 8 CC (burden of proof) apply inside contract disputes without ceremony. Second, the CC opens with a remarkable bootstrap instruction: art. 1 CC tells the judge that where the statute is silent and custom offers nothing, the judge decides according to the rule they would adopt as legislator. The system declares its own fallback rule on page one.

Above these sits the norm stack — Constitution, statutes, ordinances, cantonal law, and at the bottom, your contract — with precedence rules governing who overrides whom. Chapter 10 (ch. 10) walks the whole stack; until then, one simplification serves: for private disputes, the action is in the CC and CO, and your contract is the most local override level.

Two operating modes, one system

The machine runs two permission regimes simultaneously, and confusing them is the most common civilian error.

Horizontal (private law): allowed unless forbidden. Between private parties — you and your landlord, Ekeko and its clients — freedom is the base state. Anything not forbidden can be agreed. Most statutory rules are dispositive: defaults that apply only where your contract is silent. The law is a set of fallback defaults, not a permission system.

Vertical (public law): forbidden unless authorized, for the state. When the state acts on you — taxes, permits, fines — the polarity inverts: the state may only do what a norm explicitly authorizes. You never need a legal basis to act; the administration always does.

One event can trigger both modes at once (a data leak: contractual liability to the client, horizontal; regulatory action, vertical). Knowing which mode you are in tells you who needs to produce a legal basis and which way the presumptions point. Chapter 8 (ch. 8) develops this; this book is about the horizontal mode.

Rights on paper, enforcement in the machinery

A distinction the whole of Part II leans on: the CC and CO define what you are owed (substantive law), but a right you cannot enforce is a promise on paper with nothing behind it. Enforcement is a separate subsystem — the CPC for getting a judgment, the LP for turning a judgment (or even just a claim) into seized assets. The two subsystems have their own rules, their own deadlines, and their own strange loops (in Switzerland you can start debt collection before proving your claim — the famous poursuite circuit of Chapter 9, ch. 9).

The practical consequence, previewed now because it reframes everything: the value of a claim is not what the code says you are owed. It is what the code says, times the probability you can prove it, times the cost of the path to collect it. Chapter 12 (ch. 12) prices it out.

Where disputes actually die

Here is the distribution nobody shows you in a law course. As a citizen — tenant, employee, insured, freelancer, occasional buyer of expensive things — you will meet the machine maybe a dozen times in your life, and almost none of those encounters will look like court. They look like a funnel:

Two readings of this table, one cynical and one useful. The cynical one: the law barely matters, since almost everything is settled informally. The useful one, and the premise of this book: the law matters most in the informal phase, because informal settlements are priced in the shadow of what a court would do. Both sides are estimating the same thing — which norm, which elements, whose proof, what path — and the side that can actually run that calculation credibly sets the price. The other side is guessing.

That is what leverage means in a settlement negotiation, and it decomposes exactly along the strength formula of Chapter 12: provability × enforceability × opponent's costs. A registered letter that names the claim-foundation norm, walks its elements against documents you visibly possess, and states the next procedural step you are prepared to take is not "being formal." It is a demonstration that your probability terms are high and that continuing will cost the other side real money — which moves their expected value below your number. The Florissant CHF 4,700 claim died in this phase: no judge, no lawyer on retainer, one letter built the way Part I teaches (Chapter 13 shows the letter's anatomy). The machine's day job is not deciding disputes. It is pricing them, so that the informal market — where you will live — clears at a number that tracks your rights instead of your opponent's nerve.

The two disputes on the bench

Abstract law does not stick. This book keeps two concrete disputes on the workbench and returns to them chapter after chapter:

Florissant 31 (real, won). A departing tenant in Prilly; a régie claiming CHF 4,700 of the deposit for wear on a 13.5-month tenancy; an exit inspection report marking six of seven items « usure normale »; and a « convention de sortie » that only the agent signed because the tablet failed. Every move of Part I gets tested against this file, and Chapter 13 (ch. 13) replays the whole case end to end.

Ekeko (fictional, instructive). A small Swiss consultancy takes a modeling mandate; scope creeps, the client stops paying, the relationship walks away. This one exists to exercise the contract-drafting side: what the CO's defaults do to you when your contract is silent, and the six clauses that would have prevented the mess (Chapter 14, ch. 14).

One is defense (a claim aimed at you), one is offense (a claim you must assert and a contract you must draft). Between them they cover most of what a non-lawyer running their own affairs actually meets.

The seven moves at a glance

Part I is a toolkit of seven moves. Each is small; the power is in composition.

Part II (Chapters 8–12) supplies the five context frames sketched above — the two modes, the enforcement machinery, the norm stack, the judges, the settlement market. Part III runs everything on the two cases. Part IV is the long view for the curious: the same dispute reasoned under common law and civil law (ch. 15), and the four-thousand-year history that produced the codes (ch. 16). The appendices (ch. 17) hold the lookup tables: articles cited, deadlines, and how to search Fedlex.

What the machine is not

Three honest limits, stated once so the machine metaphor does not overpromise. The finest-grained evaluations are human: "normal wear," "good faith," "important reason" are judgment calls a judge makes, not tests a machine could run mechanically. The machine has a safety valve: art. 2 CC means a technically flawless position deployed in a shocking way can still lose (Chapter 7). And running the machine costs money and months: being right is an asset only net of the cost of proving it (Chapters 9 and 12). None of this weakens the method. It defines where the method's guarantees end, which is exactly what you want to know about any system you rely on.

The picture in one line

One designed system, two operating modes, master texts in the CC/CO with your contract as the most local override, a separate machinery for enforcement — and every dispute inside it reducible to: which norm, which elements, whose proof.

Drills

1. Your commune bills you a waste-collection fee you think is too high. Your cousin's moving company bills you CHF 400 more than the quote. Which operating mode is each dispute in, and who must produce a legal basis?

Answer: The commune is vertical: public law, forbidden unless authorized, so the commune must point to the legal norm authorizing that fee and its calculation. The moving company is horizontal: private law, and the claimant (the company) must name the foundation of the extra CHF 400 — a contract clause or a norm. In both cases someone owes you a citation before you owe them an argument.

2. A friend says: "The law says my deposit must be returned, so I'll get it back." Name the two multiplications this statement ignores.

Answer: Provability (can the elements of the restitution claim be proven, and are the landlord's counterclaims disprovable — Chapters 1–3) and the path (the claim must be run through the CPC/LP machinery, which costs time and money and may need to beat a deadline — Chapter 9). Substantive right × probability of proof × cost of enforcement is the real value (Chapter 12).

3. You need the current, authentic text of art. 267 CO in French. Where do you get it, and why not from a blog or a PDF you downloaded last year?

Answer: Fedlex (RS 220), the official consolidated collection. The three language versions there are equally authentic, and the consolidation is versioned — statutes get amended, and a stale copy can silently carry a repealed default or a changed deadline. Always resolve to the current RS text before relying on an article (Appendix C).

Exercises

Self-test

Fieldwork

• Fedlex · ~20 minOpen fedlex.admin.ch and resolve RS 220 to the Code of Obligations. Scroll the table of contents until you can point at where the general part ends (art. 183 CO) and the named contracts begin (art. 184 CO, sale). Then open art. 1 CC (RS 210) and read the famous fallback ladder in the actual text — three alinéas, one page. Goal: navigate from an RS number to an article without a search engine.
• Your own files · ~30 minDig out one official-looking demand you have actually received — an invoice reminder, an insurance denial, a régie letter. In the margins, mark the four slots of the magic question: who, what, from whom, based on what. Note whether slot 4 contains a norm, a mere document, or nothing. Keep the letter — the exercises of Chapters 1 to 3 reuse it.

What this chapter gives you. The single most useful habit in Swiss private law: stop reading articles as prose and start reading them as conditional rules — fixed forms with slots to fill. Every operative article of the Code des obligations (CO) and the Code civil (CC) reduces to: if all elements are satisfied → legal consequence. Lawyers call the condition side the Tatbestand (in French, l'état de fait légal or les conditions) and the consequence side the Rechtsfolge (la conséquence juridique). Once you can mechanically decompose an article into its element checklist, three things follow for free: you know exactly what must be proven (the burden question, see ch. 3), you know what the other side must attack, and you know what a court is actually deciding. This chapter builds the decomposition method and works it live on the two articles you have already met in real life: art. 267 CO from your deposit dispute and art. 41 CO, the general tort norm.

Why prose-reading fails

An article in the CO looks like a sentence. It is not a sentence. It is a required form wearing a sentence costume — a fixed set of slots, each of which must be filled before anything is owed.

Read art. 41 al. 1 CO the way you would read a novel and you get a vague moral vibe: "people who hurt others should pay." That reading is useless in a dispute, because a dispute is never about the vibe. It is about whether this claimant gets this money from this defendant, and the court will walk the elements one by one. If one element fails, the claim fails, no matter how sympathetic the vibe. Prose-reading hides the element boundaries, hides which elements are load-bearing, and hides the quiet exceptions ("à moins que...", "sauf convention contraire") that flip the burden of proof or the default rule.

The failure mode is familiar from any inspection trade. A casual visitor walks through a house and says "it looks solid." A building inspector asks: which walls carry the load, where does the water go when it rains, what happens if the ground shifts? Legal reading is inspector reading. Every clause is doing work, and the work is usually gatekeeping. (If you come from software: it is the difference between a junior engineer's "this function handles errors" and a senior's "which errors, on which path, and what does it return when the input is null?")

There is a second failure mode: prose-reading makes you argue at the wrong altitude. People in disputes argue fairness ("the apartment was fine!") when the legal machine is evaluating a precise yes/no question ("does the alleged damage exceed usure normale under art. 267 al. 1 CO, and who proved what?"). You won your deposit dispute precisely because you argued at checklist altitude while the régie argued at vibe altitude.

Tatbestand and Rechtsfolge: the if and the then

Every claim-granting article has this shape:

Tatbestand (conditions)                    Rechtsfolge (consequence)
condition 1 AND condition 2 AND ...   →    someone owes / may demand / acquires something
(every one must hold)

Terminology, once, so the German and French map cleanly:

Subsomption is the step everyone underestimates: taking your messy facts and testing each one against each element. Facts are the raw material; the elements are the slots on the form. A claim is well-founded when every slot is filled by at least one proven fact. The proving part is Chapter 3's territory (art. 8 CC, see ch. 3); this chapter is about extracting the checklist correctly in the first place.

One honest caveat on the analogy: elements are not crisp yes/no switches. Several are evaluative standards that a human judge resolves ("unlawful", "normal wear"). The checklist structure is exact; the evaluation of individual elements can be fuzzy. More on that below under open terms.

Worked live: art. 267 al. 1 CO

The text: at the end of the lease, the tenant must return the thing in a state that results from a use conforming to the contract ("le locataire doit restituer la chose dans l'état qui résulte d'un usage conforme au contrat", art. 267 al. 1 CO).

Decomposition:

Tatbestand: 1. A lease existed (bail, art. 253 CO defines the contract type). 2. The lease has ended (termination, expiry, whatever route; the lifecycle routes live in ch. 6). 3. The tenant returns, or must return, the leased thing.

Rechtsfolge: an obligation of the tenant, with a built-in quality standard: the state owed is not "pristine" and not "as received", it is "the state resulting from contract-conforming use."

Notice what the decomposition surfaces that prose hides:

• The article defines the content of the return obligation, and the quality standard is the whole game. Degradation from ordinary living, usure normale (normal wear and tear), is inside "use conforming to the contract", so it is by definition not damage the tenant owes for. The landlord who wants money must therefore establish something beyond that standard: excessive wear or actual damage. That is a fact the landlord alleges to derive a right, so under art. 8 CC the landlord proves it. The checklist reading of art. 267 al. 1 CO hands you the burden allocation almost automatically.
• Alinea 2 of the same article is a separate mini-rule with its own Tatbestand: any clause by which the tenant commits in advance to pay an indemnity covering more than actual damage is void ("nulle", art. 267 al. 2 CO). That is a mandatory norm, a precedence tag in the sense of ch. 4: the contract cannot override it.
• The famous inspection-and-notice rule is the next article, not this one: the landlord must check the state of the thing on return and immediately notify the tenant of defects for which the tenant is answerable, otherwise the tenant is discharged, except for defects not discoverable on an ordinary examination (art. 267a CO). Checklist reading trains you to notice article boundaries: 267 gives the obligation, 267a gives the landlord a procedural trap with a forfeiture consequence. Two rules, two checklists.

Worked live: art. 41 al. 1 CO

The text: whoever unlawfully causes damage to another, whether intentionally or by negligence or imprudence, is bound to repair it (art. 41 al. 1 CO). This is the general tort norm, one of the big claim-foundation norms from the magic question in ch. 2.

Tatbestand (four cumulative elements, the canonical decomposition): 1. Damage (dommage): a quantifiable diminution of the claimant's patrimony. 2. Unlawfulness (illicéité): violation of an absolute right (body, property, personality) or of a protective norm. 3. Causation (lien de causalité, natural and adequate): the act produced the damage. 4. Fault (faute): intent OR negligence OR imprudence.

Rechtsfolge: obligation to repair the damage (quantum and mode governed by art. 42 ff. CO).

Miss any one of the four and the claim dies. There is no partial credit, no "three out of four is pretty good." All four must hold, without exception. This is why defending a tort claim is structurally easier than bringing one: the defendant needs to knock out any single element, the claimant must hold all four, and must prove all four (art. 8 CC; damage quantum gets a softened proof regime in art. 42 al. 2 CO when exact proof is impossible, a deliberate loosening of the standard you will meet again in ch. 3).

Cumulative vs alternative elements

Art. 41 al. 1 CO demonstrates both connectives in one line:

• The four elements are cumulative: joined by AND, all required.
• Inside element 4, the fault modes are alternative: intent OR negligence OR imprudence, any one suffices. "Soit... soit..." in the French text is the giveaway, as are "ou" and enumerations introduced by "notamment".

Getting the connective wrong inverts your whole strategy. If you think alternatives are cumulative, you over-prove and waste effort. If you think cumulatives are alternative, you build a claim with a hole in it and lose. When decomposing, annotate every branch point explicitly:

art. 41 al. 1 CO:
  damage        AND
  unlawfulness  AND
  causation     AND
  fault (intent OR negligence OR imprudence)
→ duty to repair

Watch also for a third connective: the exception clause, usually flagged by "à moins que", "sauf si", "sauf convention contraire". Exceptions are not elements of the claim; they are defenses, and defenses carry their own burden of proof, borne by whoever invokes them (art. 8 CC logic, developed in ch. 3). Art. 97 al. 1 CO is the classic: the debtor escapes liability "à moins qu'il ne prouve qu'aucune faute ne lui est imputable", so absence of fault is the debtor's exception to prove, not the creditor's element. The article even says "prouve" out loud. Statutes rarely hand you the burden allocation this explicitly; when they do, take the gift.

Defined terms vs open terms

Elements come in two kinds.

Defined terms have their definition somewhere in the code, and you must chase the reference. "Bail à loyer" is defined at art. 253 CO. "Demeure" (debtor's default) is constructed by art. 102 CO. Reading an element without resolving its definition is quoting a rule while never opening the fine print it points to.

Open terms (notions juridiques indéterminées, in German unbestimmte Rechtsbegriffe) are deliberately underspecified: "usure normale", "justes motifs", "sans délai", "faute grave", "bonne foi". These are not sloppy drafting. They are deliberate blanks: the legislator ships the name of the standard and delegates its content to judges, case by case, so the rule survives contact with situations nobody anticipated in 1911. The settled answers accumulate in Tribunal fédéral case law, the reservoir of reusable answers described in ch. 11. For usure normale, practice has even grown lookup tables outside the statute entirely: the paritarian lifespan tables (tables de longévité) that say a carpet is amortized over ten years, paint over eight, and so on. None of that is in the CO. The article just says "usage conforme au contrat" and lets the ecosystem fill in the numbers.

Practical consequence: when your checklist hits an open term, your work forks. For a defined term you cite the definition and subsume. For an open term you argue the content: precedent, tables, expert practice, and the burden of proof does heavy lifting because open terms are exactly where facts get murky. You lived this: "usure normale" was the open term, the EDL plus the lifespan tables supplied its content, and art. 8 CC decided who suffered from the murk.

The decomposition method, step by step

1. Isolate the operative sentence. One alinea, one rule. Never decompose an article wholesale; art. 267 al. 1 and al. 2 CO are different rules.
2. Find the Rechtsfolge first. Look for the verb of consequence: "est tenu de", "doit", "peut exiger", "est nul", "se prescrit". The consequence tells you what kind of norm this is: claim-granting, void-making, right-shaping, deadline-setting.
3. Everything else is Tatbestand. List each factual condition as its own line item, even the ones hiding in participles and relative clauses ("celui qui...", "lorsque...", "à la fin du bail").
4. Mark the connectives. AND, OR, and exception clauses ("à moins que", "sauf"). Exceptions go on a separate list labeled defenses.
5. Type each element. Defined term (chase the definition, cite it), open term (note that its content lives in case law and practice), or plain fact (dates, payments, signatures).
6. Check the neighbors. Read the articles immediately before and after; codes cluster related rules, and the trap or the definition you need is often next door (art. 267a CO next to art. 267 CO).
7. Only then subsume. Walk your facts down the checklist. Every unchecked box is either "gather evidence" or "abandon the claim."

Steps 1 through 6 need no facts at all. You can, and should, decompose the article before you look at your dossier. Checklist first, facts second — the way a surveyor draws the grid before plotting a single point on it. (If you come from software: schema first, data second.)

Three articles, decomposed

Read the middle column twice. Art. 97 al. 1 CO is the workhorse claim-foundation norm for every contract breach in this book, and its structure, four elements for the creditor plus one exculpation defense for the debtor, is the cleanest illustration of why decomposition matters: the creditor's checklist does not contain fault. Fault is presumed, and disproving it is the debtor's problem. A prose-reader almost never spots that inversion. A checklist-reader cannot miss it.

Where the analogy breaks

Two honest limits. First, elements are not mechanically checkable conditions; open terms make individual checks judgment calls, and judges, not machines, run the evaluation. The structure is rigid, the individual evaluations are not. Second, real analysis is rarely one article deep: the Rechtsfolge of one norm is often an element of another (a valid contract under art. 1 CO is element 1 of art. 97 CO), so claims resolve through chained checklists — a chain of nested questions rather than a single form. Chapter 2 (ch. 2) gives you the entry point into that chain: which norm to start decomposing at all.

The move in one line

Never read an article as a sentence: split it into the if (elements, each typed and connective-tagged) and the then (consequence), and treat every unchecked element as either an evidence task or a dead claim.

Drills

1. Decompose art. 102 al. 1 CO: "Le débiteur d'une obligation exigible est mis en demeure par l'interpellation du créancier." List the elements and the Rechtsfolge.

Answer: Tatbestand: (1) an obligation exists, (2) it is due/exigible (exigible), (3) the creditor issues a demand (interpellation). Rechtsfolge: the debtor is in default (demeure). Bonus if you checked the neighbor: art. 102 al. 2 CO makes the interpellation unnecessary when a deadline was fixed by agreement, an alternative route into the same Rechtsfolge.

2. In art. 41 al. 1 CO, your opponent proved damage, causation, and negligence beyond doubt, but the act violated no absolute right and no protective norm. Outcome, and why?

Answer: Claim fails. The four elements are cumulative; unlawfulness (illicéité) is unproven, so the checklist is not satisfied. Three strong elements do not compensate for one missing element; there is no weighting across elements, only within the evaluation of a single open term.

3. A lease clause says: "On departure, the tenant pays a flat CHF 2,000 renovation fee regardless of the apartment's condition." Which decomposition step catches the problem, and which article kills the clause?

Answer: Step 6, check the neighbors, or step 4 applied to art. 267 CO as a whole: al. 2 is a separate rule stating that clauses where the tenant pre-commits to an indemnity covering more than actual damage are void (art. 267 al. 2 CO). It is a mandatory norm (see ch. 4), so party autonomy cannot save the clause. The flat fee is payable "regardless of condition", hence it can exceed actual damage, hence void.

Exercises

Self-test

Fieldwork

• Fedlex · ~30 minRead art. 201 CO, the buyer’s duty to inspect the goods and notify defects. Decompose al. 1 into elements, mark the connectives, then read on to find the consequence of missed notice (the goods are deemed accepted). Compare the structure with art. 267a CO: same three-part pattern — verification duty, immediate notice, forfeiture. The codes reuse their designs; learn one and you have learned several.
• Drafting · ~30 minTake the demand letter from Chapter 0’s fieldwork. Write out the element checklist of the norm it invokes (or the norm it should have invoked). For each element, note: proven by which document the sender visibly holds — or unproven. You have just produced the skeleton of the reply letter.

What this chapter gives you. One question that turns any legal mess into a tractable problem: who wants what from whom, based on what? Swiss lawyers call the underlying method the Anspruchsmethode (méthode des prétentions, claim method). It forces every argument to declare its starting point before it gets going. You will learn what counts as a valid starting point (a claim-foundation norm, fondement de la prétention), why naming it first wins arguments before they start, how claims chain into trees of main claim plus interest plus damages, and why a Geneva property manager's CHF 4,700 demand evaporated the moment this question was asked out loud.

The fixed form of every legal dispute

Strip away the letterhead and the Latin, and every legal claim must fill the same fixed form:

who + what + from whom + on what basis → enforceable obligation, or nothing

Four slots. Fill them or go home:

The first three slots are usually obvious. Anyone can write "pay me CHF 4,700." The entire game lives in the fourth slot. A demand without a foundation is a demand nothing in the legal system will act on: it reads as English, but no legal machinery picks it up and runs with it.

This is the discipline. Before you argue about fairness, evidence, tone, or who behaved badly, you ask the fourth-slot question. Often the conversation ends right there.

What qualifies as a claim-foundation norm

Not every article in the CO or CC can be the "based on what." Chapter 1 (see ch. 1) showed that norms come in two flavors: complete norms with a Tatbestand (conditions) and a Rechtsfolge (legal consequence), and auxiliary norms that only define terms or modify other norms. Only complete norms whose consequence is an obligation someone can demand qualify as claim foundations. Everything else is supporting machinery.

The main families, roughly in order of how often you will meet them:

1. The contract itself

The most common foundation is not an article at all. It is a clause. "Contractor delivers the model by March 1, client pays CHF 30,000" is a claim-foundation norm that you and your counterparty wrote yourselves. Art. 19 al. 1 CO gives you that power: within legal limits, the parties determine the content of the contract freely. When Ekeko invoices a client, the foundation of the payment claim is the contract clause fixing the fee, backed by art. 82 CO on the order of performance. You cite the statute only when the contract is silent, because the CO general part is the general rulebook whose default rules your contract overrides where the law is dispositive (see ch. 4 and ch. 5).

2. Contractual liability: art. 97 CO

When the contract was breached rather than performed, the workhorse is art. 97 al. 1 CO (inexécution des obligations): if the creditor cannot obtain performance, the debtor must repair the resulting damage, unless the debtor proves no fault is attributable to them. Elements: (1) a valid obligation, (2) its non-performance or bad performance, (3) damage, (4) causal link, and (5) fault, which is presumed, so the debtor carries the burden of disproving it. That reversed fault burden is the big difference from tort and the reason plaintiffs love contractual foundations (see ch. 3).

3. Tort: art. 41 CO

No contract between the parties? Then you need art. 41 al. 1 CO (acte illicite): whoever unlawfully causes damage to another, intentionally or negligently, must repair it. Elements: damage, unlawfulness, causation, fault. Here fault is not presumed, the victim proves everything per art. 8 CC. Tort is the fallback foundation between strangers: the cyclist who dents your car, the competitor who spreads false claims.

4. Unjust enrichment: art. 62 CO

The cleanup rule. Art. 62 al. 1 CO (enrichissement illégitime): whoever is enriched without legitimate cause at another's expense must return the enrichment. This applies when value moved but no valid contract explains it: you paid twice, you paid under a contract later voided, you paid a debt that did not exist (art. 63 CO adds conditions for recovering voluntary payments: you must have paid in error). Enrichment is subsidiary in spirit: if a contract governs the transfer, the contractual rules run first.

5. Property claims: art. 641 CC

Outside the law of obligations entirely. Art. 641 al. 2 CC gives the owner the revendication (rei vindicatio): the owner can claim the thing back from anyone who withholds it, and repel any unjustified interference. This claim does not care about contracts or fault. It attaches to the thing. If someone holds your laptop, you do not need to show they promised to return it or that they were negligent, only that you own it and they possess it without a right. Possession itself gets separate fast-track protection at art. 927 and 928 CC.

The lookup table

Named contracts add specialized foundations that override and extend the general rulebook: the buyer's warranty claims (art. 197 ss CO), the tenant's remedies for defects (art. 259a ss CO), the principal's rights against a contractor (art. 368 CO). Same method, more specific Tatbestand (see ch. 5).

The discipline: name the foundation before arguing

Here is where careful, systematic readers have an unfair advantage. Untrained people argue legal disputes the way they argue on the internet: by throwing every grievance at once. The trained move is boring and devastating: refuse to engage until the foundation is named.

When a demand arrives, run through these steps:

1. Extract the four slots. Who, what, from whom, based on what. If the letter never names a foundation, your first reply is a single question: "Sur quelle base juridique fondez-vous cette prétention?" (On what legal basis do you found this claim?) You are not being difficult. You are asking which rule they intend to invoke.
2. Check the named foundation's element checklist. Every foundation is a checklist on which every condition must hold (Chapter 1's move). List the elements. Any element that fails, the claim fails, regardless of how strong the others are.
3. Map each element to its burden. Who must prove it under art. 8 CC? Elements of the claim fall on the claimant; facts that destroy or paralyze the claim fall on the defender (see ch. 3). An element the claimant cannot prove is treated as absent.
4. Only then discuss facts. Now the argument has a shape. You are not debating "was the apartment clean enough," you are debating "can the landlord prove element 3 of their art. 97 CO claim."

The analogy, and where it breaks: the foundation norm is like a required label on an official form. An unlabeled demand does not get processed. (If you come from software: the foundation is a required type annotation, and an unannotated demand does not compile.) But unlike a strict form-checker, a court will sometimes fix the label for you: under the principle iura novit curia (the court knows the law, reflected in art. 57 CPC), you plead facts and the judge applies the correct legal foundation even if you cited the wrong article. So sloppy labeling is not always fatal in court. Out of court, though, in the negotiation phase where most disputes actually live, the party who cannot name their foundation is signaling they have none, and everyone can see it.

Claims chain: the tree, not the line

A real dispute is rarely one claim. It is a tree of claims, each with its own foundation, each needing its own check.

Take a client who does not pay Ekeko's CHF 30,000 invoice due March 1:

1. Main claim: CHF 30,000. Foundation: the contract clause fixing the fee (art. 19 CO). This claim exists from the moment the work was performed as agreed.
2. Default interest: 5% per year. Foundation: art. 104 al. 1 CO. But this claim has its own extra element: the debtor must be in default (demeure), which under art. 102 al. 1 CO normally requires a reminder (interpellation), unless a fixed due date was agreed (art. 102 al. 2 CO: dies interpellat pro homine, the deadline itself puts the debtor in default). Contract said "payable by March 1"? Interest runs from March 2 with no reminder needed. Contract silent on the date? No interest until you send the reminder. One sentence in your contract template changes when a whole branch of the claim tree activates.
3. Additional damages. Foundation: art. 106 al. 1 CO lets the creditor claim damage exceeding the default interest, and art. 103 al. 1 CO makes the debtor in default liable for damage caused by late performance. Say the missing CHF 30,000 forced you into an expensive bridge loan: that delta is a further claim, with its own elements (damage, causation) to prove.
4. Enforcement costs. Once you start debt collection, art. 68 al. 2 LP puts the poursuite costs on the debtor by allowing you to recover them from the payments.

Each node in the tree stands or falls on its own foundation and its own elements. Opposing counsel can concede the trunk and fight a branch: yes we owe the fee, but interest only runs from your reminder in June, and the bridge-loan damage is not causally linked. Seeing claims as a tree stops you from treating "they owe me" as one yes/no question when it is really a set of independently verifiable assertions.

Note also that every claim node carries an expiry clock: prescription (art. 127 ss CO, ten years by default, five for periodic payments and some others per art. 128 CO). A foundation whose clock has run out still exists but can no longer be enforced if the debtor raises the exception (see ch. 6).

The demo: Florissant, CHF 4,700, foundation not found

Now the case you lived. A tenant leaves an apartment. The régie, Gerofinance, sends a demand: CHF 4,700 for remise en état. Scary number, official letterhead. Run the method.

Slot 1 to 3: Landlord (via agent) wants CHF 4,700 from former tenant. Fine.

Slot 4: based on what? The candidate foundation is the tenant's restitution duty: art. 267 al. 1 CO, the tenant must return the thing in the state resulting from use in conformity with the contract. Normal wear, usure normale, is by definition conforming use, so it grounds nothing. A damages claim for anything beyond normal wear runs through art. 97 CO on top of art. 267 CO, with elements: damage exceeding normal wear, attributable to the tenant, quantified, and, crucially, notified. Art. 267a al. 1 CO requires the landlord to verify the state of the thing on return and immediately notify the tenant of defects the tenant answers for; defects not notified in time are waived (art. 267a al. 2 CO, subject to the hidden-defects carve-out in al. 3).

Now check the elements against the actual file:

The demand's real foundation was an unsigned piece of paper and a contractor's quote. Asked "based on what," the claim had no answer. Not a weak answer, no answer. That is why the dispute was winnable by a non-lawyer: it never required out-arguing anyone on the law. It required noticing that slot 4 was empty and refusing to argue slots 1 to 3 until it was filled. When the poursuite arrived anyway, the opposition (art. 74 LP) flipped the procedural burden back where the substantive one already sat: prove your foundation in court or drop it. They dropped it. Enforcement mechanics are Chapter 9's territory (see ch. 9); the point here is that the magic question decided the case before procedure ever mattered.

The general lesson from Florissant: a document is not a foundation. Letters, estimates, invoices, and unsigned conventions are evidence at best, and evidence only matters once a valid foundation norm gives it something to prove. People send scary paper precisely because most recipients never ask what norm stands behind it.

Using the question offensively

The question cuts both ways. Before you demand anything, as Ekeko will, run yourself through the same gate:

• Foundation check: which clause or article produces my claim? Write it in the demand letter. "En vertu de l'art. 3 du contrat du 12 janvier, et de l'art. 104 CO pour les intérêts" reads like someone who will follow through, because naming the foundation is costly signaling: it shows you know which elements you would have to prove.
• Element check: can I tick every box of that foundation with documents I already hold? If not, the gap tells you exactly what to collect before demanding, not after.
• Chain check: what does my full claim tree look like? Main claim, interest from when, extra damages, costs. Claim the whole tree; opponents rarely volunteer branches you forgot.

This is also why contract drafting is claim engineering. Every clause you write is a future foundation norm. A fixed payment date is not bureaucratic decoration, it is the element that makes art. 102 al. 2 CO kick in automatically and starts the interest clock without a reminder. Draft with the magic question in mind: if this clause is ever slot 4 of a demand, will its elements be provable?

The move in one line

Never argue a demand, yours or theirs, until slot four is filled: name the norm that produces the obligation, then check its elements, and if no norm answers "based on what," the conversation is already over.

Drills

1. Your neighbor's contractor accidentally cuts your fiber line. You have no contract with the contractor. You want the repair cost of CHF 800. Fill the four slots and name the foundation. What must you prove?

Answer: Who: you. What: CHF 800 damages. From whom: the contractor (not the neighbor, absent a specific basis against them). Based on what: art. 41 al. 1 CO, tort, since no contract links you. Elements you must prove under art. 8 CC: damage (CHF 800, invoice), unlawfulness (interference with your property), natural and adequate causation (their cut caused the outage and repair), and fault (negligence), which in tort is not presumed. Contrast with art. 97 CO where fault would be presumed against the debtor.

2. A client pays Ekeko's invoice twice by mistake. Three foundations could superficially seem relevant: the contract, art. 97 CO, art. 62 CO. Which one actually grounds the client's claim to get the second payment back, and why not the others?

Answer: Art. 62 CO, unjust enrichment (with art. 63 al. 1 CO: the payment was made in error and without debt, since the debt was already extinguished by the first payment). The contract cannot be the foundation because the contract only obliged one payment; the second transfer sits outside it. Art. 97 CO fails because receiving an accidental overpayment breaches no contractual obligation. Value moved with no valid cause: that is exactly the enrichment Tatbestand.

3. Ekeko's contract says "invoice payable within 30 days of issuance." The client is silent at day 45. From when does 5% interest under art. 104 CO run, and what extra step, if any, is needed? What if the contract had said nothing about payment timing?

Answer: With "payable within 30 days of issuance," a determinable due date was agreed, so under art. 102 al. 2 CO the debtor falls into default when that day passes, no reminder needed: interest at 5% (art. 104 al. 1 CO) runs from day 31. If the contract were silent, the obligation is due immediately (art. 75 CO) but default still requires an interpellation under art. 102 al. 1 CO: interest only runs once you send a reminder demanding payment. Same main claim, different activation condition on the interest branch of the claim tree.

Exercises

Self-test

Fieldwork

• Drafting · ~15 minWrite, in two sentences and no more, the reply to a foundation-less demand: the question “Sur quelle base juridique fondez-vous cette prétention?” plus a polite closing. The exercise is practicing what you leave out — no facts, no justifications, no indignation. You argue nothing until slot 4 is filled.
• Fedlex · ~30 minBuild your own copy of this chapter’s lookup table. For each of the five foundation families — the contract clause (art. 19 and 82 CO), art. 97 CO, art. 41 CO, art. 62 CO, art. 641 al. 2 CC — open the article on Fedlex and copy the operative sentence verbatim into your notes, in French and one other official language. You will reach for this table for years.

What this chapter gives you. The single most decision-relevant rule in Swiss private law is not about rights. It is about facts. Art. 8 CC decides which party loses when a fact cannot be established, and in real disputes facts are missing far more often than law is unclear. This chapter gives you the allocation rule (elements vs exceptions), the three degrees of proof courts actually apply, the practical ranking of evidence types, the ways contracts can pre-load the burden in your favor, and a full worked example: the four-rung ladder that won the Florissant deposit dispute. If chapter 2 taught you to find the claim-foundation norm (see ch. 2), this chapter tells you which of its elements are your problem to prove.

The rule and why it exists

Art. 8 CC, in one sentence: unless the law provides otherwise, each party must prove the facts they allege in order to derive a right from them (fardeau de la preuve, burden of proof).

Think of it as the tie-breaking rule of a system that must always produce a decision. A court is not allowed to answer "we don't know." When the evidence phase ends and a fact is neither established nor refuted (the situation lawyers call non liquet, "it is not clear"), the judge still has to decide. Art. 8 CC is the default answer assigned to every missing fact: an unprovable fact is treated as false for whoever needed it to be true. (If you come from software: art. 8 CC is the default value substituted for missing data — the system cannot return null, so it returns "false" against the party who needed "true".) That is the whole rule. It decides who absorbs the loss when the facts stay unknown, and it is brutally asymmetric: the same objective reality produces opposite judgments depending on who carried the burden.

Two consequences follow immediately:

1. Burden allocation is worth more than eloquence. Before arguing about what happened, figure out who must prove what. Often the dispute is over before it starts.
2. Evidence generation is a design activity. The time to win a proof battle is when the facts are being created (contracts signed, states of premises recorded, emails sent), not when the dispute arrives. More on this below.

One honest caveat about the picture: art. 8 CC governs the risk of unproven facts, not who must physically produce evidence at trial. Under the CPC both parties offer evidence and the judge weighs everything freely (art. 157 CPC, libre appréciation des preuves). The burden only bites at the end, when something remained unproven. It is a tie-breaker, not a turn order.

The allocation rule: elements vs exceptions

How do you know which facts are "yours"? The standard reading of art. 8 CC gives a clean two-branch rule:

The rule, as a plain decision routine:

For each disputed fact:
    is it an element of the claim-foundation norm?      → the claimant must prove it
    is it an element of a defense or exception norm?    → the defendant must prove it
    finally: does a statutory presumption override this default?

That last line matters. The opening words of art. 8 CC ("unless the law provides otherwise") are a hook for explicit overrides scattered through the codes. Three worth knowing:

• Good faith is presumed (art. 3 al. 1 CC). Whoever alleges bad faith proves it.
• Fault is presumed in contract breach (art. 97 al. 1 CO). The creditor proves breach and damage; the debtor must prove absence of fault to escape. Compare tort (art. 41 CO), where the victim proves fault as an element. Same word "fault", opposite allocation. This is why the choice of claim-foundation norm in chapter 2 is not academic.
• Public registers and public deeds provide full proof of the facts they attest, unless shown inaccurate (art. 179 CPC).

Note the structural elegance: the legislator allocates burdens by drafting position. If a fact appears in the main rule, the claimant carries it. If it appears in a proviso ("unless", "except where", a separate paragraph starting with "however"), the defendant carries it. Reading an article as an element checklist (chapter 1, see ch. 1) therefore also reads out the burden map for free. Tatbestand structure is burden structure.

Degrees of proof: three thresholds, not one

"Prove" is not binary. Swiss law runs three confidence thresholds, set by case law rather than statute:

Practical takeaway: in ordinary contract litigation you should assume certitude is required for your elements and plan your evidence accordingly. The reduced thresholds are relief valves the court grants where strict proof is structurally impossible, not a discount you can request because your paperwork is thin.

What counts as evidence, and how it ranks in practice

Art. 168 al. 1 CPC gives the closed list of admissible evidence: witness testimony (témoignage), documents (titres), inspection (inspection), expert opinion (expertise), written information (renseignements écrits), and examination of the parties (interrogatoire et déposition des parties). Legally, none of these outranks another: the judge weighs everything freely (art. 157 CPC). In practice, there is a stable informal hierarchy any litigator will confirm:

1. Contemporaneous signed documents. A document signed by your opponent at the time of the events is close to unbeatable. If they contest authenticity they must give specific grounds, and the party relying on the document then proves it is genuine (art. 178 CPC), which for an ordinary signed original is usually easy.
2. Contemporaneous unsigned records. Emails, photos with metadata, delivery receipts, bank statements. Strong, because they predate the dispute and were not created to win it.
3. Expert opinion (expertise). Decisive on technical questions (was this defect wear or damage, what would repair cost), but expensive and slow, and the court appoints the expert; a party-commissioned report is treated as mere party allegation.
4. Witnesses. Human memory, examined years later, filtered through loyalty. Courts use them and courts distrust them.
5. Party statements. You asserting your own case. Admissible (art. 168 CPC), nearly weightless when self-serving.

The working principle behind the ranking: the value of evidence decays with the distance between fact creation and record creation, and rises with the opponent's involvement in creating the record. A document your opponent signed at the time is both dated by events and counter-signed by the adverse party. Everything you do at contract time should aim to manufacture rung 1 and rung 2 material for the facts you know you will carry under art. 8 CC.

Shifting the burden by contract, and the limits

Art. 8 CC opens with "unless the law provides otherwise", and the dominant view treats the allocation itself as dispositive between the parties (see ch. 4 for the dispositive vs mandatory taxonomy): contracts can reallocate proof risk. This happens constantly, usually in disguise. Three mechanisms:

1. Contractual presumptions and acknowledgments. "The client's acceptance of the deliverable is deemed given if no defects are notified within 10 days." "Signature of this delivery note confirms receipt in good condition." These clauses do not change reality; they change who must prove the deviation from the recorded state. For your Ekeko consulting contracts this is the cheap, high-leverage move: acceptance clauses, written-form-for-changes clauses, and deliverable sign-off protocols are all burden-shifting instruments dressed up as process.

2. Documents that double as enforcement titles. A signed, unconditional acknowledgment of a specific debt (reconnaissance de dette) unlocks provisional lifting of an LP opposition (mainlevée provisoire, art. 82 LP) and, via the action en libération de dette (art. 83 al. 2 LP), flips the roles: the debtor must sue within 20 days and appear as plaintiff. The burden of moving first, paying court advances, and carrying litigation risk migrates with the paper. Signing one document reroutes the entire procedural path (see ch. 9).

3. Agreed evidence protocols. Jointly established records at defined checkpoints. The état des lieux, next section, is the canonical one.

The limits, because the system does police this:

• Mandatory substantive law cannot be laundered through burden clauses. In residential leases, agreements made before the end of the lease obliging the tenant to pay anything other than actual proven damage are void (art. 267 al. 2 CO), and derogations from the defect rules to the tenant's detriment are void (art. 256 al. 2 CO). A clause "tenant is presumed liable for all marks on walls" fails.
• Unfair terms control. In consumer general terms and conditions, clauses creating a significant and unjustified imbalance are unlawful (art. 8 LCD). Aggressive burden reversals in fine print are prime targets.
• Good faith backstop. Art. 2 CC catches manifestly abusive invocations of formal positions (chapter 7's safety valve).

So: burden shifting by contract is real, powerful, and bounded exactly where the law tags the underlying norms mandatory or semi-mandatory.

The état des lieux: a burden-shifting instrument done right or wrong

Tenancy law is the cleanest live demo, and you have lived it. The mechanics:

• At the end of the lease, the tenant must return the premises in the state resulting from use in conformity with the contract (art. 267 al. 1 CO). Normal wear (usure normale) is the landlord's cost, priced into rent. Damage beyond it is the tenant's.
• The landlord wants money, so under art. 8 CC the landlord proves every element: a defect exists, it exceeds normal wear, it is attributable to the tenant, and the amount.
• On top of that sits a guillotine: the landlord must inspect the premises on return and immediately notify the tenant of defects for which the tenant is answerable (art. 267a al. 1 CO); silence forfeits the claim except for defects not discoverable on ordinary inspection, which must be notified upon discovery (art. 267a al. 2 and 3 CO).

Now watch how the paperwork moves the burden.

Done right (from the landlord's perspective): a detailed état des lieux d'entrée (entry inspection record) signed by both parties, and at exit a convention de sortie (exit agreement) in which the tenant signs an itemized list of defects with amounts and an acknowledgment of owing them. That signature does three jobs at once: it defeats the "it was already like that" defense by comparing against the signed entry record, it satisfies the art. 267a CO immediate-notice requirement, and it manufactures a reconnaissance de dette good for mainlevée provisoire (art. 82 LP). The landlord converted a claim they would have had to prove element by element into a document the tenant must now affirmatively fight their way out of.

Done wrong: no signed entry record, an exit inspection noting usure normale, a convention de sortie the tenant declined to sign, invoices produced weeks later. Every burden now sits exactly where art. 8 CC put it, on the landlord, with no document to carry it. Refusing to sign an exit convention is not obstruction; it is declining to hand your counterparty the burden-shifting instrument. The law never obliged you to sign it.

The general lesson exceeds tenancy: whenever a counterparty pushes a document at you at a transition point (delivery, acceptance, exit, termination), ask what burden it shifts before signing. Symmetrically, when you are the one who will carry burdens later (you will, as Ekeko, when invoicing for delivered work), build your own checkpoint documents and get the counter-signature.

Worked example: the Florissant four-rung ladder

The deposit dispute, reconstructed as a burden ladder. Gérance claims CHF 4,700 from the departing tenant for end-of-lease damage. Each rung is a question; at each rung, ask who carries the proof risk and with what evidence.

Rung 1: Who wants what from whom based on what? The landlord wants CHF 4,700 based on art. 267 al. 1 CO (return obligation) combined with art. 97 CO. The tenant wants the deposit released (art. 257e CO). The landlord is the claimant on the damage claim, so per art. 8 CC the landlord carries the elements: defect, beyond usure normale, attributable, quantified.

Rung 2: The element check against the record. The exit état des lieux itself recorded usure normale. The landlord's own contemporaneous document contradicts the first element. Under the informal hierarchy above, that record outranks any later-produced invoice or estimate. Worse for the landlord: art. 267a al. 1 CO required immediate notice of tenant-attributable defects, and a record noting normal wear is the opposite of such notice, so the claim is not just unproven, it is arguably forfeited.

Rung 3: The missing signature. The convention de sortie was never signed. Therefore no reconnaissance de dette, therefore no mainlevée provisoire under art. 82 LP. The one instrument that could have shifted the procedural burden onto the tenant does not exist. The refusal to sign, weeks earlier, was the decisive move of the whole dispute, made before any lawyer was involved.

Rung 4: Enforcement mechanics as burden amplifier. Suppose the landlord starts debt enforcement anyway (commandement de payer, art. 69 LP; poursuite can be launched without any proof). The tenant files opposition within 10 days (art. 74 al. 1 LP), one signed line, no reasons needed. Now the landlord must either obtain mainlevée (impossible without a title, art. 80 and 82 LP) or sue on the merits (art. 79 LP) and prove, to the certitude standard, every element from rung 1, against their own exit record, with court advances paid up front. The expected value of that lawsuit is negative (see ch. 12), which is why such claims quietly die after opposition.

Read the ladder bottom-up and the pattern is stark: at no point did the tenant prove innocence. The tenant never carried a burden. The entire defense consisted of declining to sign, pointing at the opponent's own record, and filing a one-line opposition. Art. 8 CC did the rest, treating every unprovable fact as false for the party who needed it.

The move in one line

Before arguing about what is true, work out who loses if it stays unknown: an unprovable fact is false for whoever needed it, so allocate burdens first and manufacture counter-signed contemporaneous documents for the facts that are yours to carry.

Drills

1. Your Ekeko client refuses to pay an invoice, claiming the delivered model "never worked as promised". You sue for the fee. List who must prove what, with articles.

Answer: You are the claimant on the fee claim (art. 82 CO context, claim founded on the contract): you prove the contract, its content, and your performance (art. 8 CC). "It never worked" is the client's story, and how it plays depends on its legal shape: as a defect objection under warranty-type rules or as a breach defense, the client carries the facts establishing the defect. If the client counterclaims damages under art. 97 CO, the client proves breach, damage, and causation, while you would carry absence of fault (art. 97 al. 1 CO, presumed fault). Practical prophylaxis: a signed acceptance protocol at delivery moves the whole fight onto rung 1 evidence you control.

2. Why did refusing to sign the convention de sortie matter more than the usure normale notation itself?

Answer: The notation is strong rung 2 evidence against the landlord's elements, but evidence can in principle be outweighed under free assessment (art. 157 CPC). The unsigned convention removes an entire procedural weapon: without a signed acknowledgment there is no reconnaissance de dette, so no mainlevée provisoire (art. 82 LP), so after a simple opposition (art. 74 LP) the landlord must sue as plaintiff (art. 79 LP) carrying full burdens at full cost. The signature would have inverted plaintiff and defendant roles regardless of the underlying facts; its absence locked the burden where art. 8 CC placed it.

3. A clause in your consulting GTC states: "Deliverables are deemed accepted unless the client notifies specific defects in writing within 10 business days of delivery." Is this a burden shift, and is it valid?

Answer: Yes, it is a burden-shifting device: after the window closes, the client can no longer simply deny conformity; the client must prove timely notice or facts escaping the fiction, while you point at the deemed acceptance. Between businesses it is generally valid because the underlying rules are dispositive (art. 8 CC's allocation and the CO acceptance rules admit contrary agreement; compare art. 370 CO's own acceptance fiction in works contracts). Limits: against consumers in GTC it must survive art. 8 LCD unfairness control, an unreasonably short window risks censure, and invoking it against a defect you actively concealed collides with good faith (art. 2 CC). See ch. 4 for the general validity test.

Exercises

Self-test

Fieldwork

• Drafting · ~45 minWrite the acceptance-protocol block for your own consulting template: deliverables list, verifiable acceptance criteria, a 10-business-day deemed-acceptance window, a signature line. Then audit your draft against the three limits from this chapter: does it collide with any mandatory rule, would it survive art. 8 LCD in a consumer setting, and could invoking it ever look like bad faith (art. 2 CC)?
• Your own files · ~30 minEvidence generation is a design activity, so design some: locate your current apartment’s entry EDL, your photos from move-in day, and the deposit account statement. Anything missing? Create the rung-2 record now — dated photos, a short email to the régie fixing a fact in writing. The dispute you are preparing for does not exist yet; that is the point.

What this chapter gives you. Move 4 of the seven: before you rely on any clause, or on any article of the CO, you ask one question: is this rule overridable? Swiss private law sorts every norm into three precedence tiers: dispositive (droit dispositif, the default that your contract can replace), semi-mandatory (droit semi-impératif, overridable in one direction only), and mandatory (droit impératif, not overridable at all). Get the tier wrong and you either draft clauses that are dead on arrival, or you concede rights you never had to concede. This chapter teaches you to read the tier off the text, walks through the classic mandatory traps that bite consultants and founders, and ends with a clause-audit checklist you can run on any contract, including your own Ekeko templates.

The three tiers, as a precedence system

Think of the CO as a book of defaults. Your contract is a layer of choices written on top of it. The question is which of the underlying rules are locked. (If you come from software: the CO is the default config, your contract is the user config layered on top, and some keys are locked.)

The base grant of override power is freedom of contract: parties may set the content of their agreement freely within the limits of the law (art. 19 al. 1 CO). The limits: a contract may deviate from statutory rules only where the statute does not lay down an inalterable rule or where deviation would not offend public policy, morality, or personality rights (art. 19 al. 2 CO). A contract whose content is impossible, unlawful, or immoral is void (art. 20 al. 1 CO), and if the defect affects only parts of it, only those parts fall, unless the deal would not have been done without them (art. 20 al. 2 CO, partial nullity).

That last rule matters for drafting: Swiss law usually does surgical excision, not demolition of the whole contract. A void liability clause dies alone; the statute's default takes its slot. The failure mode is silent: nobody tells you at signature time. You find out at dispute time, which is exactly when you wanted the clause to work.

Where the picture breaks down: none of this is checked at the moment of signing. Tier membership is sometimes written in the statute, sometimes established only by case law (art. 404 CO is the famous example below), and semi-mandatory rules require you to know who the protected party is. Nothing warns you while you draft; the courts do the checking, after the fact, once a dispute is already live. (If you come from software: there is no linter. The case law is the linter, and it runs in production.)

How to read the tier off the text

Signals, in rough order of reliability:

1. Explicit nullity language. "Est nulle toute stipulation..." (any stipulation is void) or "les dérogations... sont nulles" (derogations are void). This is the statute marking its own rule as locked. Example: art. 100 al. 1 CO on liability exclusions, art. 199 CO on warranty disclaimers obtained by fraud.
2. One-directional nullity. "Il ne peut être dérogé... au détriment du travailleur/locataire" (no derogation to the detriment of the worker/tenant). That is the semi-mandatory signature: a floor for the weak party, open sky above it. Example: art. 362 CO (employment), art. 256 al. 2 CO (lease).
3. Explicit tier lists. Employment law is unusually kind to systems thinkers: art. 361 CO is a literal enumeration of the fully mandatory provisions (neither party may derogate), and art. 362 CO enumerates the semi-mandatory ones (no derogation to the employee's detriment). Two ready-made lists, printed in the statute itself.
4. "Sauf convention contraire" and cousins. "Unless otherwise agreed", "à défaut de convention", "sauf usage ou convention contraire": the statute announcing its own default status. Dispositive.
5. Silence. No signal at all. Default assumption in the CO general part and in commercial contracts between professionals: dispositive. But check case law before betting money, because tier 6 exists:
6. Judge-made mandatory character. Some articles carry no textual marker and are mandatory anyway because the Tribunal fédéral says the rule's purpose demands it. Art. 404 CO is the canonical trap, next section.

Structural heuristic: the closer a contract type sits to a protection regime, the more locked rules it has. Employment (art. 319 ss CO), residential lease (art. 253 ss CO, plus art. 271 ss CO on abusive rents and termination protection), consumer credit, package travel: dense with mandatory and semi-mandatory rules. Sales and mandate between two businesses: mostly dispositive, with a few sharp exceptions. You lived this asymmetry in the deposit dispute: tenancy law kept tilting the table your way by design, not luck.

The classic traps

These are the rules people draft against, in good faith, and lose.

Art. 404 CO: the mandate is always revocable

The mandate (mandat) may be revoked or terminated by either party at any time (art. 404 al. 1 CO). The only sanction for doing it at an inopportune moment is damages for the harm caused by the timing (art. 404 al. 2 CO), not performance of the rest of the contract, and per case law not the lost profit on the unperformed remainder.

The trap: nothing in the text says "mandatory". Yet the Tribunal fédéral holds art. 404 CO mandatory (ATF 115 II 464): you cannot contract it away, not with a fixed term, not with a termination penalty that economically forces continuation. And because many service contracts (contrats de services) are classified as mandates or mandate-heavy innominate contracts (see the composition logic in ch. 5), this reaches further than "agency agreements". For Ekeko this is personal: a consulting agreement is typically a mandate. A clause saying "12-month engagement, non-cancellable" or "early termination fee of 50% of remaining fees" is at serious risk of being void under art. 404 CO if the contract is a mandate. Mitigations exist (structuring around work-product deliverables so contrat d'entreprise rules under art. 363 ss CO apply, milestone-based payment so revocation costs the client the milestones already earned, genuine timing-damage clauses), but "we agreed it's binding" is not one of them.

Art. 100 and 101 CO: the liability-cap floor

Any advance exclusion of liability for unlawful intent (dol) or gross negligence (faute grave) is void (art. 100 al. 1 CO). That is the floor under every limitation-of-liability clause in every Swiss contract. "Liability excluded to the maximum extent permitted by law" translates to: liability for slight negligence excluded, liability for gross negligence and intent fully intact. The judge may additionally strike exclusions even for slight negligence where the releasing party was in the other's service, or the liability arises from a business conducted under official license (art. 100 al. 2 CO).

The subtlety worth knowing: liability for auxiliaries (auxiliaires, your subcontractors and employees) under art. 101 CO can be excluded more broadly, in principle even for their gross fault (art. 101 al. 2 CO), except again for licensed industries (art. 101 al. 3 CO). So a well-drafted cap distinguishes your own fault from your subcontractors' fault. Most templates copied from common-law forms do not, and common-law "consequential damages" waivers map poorly onto Swiss categories anyway. Audit any LoL clause with this pair of articles open.

Art. 257e CO: the deposit cap

For residential leases, the security deposit (sûretés) may not exceed three months' rent, and cash or securities must sit in a bank account in the tenant's name (art. 257e al. 1 and 2 CO). The bank releases it only with both parties' agreement, or on a final payment order or enforceable judgment (art. 257e al. 3 CO). Semi-mandatory in structure: the parties can agree on less than three months, never more. Your Gerofinance episode ran on the release mechanics of al. 3: the landlord cannot just take the deposit, it must obtain your signature or a title, which is why the unsigned convention de sortie was load-bearing. Note the scope limit: the three-month cap covers residential premises; commercial leases have no statutory cap, though cantonal rules can add constraints.

Art. 82 LP: fine print does not make a debt enforceable

This one is a procedure-side trap, the mirror image of the others (substance versus procedure: see ch. 9). Provisional lifting of opposition (mainlevée provisoire) in debt enforcement requires a signed acknowledgment of debt, by authentic deed or under private signature (art. 82 al. 1 LP), and case law requires that the signed document evidence an unconditional obligation to pay a determinate, or readily determinable, sum. A beautifully drafted clause saying "client shall indemnify all damages arising from X" creates a right but not a mainlevée title: the sum is not determinate, so at enforcement time the creditor is thrown back into ordinary litigation, and the debtor keeps the opposition alive (the debtor may still escape mainlevée by making a defense plausible, art. 82 al. 2 LP). Drafting consequence for Ekeko invoiceable work: fixed fees, signed offers, signed acceptance of amounts. A signed contract stating "CHF 12,000 payable on delivery" is worth more at enforcement than three pages of indemnity prose. Legal strength is provability times enforceability (see ch. 12), and art. 82 LP is where enforceability gets priced.

Honorable mentions

• Art. 341 al. 1 CO: during the employment relationship and for one month after it ends, the worker cannot validly waive claims arising from mandatory or semi-mandatory provisions. Settlement agreements signed at exit get audited against this.
• Art. 129 CO: the prescription periods of the CO's third title cannot be modified by agreement. Your contract cannot shorten the ten-year default (art. 127 CO) or the five-year period for periodic payments and certain claims (art. 128 CO). Waiver of an already-running prescription is a separate, regulated move (art. 141 CO).
• Art. 199 CO: a clause excluding sales warranty is void if the seller fraudulently concealed the defect. Every "sold as seen" clause carries this hidden exception.
• Art. 27 al. 2 CC: nobody can contractually alienate their freedom to a degree that violates law or morality. The backstop against perpetual exclusivity, lifetime non-competes, and other excessive commitments.
• Art. 2 CC: even a valid clause can be refused effect in a given case if invoking it is a manifest abuse of right. The safety valve that catches what the tier system misses (Chapter 7).

Drafting in the dispositive space

Once you see the tiers, drafting stops being prose and becomes deliberate settings work: defaults, overrides, floors:

1. Enumerate the defaults. For your contract type, list what the CO already provides (the checklist reading from ch. 1). Much of a good mandate contract is already written in art. 394 ss CO for free.
2. Override only where the default hurts you, and only dispositive rules. Every clause restating the statute is noise; every clause fighting a mandatory rule is a latent nullity.
3. Respect the floors. Where a rule is semi-mandatory, drafting energy should go into the open direction, or into scope (is this even a residential lease? is this person even a worker?), because classification, not clause wording, decides which protection regime applies.
4. Draft for the enforcement path, not just the right: determinate sums, signatures, dates (art. 82 LP again).
5. Add severability deliberately. Art. 20 al. 2 CO gives partial nullity by default, but the fallback that replaces a void clause is the statutory default, which may be worse for you than a lawful middle ground. A well-built clause includes its own fallback: "liability capped at fees paid; where and to the extent this cap is invalid, at the lowest cap permitted by law."

Clause-audit checklist

For any clause, yours or theirs, run this:

• [ ] Which statutory rule does this clause displace? Name the article. If you cannot, you do not yet understand the clause.
• [ ] What tier is that rule? Check text signals (nulle, au détriment de), the explicit lists (art. 361 and 362 CO), and known case-law locks (art. 404 CO).
• [ ] If semi-mandatory: which direction does the clause move, and who is protected? Toward the protected party: valid. Away: void.
• [ ] If mandatory: what default snaps back when the clause is voided (art. 20 al. 2 CO), and can you live with it?
• [ ] Does a hidden exception gut it? Fraud (art. 199 CO), gross fault (art. 100 al. 1 CO), abuse of right (art. 2 CC).
• [ ] Is it enforceable, not just valid? Signed, determinate sum, clear due date (art. 82 al. 1 LP), and remember who must prove what if it is contested (art. 8 CC, see ch. 3).
• [ ] Classification check: could a court recharacterize the contract (mandate instead of contrat d'entreprise, employment instead of freelance) and load a different protection regime over your clause?

The move in one line

Before trusting any clause, ask which statutory rule it overrides and whether that rule was ever yours to change: dispositive rules yield, semi-mandatory rules move only toward the protected party, mandatory rules silently revert to the statute at the worst possible moment.

Drills

1. Your Ekeko template says: "This agreement runs for 12 months and may not be terminated early by either party. Early termination triggers a penalty of the full remaining fees." A client cancels in month 3. What do you actually recover?

Answer: If the contract is a mandate, art. 404 al. 1 CO makes it revocable at any time and is mandatory per case law (ATF 115 II 464), so both the no-termination clause and the penalty forcing continuation are at risk of nullity (art. 20 CO). You keep fees for work already performed and, at most, damages for termination at an inopportune moment under art. 404 al. 2 CO, which per case law does not cover lost profit on the unperformed remainder. Structural fixes beat clause fixes: milestone-based deliverables and payment, or genuine contrat d'entreprise scoping.

2. A supplier's terms state: "Liability is excluded for all damage, howsoever caused." Their technician, grossly negligent, fries your GPU cluster. Two sub-questions: does the exclusion hold, and does the answer change if the fryer was a subcontractor?

Answer: Against the supplier's own gross negligence the exclusion is void (art. 100 al. 1 CO): partial nullity (art. 20 al. 2 CO) strikes the clause to the extent it covers dol and faute grave, leaving statutory liability (art. 97 CO) intact. If the damage was caused by an auxiliary under art. 101 CO, liability can in principle be excluded even for the auxiliary's gross fault (art. 101 al. 2 CO), unless the supplier runs an officially licensed business (art. 101 al. 3 CO). Whether this blanket clause is interpreted to cover auxiliary fault at all is a construction question courts read restrictively.

3. A Lausanne landlord demands a five-month deposit for a flat, "agreed by both parties in the lease". Valid? And what stops the landlord from simply withdrawing it after a messy exit?

Answer: Invalid above three months' rent for residential premises (art. 257e al. 2 CO): the rule is semi-mandatory, the parties' agreement moves in the forbidden direction, so the excess is void and reclaimable. On release: the deposit sits in a bank account in the tenant's name, and the bank pays out only with both parties' consent, a final payment order, or an enforceable judgment (art. 257e al. 3 CO). Without the tenant's signature (an unsigned convention de sortie counts for nothing), the landlord must sue and prove damage beyond usure normale, bearing that burden under art. 8 CC (see ch. 3).

Exercises

Self-test

Fieldwork

• Fedlex · ~20 minRead art. 361 and 362 CO — the statute’s own printed tier lists for employment. Pick any three provisions enumerated in art. 362 CO, look each up, and state in one line what the one-directional protection means for an employment contract you know (yours, a friend’s, a standard template).
• Audit · ~45 minRun the clause-audit checklist on a contract you have actually signed — your lease, a client agreement, a supplier’s GTC. Pick three clauses: the liability clause, the termination clause, and the weirdest one you can find. For each, name the statutory rule it displaces, its tier, and what snaps back if the clause is void. Expect at least one surprise.

What this chapter gives you. The Code of Obligations is not a flat list of rules. It is a layered system: a general part that works as one shared rulebook of defaults for every contract (arts. 1 to 183 CO), a special part of named contracts that override those defaults with their own specialized rules (vente, bail, travail, entreprise, mandat), and a large open world of innominate contracts built by composing the pieces. Once you see this structure, two practical skills follow: you can find the applicable rule fast (start from the specialized rules, fall back to the general ones), and you can predict how a court will treat a contract that does not fit any named box (qualification, then composition). The chapter closes with the qualification that will matter most for Ekeko: is your AI consulting gig a mandat or a contrat d'entreprise, and why the difference is worth real money.

The general rulebook: CO arts. 1 to 183

The general part of the CO (Partie générale, arts. 1 to 183 CO) defines what every obligation gets for free, regardless of contract type. Think of it as the standard equipment every contract carries before any specialized rule touches it. (If you come from software: this is a base class with subclasses overriding its defaults.)

Two consequences of "one shared rulebook" thinking:

1. Every contract ultimately runs on these. An NDA, a SaaS subscription, a handshake deal for a logo: formation, error, breach damages, and prescription all resolve through the general part unless a special rule overrides them. This is why the general part is the highest-leverage 183 articles in Swiss private law.
2. The general part is mostly dispositive (see the precedence tags in ch. 4): parties can override many defaults in their contract. The special part is where the mandatory and semi-mandatory overrides cluster, because that is where the legislator decided one party typically needs protection.

The picture's honest limit: nothing checks a contract for internal coherence at the moment of signing. A contract can be internally contradictory, and nothing breaks until someone litigates. The real consistency check happens at dispute time, by a judge, using art. 18 CO.

The specializations: five named contracts

The special part (Partie spéciale, art. 184 CO onward) defines the contrats nommés: named types with their own element checklists and their own overrides. One-line definitions for the five you will actually meet:

What "override" means concretely, with examples:

• Vente overrides the general breach regime with a specialized warranty for defects (garantie pour les défauts, arts. 197 to 210 CO), including a duty to inspect and notify promptly (art. 201 CO) and a short prescription of two years (art. 210 al. 1 CO). Miss the notification and your general art. 97 CO instincts will not save you.
• Bail is dense with semi-mandatory tenant protection: challenge of abusive rent (arts. 269 ff. CO), protection against termination (arts. 271 ff. CO), and the deposit cap of three months' rent for housing (art. 257e al. 2 CO). You met this specialized regime in your deposit dispute.
• Travail carries explicit lists of provisions that cannot be waived at all (art. 361 CO) or not to the employee's detriment (art. 362 CO). This is the most protective of the named contracts.
• Entreprise overrides breach with its own defect machinery: inspection and notice after delivery (art. 367 CO), a menu of remedies (repair, price reduction, rescission, art. 368 CO), acceptance effects (art. 370 CO), and prescription of two years from delivery for movable works, five years for works integrated into an immovable (art. 371 CO).
• Mandat carries the most famous mandatory override in the special part: either party may terminate at any time (art. 404 al. 1 CO), and the Tribunal fédéral holds this rule mandatory even between sophisticated businesses (ATF 115 II 464). You cannot contract your client into staying. Only termination "en temps inopportun" triggers damages (art. 404 al. 2 CO), and those cover the negative interest, not your lost profit on the remaining engagement.

And one routing rule that makes mandat the designated fallback: services that fit no other named contract are governed by the rules of mandate (art. 394 al. 2 CO). Mandat is the catch-all bucket at the bottom of the services sorting machine: whatever no other named type claims lands there.

Qualification: how the court decides which specialized regime you are in

You do not choose your contract type by naming it. The court qualifies the contract from what the parties actually agreed and did, using real intent under art. 18 CO. Substance over label: art. 18 al. 1 CO says the judge looks at "la réelle et commune intention des parties" without stopping at inexact designations, even ones used deliberately to disguise the deal (simulation, art. 18 CO).

The practical loop a court runs:

1. Extract the obligations. What did each side actually promise? Ignore the title of the document.
2. Match against the named-contract checklists. Each named type is a Tatbestand like any other article (see ch. 1): if the essential elements of vente are present, vente rules apply.
3. If one type matches, apply its regime, including its mandatory overrides. Calling an employment relationship a "freelance mandate" does not switch off arts. 361 and 362 CO if the subordination, integration, and economic dependence of employment (art. 319 CO) are present in fact. Courts and social insurance authorities requalify sham independence routinely.
4. If no single type matches, go to composition (next section).

Why parties fight about qualification: the label decides which cargo of rules arrives with it. Qualify as travail and you get notice periods, holiday pay, and social insurance. Qualify as entreprise and the client gets defect remedies plus your promised result. Qualify as mandat and either side can walk at any time (art. 404 CO). Same facts, very different rule sets. Qualification is the routing step that decides which rule set your dispute will be decided under.

Innominate and mixed contracts: composition

Freedom of contract (art. 19 al. 1 CO) means you can build contracts the legislator never named: contrats innommés. Modern commercial life mostly runs on them. Three patterns:

• Contrats sui generis: genuinely new types with their own case-law-built regime. Licensing is the classic: no CO chapter, rules assembled by courts and doctrine.
• Contrats mixtes (mixed contracts): one contract combining elements of several named types. A SaaS agreement typically mixes license-like use rights, mandat-like support services, sometimes entreprise-like implementation work, and bail-like ideas about continuous availability.
• Contrats composés (linked contracts): several distinct contracts bundled into one economic operation, each keeping its own regime but standing or falling together.

How courts fill the rule gaps for these, in resolution order:

1. The contract itself. Party agreement governs everything dispositive (art. 19 al. 1 CO). For innominate contracts this is the main rulebook, which is why NDAs and SaaS terms are long: they are writing the special part the CO never wrote.
2. Named-contract rules applied by analogy, per obligation. For mixed contracts the dominant Swiss approach is to route each disputed question to the named type whose element it resembles. Support outage in a SaaS deal: reason from mandat diligence (art. 398 al. 2 CO). Botched custom implementation: reason from entreprise defect rules (arts. 367 ff. CO). The mandatory rules of a named type reach into a mixed contract when the protected element is present: you cannot escape art. 404 CO by wrapping a pure mandate inside a fancier label if the mandate element dominates the disputed question.
3. The general part, which always applies as the shared rulebook underneath: formation, breach under art. 97 CO, prescription under arts. 127 and 128 CO.
4. Art. 2 CC good faith as the safety valve of last resort (see ch. 7).

The conflict rule inside this stack is lex specialis: when a special-part rule and a general-part rule both cover the situation, the specific rule replaces the general one for its scope, and the general part fills everything the special rule does not address. Sale's two-year defect prescription (art. 210 CO) displaces the ten-year default (art. 127 CO) for defect claims, but formation of that same sale still runs on arts. 1 ff. CO. Override of particular defaults, not replacement of the whole rulebook.

Worked example: qualifying an AI consulting gig

Ekeko signs a client. The statement of work says: "Consultant will build and deliver a working document-classification pipeline meeting the accuracy targets in Annex A, acceptance upon successful evaluation on the client's holdout set. Fixed price CHF 40,000." Mandat or entreprise?

Run the checklists.

Mandat (art. 394 CO): services rendered diligently in the principal's interest, obligation de moyens: you promise careful effort, not an outcome. A doctor, a lawyer, a strategy consultant advising on architecture: mandat.

Entreprise (art. 363 CO): you promise an ouvrage, a result, obligation de résultat. The Tribunal fédéral accepts immaterial results as ouvrages when the outcome is objectively measurable against the contract. Software developed to specification is the standard modern example.

The discriminating feature is the acceptance-of-deliverable criterion. Ask: did the parties define a result precisely enough that a third party could verify, yes or no, whether it was achieved, and did they build in delivery and acceptance of that result? Annex A accuracy targets plus a holdout-set evaluation is exactly that. This gig is entreprise, whatever the invoice says.

Now feel the difference in consequences:

Practical drafting moves that fall out of this, for Ekeko:

• Decide your type on purpose, then draft consistently. If you want mandat treatment (advisory, R&D, exploratory model work where outcomes are genuinely uncertain), write obligations of effort: "advise", "investigate", "iterate", time-based billing, no acceptance gate. If you want entreprise (you are confident in the result and want art. 377 CO's full indemnity if the client bails), define the ouvrage sharply and put acceptance criteria in an annex. A vague hybrid gives the qualification decision to a judge later.
• In entreprise mode, love the acceptance protocol. Written acceptance starts the defect clocks (arts. 367, 370, 371 CO) and triggers payment (art. 372 CO). It is a dated, signed, unalterable marker that the state of the work at that moment was approved. (If you come from software: written acceptance is your git tag.)
• Remember what you cannot draft away. If the disputed element is mandate-like, art. 404 CO applies regardless of a "non-cancellable, 12-month minimum" clause. Structure long engagements as entreprise milestones or accept the walk-away risk and price it in.
• Mixed gigs are normal. A typical AI engagement is advisory (mandat) plus build (entreprise) plus maintenance (mandat again, or a duration element). Split them into separately priced work packages so each disputed question routes cleanly. Composition beats one blurry blob.

Where the burden of proof lands in each branch, and who must prove diligence versus defect, is Chapter 3's territory: see ch. 3. How each qualification changes your settlement leverage is Chapter 12's.

The move in one line

Find the specialized regime first, because its overrides beat the general rulebook, and when no named type fits, compose: route each disputed obligation to the named type it resembles, and let arts. 1 to 183 CO catch everything else.

Drills

1. Your client's contract is titled "Mandat de conseil" but obliges you to deliver a chatbot passing a defined test suite, fixed price, acceptance protocol included. The client cancels halfway and cites art. 404 CO, claiming they owe only work done to date. What is your counter?

Answer: Qualification runs on substance, not label (art. 18 CO). A precisely specified, objectively verifiable deliverable with an acceptance gate is an ouvrage, so the contract is a contrat d'entreprise (art. 363 CO), not a mandat. The client may still withdraw, but under art. 377 CO, which requires paying for work done and fully indemnifying the contractor, including lost profit on the remainder. Art. 404 CO never enters the frame because the mandate checklist was never satisfied.

2. A SaaS contract you signed has no clause about what happens when a bug in the vendor's custom-built integration corrupts your data. Which rules fill the gap, and in what order?

Answer: First the contract itself for anything it does regulate (art. 19 CO). Then, since SaaS is a mixed innominate contract, the disputed element is routed by analogy to the named type it resembles: a defective custom-built integration looks like entreprise, so arts. 367 ff. CO reasoning applies (notice of defect, remedies, acceptance effects); ongoing operational diligence questions would route to mandat (art. 398 al. 2 CO). Behind that, the general part supplies the damages claim (art. 97 CO) and prescription (art. 127 CO). Art. 2 CC good faith is the final backstop.

3. Why does the legislator concentrate mandatory and semi-mandatory rules in the special part rather than the general part, and what does that imply for how carefully you must read each layer?

Answer: The general part addresses abstract parties in symmetric positions, so autonomy is the default and most of it is dispositive. The named contracts encode situations where the legislator identified a typically weaker party (tenant in bail, employee in travail, principal's trust in mandat) and locked protections in place (for example arts. 361 and 362 CO, arts. 271 ff. CO, art. 404 CO). Implication: in the general part you read mostly to learn the defaults you may rewrite; in the special part you must additionally check each rule's precedence tag, because some overrides survive any clause you draft. See ch. 4.

Exercises

Self-test

Fieldwork

• Fedlex · ~30 minOpen the five definition articles — art. 184 (vente), 253 (bail), 319 (travail), 363 (entreprise), 394 (mandat) CO — and copy each defining exchange into one line of your notes. This is your routing table; every contract you ever qualify gets held against these five lines.
• Drafting · ~40 minTake a real project of yours, current or past. Split it into work packages and tag each mandat (open-ended, effort-based) or entreprise (defined, verifiable result). Then write the one-sentence scope line for each package that makes your intended qualification hard to dislodge — “advise and iterate on…” versus “deliver X passing the tests in Annex B”. Composition beats one blurry blob.

What this chapter gives you. Moves 1 through 5 taught you to read a single norm and locate a claim. This chapter zooms out to the claim as a thing with a lifecycle: it is born (formation), it runs its course (performance), it hits trouble (breach, demeure), and it dies (payment, set-off, remission, prescription). Once you see the lifecycle, every dispute becomes a question of stages: what stage is this claim in right now, and which transitions out of that stage are still available? The single most practical payoff is prescription (prescription, statute of limitations): an expiry clock that runs silently in the background and, if you ignore it, kills a perfectly valid claim while you were busy being right. (If you come from software: prescription is a TTL — a time-to-live counter attached to every claim.)

The lifecycle at a glance

            offer + acceptance             performance due
            (art. 1 ss CO)                 (art. 75 CO default: immediately)
                  |                              |
   [NOTHING] ---> [FORMED] -------------------> [DUE] ----------------------+
                  |                              |                          |
                  | vice du consentement         | debtor performs          |
                  | (arts. 23-31 CO)             v                          |
                  | declare within 1 year   [EXTINGUISHED]                  |
                  v (art. 31 CO)             by payment,                    |
             [VOIDABLE ->                    set-off (art. 120 CO),         |
              UNWOUND]                       remission (art. 115 CO)        |
                                                                            |
                       mise en demeure (art. 102 CO)                        |
                                                 +--------------------------+
                                                 v
                                            [EN DEMEURE]
                                                 |
                                     art. 107 CO fork (see below)
                                                 |
                            +--------------------+-------------------+
                            v                    v                   v
                    performance +         damages for         rescission +
                    delay damages        non-performance      restitution
                                                              (art. 109 CO)

   ALWAYS RUNNING IN THE BACKGROUND once the claim is exigible:
   the prescription expiry clock (arts. 127-142 CO). Reaching zero does
   not delete the claim; it arms a kill-switch the debtor may pull
   (art. 142 CO).

The diagram is honest with one caveat: unlike a mechanism whose stage is always plainly readable, a claim can occupy fuzzy stages while a dispute runs, precisely because the facts that drive transitions must be proven, not just asserted. Which stage the claim is in is itself decided under the burden-of-proof rules of art. 8 CC, see ch. 3.

Birth: formation

A contract exists when the parties have exchanged mutual and concordant manifestations of will (manifestation de volonté réciproque et concordante, art. 1 CO). The mechanics are offer and acceptance (arts. 3 to 9 CO), with defaults you would expect from a well-designed set of rules: an offer with a deadline binds until the deadline (art. 3 CO), an offer between absent parties binds until a reply could normally be expected (art. 5 CO), and silence counts as acceptance only where an explicit acceptance is not to be expected given the nature of the deal or the circumstances (art. 6 CO). Agreement is needed on all essential points; secondary points left open do not prevent formation, the judge fills them in (art. 2 CO, the Code of Obligations one, not the Civil Code one).

Form: the default is formlessness

The validity of a contract requires a special form only where the law prescribes one (art. 11 al. 1 CO). This is the big default: your Ekeko consulting contract concluded over email, or even orally on a call, is fully valid. Form requirements are the exception and are always flagged by a specific article, for example the assignment of a claim needs written form (art. 165 CO) and a contract promising a real-estate sale needs a notarial act (art. 216 CO). Where the law does require a form, non-compliance means invalidity (art. 11 al. 2 CO). Parties can also contractually opt in to a reserved form (art. 16 CO). Practical consequence for a solo consultant: writing is for proof and for clarity, not for validity. You sign contracts because of art. 8 CC, not because of art. 11 CO.

Vices du consentement: formation with tainted consent

Sometimes the handshake completes but one party's consent was tainted at the source. The Code recognizes three ways consent can be corrupted (arts. 23 to 31 CO):

Here is the part people get wrong, and it matters operationally. A contract tainted by a vice is not void by itself. It becomes definitively ratified unless the victim declares within one year that they do not maintain the contract, or reclaims what they performed (art. 31 al. 1 CO). The year runs from the discovery of the error or the fraud, or from the moment the fear ceased (art. 31 al. 2 CO). Two practical points:

1. The invalidation is an actual declaration, not a mental state. You must send the message, and you carry the burden of proving it was sent within the window (art. 8 CC). Registered mail exists for a reason.
2. The default is ratification. Do nothing for a year and the tainted contract becomes as solid as a clean one. The system fails toward stability, not toward fairness. Same design philosophy as prescription below: legal certainty beats individual justice after a timeout.

For duress and fraud, invalidation does not exclude damages on top (art. 31 al. 3 CO).

Life: performance and the demeure machinery

A formed claim sits waiting for performance. Absent a term, performance can be demanded and rendered immediately (art. 75 CO). Where and how to perform follows arts. 68 ss CO (personal performance only where the creditor has an interest in it, art. 68 CO; place of performance rules in art. 74 CO, money debts are portable to the creditor's domicile, art. 74 al. 2 ch. 1 CO).

Demeure: putting the debtor formally in default

Late performance is not automatically actionable delay. The debtor of a due obligation enters default (demeure) through a formal reminder, the interpellation (art. 102 al. 1 CO). Exception: if a precise due date was fixed by agreement or by a valid unilateral notice, the deadline itself does the reminding, dies interpellat pro homine (art. 102 al. 2 CO). This is why your invoices should carry an explicit payment date: it turns the calendar into your process server.

Once in demeure, consequences stack:

• For money debts, default interest at 5 percent per year, even if the contract stipulated less (art. 104 al. 1 CO), and higher if a higher commercial rate can be shown between merchants (art. 104 al. 3 CO).
• The debtor answers even for accidental loss occurring during the default, unless they prove the loss would have hit anyway (art. 103 CO).
• Damages for the delay itself, unless the debtor proves no fault (art. 106 and 103 CO).

The art. 107 fork

For bilateral contracts, demeure unlocks the most important decision point in the whole lifecycle. The creditor sets, or has the judge set, a suitable grace period (délai convenable, art. 107 al. 1 CO). If the deadline passes unperformed, the creditor chooses one of three branches (art. 107 al. 2 CO):

1. Keep the contract, demand performance, plus damages for the delay.
2. Keep the contract, renounce performance, and claim damages for non-performance (positive interest: be put where proper performance would have put you). Requires an immediate declaration.
3. Rescind (résolution), unwinding the contract: each side returns what it received, and the creditor can claim negative interest, damages for having relied on a contract that collapsed, unless the debtor proves no fault (art. 109 CO).

The grace period can be skipped where it would be pointless: the debtor's attitude shows it would be useless, performance has become useless to the creditor due to the delay, or a strict deadline (terme fatal) was agreed (art. 108 CO). Note the branch choice is a real commitment, not a menu you can keep hovering over. Declare, and you have moved the claim into a new stage for good.

Breach: the failure path

When performance fails or is defective, the general claim-foundation norm is art. 97 al. 1 CO: the debtor owes damages for non-performance or improper performance unless the debtor proves that no fault is attributable to them. Elements and burden allocation, including the famous fault presumption, are Move 2 and Move 3 territory, see ch. 3. For the lifecycle view, what matters is the damages machinery attached to every breach:

• Existence and amount of damage: proved by the claimant; where exact proof is impossible, the judge estimates equitably (art. 42 al. 2 CO). This is the article that saved many claims from dying on quantification.
• Extent of compensation: set by the judge according to the circumstances and the gravity of fault (art. 43 al. 1 CO).
• Reduction: the award shrinks if the injured party consented, or if facts attributable to them caused or aggravated the damage (art. 44 al. 1 CO). Contributory negligence works as a dial that turns the award down.

These three articles (42 to 44 CO) sit in the tort chapter but apply to contractual damages via the bridge of art. 99 al. 3 CO. General rules reused by a specialized regime, exactly the pattern of Move 5.

Death: how claims are extinguished

A claim can die satisfied, die traded, die forgiven, or die of old age.

Satisfied: payment

Performance extinguishes the obligation, the normal and expected ending (arts. 68 ss CO). Partial payments impute first to interest, then to capital, unless agreed otherwise (art. 85 CO).

Traded: set-off (compensation)

Where two persons owe each other sums of money or performances of the same kind, and both debts are due, each may set off their debt against their claim (art. 120 al. 1 CO). Set-off happens by declaration to the other party (art. 124 al. 1 CO), and then both debts are deemed extinguished, up to the smaller amount, retroactively from the moment they first coexisted as set-off-capable (art. 124 al. 2 CO). Beautiful detail for the expiry-clock section below: a prescribed claim can still be set off if it was not yet prescribed at the moment it could first have been set off against the other debt (art. 120 al. 3 CO). Some claims resist set-off against the creditor's will, notably maintenance and salary claims strictly necessary for living (art. 125 CO).

Forgiven: remission and novation

A debt can be extinguished by simple agreement, the remise de dette (art. 115 CO), and this agreement is formless even if the original obligation required a form. Replacing an old debt with a new one, novation, is never presumed (art. 116 al. 1 CO): signing an acknowledgment or a bill of exchange for an existing debt does not by itself kill the old claim.

Old age: prescription as the expiry clock

Prescription is the expiry clock. Every claim carries a countdown; when it hits zero the claim does not vanish, it becomes unenforceable if, and only if, the debtor raises prescription, because the judge may not apply it ex officio (art. 142 CO). A prescribed debt survives as a natural obligation: pay it voluntarily and you cannot reclaim the payment as unjust enrichment (art. 63 al. 2 CO). So prescription is less an automatic disposal and more a kill-switch the debtor must consciously pull. Many do not, out of ignorance. You, reading this, always will.

The expiry table (all periods verified against the post-2020 revision, in force since 1 January 2020):

The countdown starts when the claim becomes exigible (art. 130 al. 1 CO). Parties cannot shorten or otherwise modify the periods in advance by contract (art. 129 CO), and waiving prescription in advance is void, though the debtor can waive an already running defense for up to 10 years at a time (art. 141 al. 1 and 1bis CO). Precedence tag: mandatory, see Move 4.

Interrupting the clock resets the counter to full (art. 137 al. 1 CO). Two families of interruption events (art. 135 CO):

1. By the debtor: any acknowledgment of the debt, in particular paying interest or an installment, or giving a pledge or a surety (art. 135 ch. 1 CO). This is why collection agencies love getting you to pay "just 50 francs as a gesture."
2. By the creditor: starting debt collection (poursuite, the same LP machinery from your deposit dispute), filing a conciliation request, submitting the claim to a court or arbitral tribunal, or filing in a bankruptcy (art. 135 ch. 2 CO). A mere reminder letter does NOT interrupt prescription. This is the classic trap: people send increasingly stern emails for years while the clock quietly runs out. Only the formal machinery of ch. 9 resets the clock, and a réquisition de poursuite is a cheap, fast reset button that needs no lawyer.

Certain relationships suspend the countdown entirely while they last, for example claims between spouses during marriage and claims of employees against a household employer during the employment (art. 134 CO).

Lifecycle discipline for Ekeko

Concretely, for a solo consultancy: your fee claims against clients likely prescribe in 5 years as professional services or in 10 as general contract claims, and the safe operating assumption is the shorter one (arts. 128 and 127 CO). Invoice with explicit due dates (art. 102 al. 2 CO), charge the 5 percent default interest you are entitled to (art. 104 al. 1 CO), and calendar every unpaid invoice with a reminder well before year 4, at which point the move is a poursuite, not a fifth polite email (art. 135 ch. 2 CO).

The move in one line

Locate the claim's current stage in formation, performance, demeure, breach, or extinction, then act on the transitions that are still open before the expiry clock closes them for you.

Drills

1. On 10 March 2025 you discover that your counterparty fraudulently misrepresented the dataset quality that made you sign a licensing deal in June 2023. It is now August 2026 and you have said nothing to them. What is the state of the contract, and which article decides it?

Answer: The contract is definitively ratified. Fraud (art. 28 CO) made it voidable, but invalidation required an actual declaration within one year of discovering the fraud (art. 31 al. 1 and 2 CO). The window opened 10 March 2025 and closed 10 March 2026. Silence defaulted to ratification. Damages under art. 31 al. 3 CO may still be arguable, but the contract itself stands.

2. A client owes you CHF 12,000, invoice due date 1 September, printed on the invoice they accepted. On 2 September, are they in demeure, and what can you charge? What is your next structural move if they still have not paid by October?

Answer: Yes. A fixed due date dispenses with the interpellation, dies interpellat pro homine (art. 102 al. 2 CO), so demeure started automatically on 2 September. You can charge 5 percent default interest (art. 104 al. 1 CO). Next move: set a suitable grace period under art. 107 al. 1 CO, and if it lapses, choose a branch of the art. 107 al. 2 CO fork; for a pure money claim, in practice you launch a poursuite, which also interrupts prescription (art. 135 ch. 2 CO).

3. Your claim against a client prescribed last month. They separately owe themselves nothing, but you owe them CHF 3,000 for a subcontracted task from last year. Can your dead claim still do work?

Answer: Yes, twice over. First, prescription only bites if the client raises it, since the judge cannot apply it ex officio (art. 142 CO); if they pay without raising it, the payment is valid and unrecoverable (art. 63 al. 2 CO). Second, set-off: a prescribed claim can still be set off provided it was not yet prescribed at the moment it first coexisted, set-off-capable, with the cross-debt (art. 120 al. 3 CO). Since both debts coexisted last year while your claim was alive, you can declare set-off (art. 124 CO) and extinguish your CHF 3,000 debt against it.

Exercises

Self-test

Fieldwork

• Calendar · ~30 minInventory every claim you currently hold: deposits, unpaid invoices, loans to friends, warranty claims. For each, note when it became exigible, which period applies (art. 127, 128, 60 or 67 CO), and the resulting expiry date. Diary a hard reminder at year 4 of any 5-year claim — the move at that point is a poursuite, not another polite email.
• Fedlex · ~20 minRead arts. 135 to 138 CO in full. Then list your habitual collection moves — reminder emails, phone calls, registered letters, payment plans — and mark which ones actually interrupt prescription and which are legally decorative. (Spoiler from the chapter: the letters, however stern, do nothing; only the formal machinery resets the clock.)

What this chapter gives you. The first six moves treat the CC and CO as a predictable machine: element checklists in, legal consequences out. Move 7 is the layer that keeps that machine from producing monstrous results when someone feeds it hostile ingenuity. Art. 2 CC (good faith, la bonne foi / Treu und Glauben) is the system-wide safety valve: al. 1 sets a behavioral norm that colors every right and duty, al. 2 refuses legal protection to the manifest abuse of a right (abus manifeste d'un droit / offenbarer Rechtsmissbrauch). (If you come from software: this is the exception handler wrapped around every rule in the book.) You will learn where this valve concretely opens (surprising clauses in fine print, self-contradictory behavior, weaponized formalities), how good faith drives contract interpretation through the principe de la confiance, why hyper-literal clever tricks reliably lose, and, just as important, why you should almost never build a case plan on art. 2 CC alone.

The architecture: two paragraphs, two jobs

Art. 2 CC is short enough to memorize:

Two design decisions matter here.

First, art. 2 CC is not a claim-foundation norm. Recall the magic question from move 2: who wants what from whom based on what. You cannot answer "based on art. 2 CC". Good faith never grants you a payment claim the CO does not give you. What al. 2 does is defensive and corrective: your opponent invokes a right (a clause, a form defect, a prescription defense), and the court refuses to enforce that invocation. Think of it as a filter fitted around every right in the system: the right still exists, but this particular use of it is turned away at the door.

Second, the trigger threshold is "manifest". Not "somewhat unfair", not "I would have drafted it differently". The Federal Court applies art. 2 al. 2 CC restrictively, as an emergency valve for cases where the formal application of law produces a result that shocks the legal feeling the norm was supposed to serve. The judge applies it ex officio, because the law is applied ex officio (art. 57 CPC, iura novit curia), but the facts that make the abuse manifest still have to be in the record, and under art. 8 CC whoever benefits from the abuse finding carries the burden of proving those facts (see ch. 3).

Where the valve concretely opens

Swiss doctrine has stabilized a small taxonomy of recognized abuse patterns. Treat it as a closed list of known cases, not an open invitation.

Three of these matter constantly in contract practice. Take them one at a time.

1. The surprising clause: clause insolite

General conditions (conditions générales, CG/AGB) are accepted globally: nobody reads them, and Swiss law accepts that as consent anyway, because consent is judged by the principe de la confiance (what the drafter could in good faith understand from your signature). But good faith cuts both ways. The drafter cannot in good faith assume you consented to a clause that is unusual, one that an inexperienced party would never expect in that type of contract, unless the clause was specifically flagged to you.

That is the clause insolite doctrine (Ungewöhnlichkeitsregel). The test, as consolidated in ATF 138 III 411, has two components:

• Subjective: the clause is unusual for this accepting party, given their experience in the field. A reinsurance broker cannot claim surprise at a standard exclusion. A first-time consumer can.
• Objective: the clause materially changes the nature of the contract or departs significantly from the legal default frame of that contract type. In insurance, the paradigm case is a clause that guts coverage so that the most typical risks are no longer covered.

If both hold and no special attention was drawn to the clause, it is simply not part of the contract. Note the mechanism: this is not the court rewriting an unfair term. It is the court reckoning consent honestly under good faith and concluding the term never made it into the agreed content. Formation-level surgery, not fairness review. (Content-level review of abusive general conditions against consumers runs through art. 8 LCD, a different tool, gestured at here and left to the unfair-competition literature.)

For Ekeko this is directly practical in both directions. When a client's procurement department slides you 40 pages of standard terms, the truly weird clause buried on page 31 may not bind you if nobody flagged it. And symmetrically: if your own CG contain anything a client would not expect (broad IP assignment to you, unusual liability shifts), put it in bold, name it in the offer email, get it initialed. The doctrine punishes drafters who rely on nobody reading.

2. Self-contradiction: venire contra factum proprium

Good faith forbids taking a position in flagrant contradiction with your own prior conduct when that conduct created legitimate reliance in the other party. The Federal Court treats this venire contra factum proprium as a standard case of abuse of rights under art. 2 al. 2 CC. The elements, read as a checklist:

1. Prior conduct that objectively signaled a position (tolerating late payments for years, confirming in writing that work was accepted, negotiating renewal terms).
2. Legitimate reliance by the other party on that signal, in good faith.
3. Disposition based on that reliance (they invested, forbore, let a deadline pass).
4. A new position that contradicts the signal and damages the relying party.

Contradiction alone is never enough. Swiss law protects your right to change your mind; what it refuses to protect is inducing reliance and then cashing in on the betrayal. This is the honest boundary of the doctrine, and courts police it strictly: without proven reliance and detriment, the venire argument fails.

You have already fought adjacent terrain. In your deposit dispute, a landlord who signs an exit inspection noting only usure normale and later invents new damage claims is walking straight into this pattern: the inspection conduct signaled a position, you relied on it, the later claim contradicts it. You did not need art. 2 CC because the burden of proof (art. 8 CC) and the missing signed convention de sortie already killed the claim upstream. That ordering is the general lesson: the specific tools usually resolve the case before the safety valve is ever reached. A well-built case rarely needs the override of last resort.

3. Weaponized formalities and abuse in enforcement

Some rights exist to protect a specific interest. Invoking them for a foreign purpose is abuse. The classic line of cases concerns form requirements: a land sale must be in authentic form (art. 216 al. 1 CO), and a defective deed makes the contract void. But in ATF 112 II 330 the Federal Court confirmed that a party who knowingly and voluntarily performed the contract, and only later discovers the form defect as a convenient exit, invokes the defect abusively. The authentic-form rule exists to protect parties from rash real-estate commitments, not to hand a fully served party a free undo button. Two limits from the same case law keep this honest: abuse is generally excluded if the party acted in ignorance of the defect, or if performance never substantially happened. The form rule stays intact; only its parasitic invocation is refused.

The same logic runs through enforcement (see the procedure chapter on the CPC/LP machinery). Raising the prescription defense (art. 127 ss CO, the claim's expiry clock from move 6) is normally a plain right, but raising it after you actively induced the creditor to hold off suing, precisely to let the clock run out, is abusive. Filing an opposition (opposition, art. 74 LP) to a poursuite is a right; conducting enforcement steps purely to harass, with no protectable interest, is not. And on the corporate side, the Durchgriff cases pierce a company's separate legal personality when the shell is used contrary to the purpose of the institution, typically to dodge obligations the controller personally incurred. In every variant the pattern is identical: legal form correct, purpose inverted, protection withdrawn.

Good faith as the interpretation engine: art. 18 CO and the principe de la confiance

Good faith does not only veto abuse. It is also the method for reading contracts, which makes it quietly the most frequently used machinery in this whole book.

Art. 18 al. 1 CO instructs the judge to look for the real and common intent of the parties (volonté réelle), not the inexact labels and words they used. The full procedure, as restated in ATF 144 III 93, runs in two stages with a strict priority:

1. Subjective interpretation first. Establish what the parties actually, commonly meant, using any evidence: drafts, emails, how they performed the contract afterward. If real common intent is proven, it wins, full stop, even against the plain text.
2. Objective interpretation as fallback. If real intent cannot be established (evidence missing or inconclusive), apply the principe de la confiance (Vertrauensprinzip): each declaration means what its recipient could and should in good faith have understood it to mean, in context, at the time of contracting.

Notice what stage 2 does: it imputes to you the objective meaning of your words even if your inner intent was different. You are bound by the reasonable reading of what you said, not by your private mental state. The other side relies on the message you actually sent into the world, in its context, not on your unspoken reservations. (If you come from software: think API contract — callers rely on the documented interface, not on your internal implementation notes, and when the docs are ambiguous, the reasonable reading of the docs governs.) But stage 1 adds a twist with no counterpart in any technical system, and it is the interesting one: if both sides provably shared a private meaning that contradicts the document, the shared private meaning wins (falsa demonstratio non nocet, the wrong label does no harm). Text does not have the last word; contracts bind persons, not paper.

This is also exactly why hyper-literal clever tricks fail. The trickster's plan is always the same: find a reading of the text that is grammatically defensible and commercially absurd, then insist on it. The principe de la confiance kills it at interpretation (no good-faith recipient would understand the clause that way, so that is not what it means), and art. 2 al. 2 CC stands behind as a second layer if the trick somehow survives (an exercise of the clause contrary to its evident purpose is abuse). Swiss contract law is, by construction, not exploitable through pure wordplay. The court does not read the text mechanically; it reads it the way an honest counterparty would.

The honest limits: why you cannot build on the safety valve

Everything above can seduce you into thinking art. 2 CC is a general fairness override. It is not, and the Federal Court says so constantly. The constraints:

• Subsidiarity in practice. If a specific norm handles the problem (a mandatory rule from move 4, defect-of-consent rules in art. 23 ss CO, the clause-formation analysis above), that norm applies, and art. 2 CC adds nothing. Courts reach for al. 2 only when no specific tool exists.
• "Manifest" means manifest. The abuse must be evident on the established facts. Borderline unfairness, hard bargains, and asymmetric outcomes are all legal. Private law assumes autonomous adults allocating risk; a bad deal is not an abused right.
• No new rights. Art. 2 CC can block your opponent's move. It cannot generate the claim you forgot to secure in the contract, resurrect a claim whose expiry clock has run out absent inducement, or substitute for evidence you cannot produce.
• Case-by-case, thin precedent value. Abuse findings are heavily fact-bound. The published ATF decisions (see the precedent chapter) store the categories as settled answers that get reused, but each new case re-litigates whether these facts reach "manifest". Predictive power is correspondingly low, which matters for your expected settlement value: an argument that depends on art. 2 al. 2 CC alone should be discounted hard in the provability term of move 12.

The planning rule that falls out: argue art. 2 CC last, plan as if it does not exist. Draft contracts so the specific tools protect you (clear clauses, flagged unusual terms, documented intent). Litigate on elements and burden. Keep good faith as what it is in the statute: the valve for the cases nobody could specify in advance.

Field checklist: before you cry "abuse"

• [ ] Have I exhausted specific norms first (formation, defects of consent, mandatory rules, interpretation)?
• [ ] Can I name the recognized abuse category from the table above? If it fits none, stop.
• [ ] Can I prove the facts making the abuse manifest (prior conduct, reliance, detriment, knowledge), given that art. 8 CC puts that burden on me?
• [ ] Would a neutral reader call the opposing move shocking, not merely tough?
• [ ] Conversely, before my own clever move: would its success depend on the other side's ignorance, my own contradiction, or a formality serving no protected interest here? If yes, expect the valve to open against me.

The move in one line

Art. 2 CC is the safety valve fitted to every right in the system: it creates nothing, it opens rarely, and it exists precisely so that being technically right in a manifestly rotten way wins nothing.

Drills

1. Your client signed your 12-page general conditions without reading them. Clause 9.4 assigns you all IP in the client's pre-existing codebase you merely touched during the mandate. The client never noticed it and you never mentioned it. Enforceable?

Answer: Almost certainly not, via the clause insolite doctrine (ATF 138 III 411). Global acceptance of general conditions does not cover objectively unusual clauses that were not specifically flagged. The clause is objectively insolite (it inverts the expected allocation, a service provider grabbing the client's pre-existing IP departs radically from the legal frame of a mandate) and subjectively surprising to a client without reason to expect it. Under the principe de la confiance you could not in good faith read their unread signature as consent to it, so the clause never became part of the contract. Note this is formation analysis, not an art. 2 al. 2 CC abuse finding: the valve's logic operates here through the honest reckoning of consent.

2. A buyer and seller execute a real-estate deal with a defective authentic form. Both perform fully and the buyer lives in the house for three years. The market rises and the seller sues for the property back, invoking the form defect (art. 216 al. 1 CO). Outcome, and what single fact could flip it?

Answer: The seller loses. Invoking a form defect after knowing, voluntary, and complete performance, for a purpose foreign to what the form protects (protection against rash commitment, not against market regret), is a manifest abuse of rights under art. 2 al. 2 CC (ATF 112 II 330). The flip fact: ignorance. If the seller genuinely did not know of the defect at the time of performance, or if performance had not substantially occurred, the abuse qualification generally falls away and the nullity can be invoked.

3. You and a client both testify credibly that "delivery" in your contract meant deployment to staging, though any outside reader of the text would understand production. Separately: if the client had understood production all along, what changes? Cite the pipeline.

Answer: Stage 1 of art. 18 al. 1 CO (subjective interpretation, ATF 144 III 93): a proven real and common intent beats the text, so "delivery" means staging, full stop. If the client actually understood production, there is no common real intent, so the judge falls back to stage 2, the principe de la confiance: the declaration means what its recipient could in good faith understand, which here is production, and you are bound by that objective meaning even though your inner intent differed. Same words, opposite results, and the branch point is purely evidentiary, which is why move 3 (see ch. 3) decides interpretation fights more often than dictionaries do.

Exercises

Self-test

Fieldwork

• bger.ch · ~30 minSearch bger.ch (Jurisprudence) for ATF 138 III 411 and read the regeste, the headnote summary. Notice how abstractly the court states the clause insolite rule — almost like a statute amendment. That format is the “settled answer” of Chapter 11, and this is your first live retrieval of one.
• Audit · ~30 minReread your own GTC or contract template with a first-time client’s eyes. List every clause such a client would not expect in this type of contract. For each: either flag it explicitly in your offer email (name it, bold it, have it initialed) or drop it. The doctrine punishes drafters who rely on nobody reading — make sure that is not you.

What this chapter gives you

Part I gave you seven moves for working inside the CC and CO. This chapter zooms out to the first context frame: the legal system runs two fundamentally different modes at once. Private law (droit privé) is horizontal, it governs relations between equals and runs on autonomy: everything not forbidden is allowed, and dispositive rules only fill the gaps you left open. Public law (droit public) is vertical, it governs the state acting on you and runs on legality: the state may only do what a legal norm authorizes it to do. Same country, same courts building, inverted default rules. Knowing which mode you are in tells you who needs a legal basis, which way presumptions point, and what kind of trouble one single act can generate. For a founder registering a company in Vaud, this is not philosophy. It is the difference between "no rule says I can't, so I can" and "no rule says they can, so they can't."

Two permission models

Think of it as two opposite permission models: between private parties, everything not forbidden is allowed; for the state, everything not authorized is forbidden. (If you come from software: two access-control policies — private law is default allow, state power is default deny.)

Private law: everything not forbidden is allowed. Between private parties, freedom of contract (liberté contractuelle) is the base state. The object of a contract can be freely determined within the limits of the law (art. 19 al. 1 CO), and a contract is only void where its content is impossible, illicit, or contrary to morals (art. 20 al. 1 CO). The constitutional anchor is economic freedom (liberté économique, art. 27 Cst.). The statute is mostly a library of defaults: if you and your client wrote nothing about a topic, a dispositive rule fills the gap (see ch. 4 for the dispositive / semi-mandatory / mandatory tags). The law is a book of fallback answers, not a list of permissions.

Public law: for the state, everything not authorized is forbidden. The state's activity has law as both its basis and its limit (principe de la légalité, art. 5 al. 1 Cst.). An administration cannot tax you, fine you, refuse you a permit, or demand a document unless a norm authorizes that act. For taxes the principle is explicit and strict: the status of taxpayer, the object of the tax, and its calculation must be defined by law (art. 127 al. 1 Cst.). For criminal law it is strictest of all: no penalty without a law that predates the act (nulla poena sine lege, art. 1 CP).

The inversion is the whole point. In private law, the person who wants to block something needs a norm (a mandatory rule, a vice of consent, art. 20 CO illiceity). In public law, the state needs a norm to act at all, and you can demand it show one. When a caisse or an office asks you for something, "on what legal basis?" (sur quelle base légale?) is not rude. It is the vertical-law equivalent of the magic question from Part I.

The picture has a known limit: private law is not a lawless free-for-all with a fence around it. Mandatory norms (droit impératif) run straight through contract space, and art. 2 CC good faith acts as a safety valve against abuses case by case (see chapter 7). "Everything not forbidden is allowed" describes the starting posture, not the absence of rules.

One act, several regimes

Here is where the split stops being abstract. A single real-world event can be evaluated independently by three separate machineries, each with its own procedure, standard of proof, and output. None of them automatically consults the others' findings.

Say a hypothetical consultant invoices a client for work never performed, using fabricated deliverables.

1. Civil (horizontal). The client wants its money back: unjust enrichment (art. 62 CO) or damages for breach (art. 97 CO), possibly tort (art. 41 CO). Claim runs through the CPC, proof burden allocated by art. 8 CC, output is a money judgment enforced through the LP.
2. Criminal (vertical). The same facts may be escroquerie, fraud (art. 146 CP): astute deception causing a self-damaging disposition of assets. The state prosecutes under the CPP. Output: a sentence. The client's consent to pay does not run this machinery, the prosecutor does.
3. Administrative (vertical). The tax administration may reassess the fraudulent revenue; a professional regulator, where one exists, may act on fitness to practice.

Crucially, the machineries are decoupled by statute. The civil judge is not bound by the criminal law's notions of imputability, nor by an acquittal, when deciding fault and damages (art. 53 al. 1 and 2 CO). An acquittal for fraud does not delete the art. 62 CO restitution claim, because "not proven beyond reasonable doubt" and "more probable than not" are different measuring sticks applied to the same facts. Conversely, a criminal conviction is powerful evidence in the civil case but the civil court still runs its own evaluation.

Practical consequence for a founder: when something goes wrong with a counterparty, ask which machineries are even switched on. Most business disputes are civil-only, and threatening criminal complaints to collect a civil debt is both usually hollow and potentially itself an offense (contrainte, coercion, art. 181 CP, when the threatened harm has no connection to the claim). Keep the layers separate in your head and in your emails.

The vertical layer a founder actually touches

Founding Ekeko puts you inside several administrative relationships whether you sign anything or not. This is the defining feature of vertical law: it attaches by operation of law (ex lege), not by agreement. A quick map, each item being a separate vertical channel with its own authority, procedure, and deadlines:

Two observations about how that table works as a system.

First, note that the AVS caisse qualifies you. Whether you are self-employed or a disguised employee of your main client is not something you and the client can contract into existence. You are in vertical territory: the caisse applies the legal criteria (own economic risk, freedom of organization, multiple clients) and issues a unilateral decision. Your contract label "independent contractor" is an input, not a controlling one. This regularly surprises people who assume the contract settles the question. In horizontal law it mostly does. In vertical law it is one piece of evidence about facts the authority evaluates independently.

Second, vertical relationships communicate through a specific artifact: the decision (décision, Verfügung). It is the administrative analog of a judgment, it comes with a deadline to object or appeal printed at the bottom (indication des voies de droit), and if you let the deadline pass, the decision becomes final (entre en force) even if it is wrong. There is no "I'll respond when I get to it" in vertical law. The enforcement chapter covers deadlines as a general phenomenon (see ch. 9); here just register that administrative deadlines are typically 30 days and unforgiving.

The blurry middle: hybrid statutes

The clean horizontal/vertical split degrades at the edges. Two statutes a solo AI consultant will actually meet are deliberately hybrid: one text, multiple enforcement machineries.

LCD (Loi contre la concurrence déloyale, unfair competition). The general clause declares unfair any conduct or business practice that is deceptive or otherwise contrary to good faith and that affects competition (art. 2 LCD). Enforcement is triple-headed. Competitors and customers can sue civilly for injunctions and damages (art. 9 LCD), so it behaves horizontally. Many violations are also crimes prosecuted on complaint (art. 23 LCD), so it behaves vertically. And the Confederation itself can sue when collective interests or Switzerland's reputation abroad are at stake (art. 10 al. 3 LCD). If a competitor copies your Ekeko branding or you are tempted to lift a rival's client list, the same act can light up all three channels.

LPD (Loi fédérale sur la protection des données, data protection). As an AI consultant handling client data, this is your most important hybrid. The revised LPD, in force since 1 September 2023, mixes:

• Administrative: the federal commissioner (PFPDT) can open investigations and issue binding administrative measures against controllers (art. 49 ss LPD).
• Criminal: intentional violation of specific duties (information, disclosure, cooperation, data security minima) is punished with fines up to CHF 250'000, and the fines target private natural persons, meaning you personally, not just a corporate shell (art. 60 ss LPD).
• Civil: data subjects enforce their rights through the personality-protection actions of the civil code (art. 32 LPD, routing into art. 28 ss CC).

The design lesson: do not classify a statute once and reuse that label forever. Ask per provision: who enforces this line, against whom, through which procedure? A statute is a container of many rules, not a single regime.

Which way presumptions point, and why you should care

This is the payoff section, because the burden-of-proof machinery from ch. 3 changes shape across the boundary.

Civil: symmetric allocation. Art. 8 CC is regime-neutral between the parties: each side proves the facts it derives rights from, elements for the claimant, defenses for the defendant, and an unprovable fact counts as untrue for whoever needed it. The standard is ordinary proof, with high probability sufficing where strict proof is impossible by the nature of things. Nobody is presumed right. Your deposit dispute ran exactly this machinery: the régie needed to prove damage beyond usure normale and could not, so its claim came to nothing.

Criminal: hard asymmetric presumption. The accused is presumed innocent until final conviction (art. 32 al. 1 Cst., art. 10 al. 1 CPP), doubt benefits the accused (in dubio pro reo, art. 10 al. 3 CPP), and the accused cannot be forced to incriminate himself (nemo tenetur, art. 113 al. 1 CPP). The state carries essentially the entire burden at a demanding standard. This tilt is deliberate: when liberty is at stake, the system would rather let a guilty person walk than punish an innocent one.

Administrative: legality for the state, cooperation from you. Here the geometry is subtler and trips people up. The authority bears the burden of establishing that the legal conditions for a charge or an obligation are met, and the maxim inquisitoire means the authority investigates the facts itself (art. 12 PA at federal level). But you carry a statutory duty to collaborate (obligation de collaborer, art. 13 PA; for social insurance art. 28 LPGA; tax law has its own extensive declaration duties). Refuse to hand over your accounting and the authority may decide on the file as it stands, or estimate your income by appreciation. There is no administrative in dubio pro reo for ordinary assessments.

Why this matters practically, in three rules of thumb:

1. Never import the criminal presumption into a civil dispute. "They can't prove it" is only half a strategy in civil litigation, because you may carry the burden on your own defenses, and at the ordinary standard, plausible evidence can be enough against silence. Conversely, an opponent who barks "that's fraud!" at a civil non-payment is usually bluffing across a regime boundary.
2. In vertical proceedings, silence is rarely free. In a criminal file, silence is a protected right. In a tax or AVS file, silence invites a discretionary assessment (taxation d'office) that then becomes your burden to overturn. Same behavior, opposite payoff, depending on the regime. Check which machinery you are in before choosing a posture.
3. Presumption direction changes what a settlement is worth. The settlement calculus of chapter 12 (see ch. 12) weighs provability. A fact pattern that is weak for a prosecutor at "beyond reasonable doubt" can still be strong for a civil claimant at "more probable than not", and overwhelming for a tax authority that can estimate. Price each channel separately.

The founder's mental switch

Compress the chapter into a quick check you run whenever a legal question appears:

• Who is on the other side? A private party: horizontal mode, think contract, defaults, art. 8 CC. An authority: vertical mode, think legal basis, decisions, 30-day deadlines, cooperation duties.
• Who needs a norm? Horizontal: whoever wants to override the parties' freedom. Vertical: always the state, and asking for the basis is legitimate.
• How many machineries can this one act trigger? Enumerate civil, criminal, administrative. They run independently (art. 53 CO) and settle in one channel without automatically closing the others.
• Which way does the presumption point here? Neutral (civil), pro-you (criminal defense), pro-completeness-of-the-file (administrative, where your cooperation shapes the file).

The move in one line

Before analyzing any legal question, identify the axis: between equals, everything not forbidden is allowed and defaults fill the gaps; against the state, nothing is allowed to the state without a norm, so always ask which regime you are in and which way its presumption points.

Drills

1. Ekeko's first year closes with CHF 85'000 in revenue. A fiduciaire tells you that you were legally obligated to register in the commercial register from day one. True? And is the AVS caisse bound by your consulting contract's clause stating you are an independent contractor?

Answer: False on the first point. A sole proprietorship must register only once prior-year revenue reaches CHF 100'000 (art. 931 al. 1 CO); below that, registration is voluntary (art. 931 al. 2 CO). On the second point, no: self-employed status is a vertical-law qualification decided by the caisse de compensation against the legal criteria of art. 12 LPGA. The contract clause is evidence about the facts, not a binding label, because parties cannot contract their way into or out of a public-law status.

2. A client refuses to pay a CHF 20'000 invoice and you are certain he never intended to pay. Your friend suggests filing a criminal complaint for fraud to "add pressure" to your civil claim. Map the regimes and name the risk.

Answer: Two independent machineries. The civil claim for the price runs through art. 97 CO (or the contract-specific norm) with art. 8 CC allocating proof at the ordinary standard, enforced via the LP. Fraud (art. 146 CP) requires astute deception, and mere non-payment or even initial bad intent is hard to establish at the criminal standard, where the presumption of innocence protects the client (art. 10 CPP). An acquittal would not kill the civil claim (art. 53 CO), but the reverse risk is real: threatening a criminal complaint to extract payment can constitute coercion (art. 181 CP) if wielded as leverage disconnected from a genuine criminal suspicion. Keep the channels separate.

3. The tax office sends a taxation d'office estimating your self-employment income far above reality, because you missed the declaration deadline. You reason: "they bear the burden of proof, I'll just deny it." What did you get wrong?

Answer: You imported the wrong regime's presumption. In administrative proceedings the authority investigates ex officio, but you owe a duty to collaborate (art. 13 PA, and specific tax declaration duties). Having breached it, you allowed the authority to estimate by appreciation, and the burden effectively flips: it is now on you, within the strict objection deadline printed on the decision, to demonstrate the assessment is manifestly wrong, with the decision becoming final if you let the deadline pass. Silence that would be a protected right in a criminal file (art. 113 CPP) is an expensive strategy in a vertical file about money.

Exercises

Self-test

Fieldwork

• Your own files · ~20 minTake the most recent letter you received from any authority — tax, commune, caisse, Serafe. Find two things: the legal basis it cites for the obligation, and the indication des voies de droit (the objection route and deadline, usually at the bottom). If no basis is cited, draft the one-line question asking for it. Note the deadline in your calendar even if you plan to comply.
• Fedlex · ~30 minRead the four verticals a founder actually meets: art. 5 and 127 Cst. (legality, taxation by law), art. 931 CO (commercial register threshold), art. 10 LTVA (VAT liability and the CHF 100’000 exemption). For each, note in one line what event triggers the obligation — these attach by operation of law, not by anything you sign.

What this chapter gives you. Everything in Part I lived in substance land: does a claim exist, who must prove what, which norms apply. This chapter is about the other half of the machine: the machinery that enforces it. A right that exists on paper but has no working path to money or performance is a key to a door that was never built. Switzerland gives you two pieces of enforcement machinery: the LP (Loi fédérale sur la poursuite pour dettes et la faillite, debt enforcement) for claims payable in money, and the CPC (Code de procédure civile) for getting a court to say who is right. They interlock, and the single most practical thing you can learn about Swiss law is how a claim travels through both. You have already travelled this path once, on Florissant, from the debtor's seat. This chapter shows you the whole board, including the creditor's seat you will occupy when Ekeko's first client does not pay an invoice.

Substance vs procedure: the rules vs the machinery

Substantive law (CO, CC) defines what obligations exist. Procedural law (CPC, LP) defines how they get enforced against someone who does not cooperate. The two are different bodies of law with different logic. (If you come from software: substance is the source code, procedure is the runtime that executes it.)

The critical coupling: procedure can kill a perfectly good substantive claim (missed deadline, no title, no provable facts), and procedure can temporarily advance a bad one (an unopposed commandement de payer proceeds even if the debt is fiction). The machinery does not check whether your claim is true. It checks your paperwork.

A peculiarity that surprises everyone: for money claims in Switzerland, you can start enforcement before any court has looked at the merits. The debt-collection sequence begins on the creditor's bare assertion. The merits check happens later, and only if someone forces it.

The LP sequence, end to end

For any claim in Swiss francs, enforcement runs through the office des poursuites. Here is the full sequence, stage by stage.

1. Réquisition de poursuite (art. 67 LP)

The creditor files a request, written or oral, naming creditor, debtor, amount, and the cause of the claim (art. 67 LP). No evidence required. No filter. This is why anyone can put anyone into poursuite: the office does not evaluate the claim, it just follows the procedure. Cost: a fee scaled to the amount, advanced by the creditor but ultimately charged to the debtor if the poursuite succeeds (art. 68 LP).

2. Commandement de payer (art. 69 LP)

The office drafts and serves the commandement de payer: pay within 20 days, or oppose within 10 (art. 69 al. 2 LP). This is the packet that landed in your mailbox from Gerofinance. It looks official because it is, but remember what it certifies: only that someone filed a réquisition. Nothing about the merits.

3. Opposition (art. 74 LP): the 10-day emergency brake

The debtor has 10 days from notification to declare opposition, orally or in writing (art. 74 al. 1 LP). No reasons needed (art. 75 al. 1 LP). One word on the form suffices. Effect: the poursuite freezes (art. 78 al. 1 LP). The ball flies back to the creditor, who now must actively remove the opposition or the poursuite dies.

This is the cheapest, highest-leverage move in Swiss law. Ten days, zero francs, zero justification, and it forces the counterparty to either lawyer up or walk away. Miss the 10 days and the poursuite continues as if the debt were established. The deadline is the whole game. (Late opposition exists, art. 77 LP, but only on narrow grounds. Treat 10 days as hard.)

4. The creditor's fork: three ways to clear an opposition

Now the merits enter, but how they enter depends entirely on what paper the creditor holds. This is the concept of the titre, the title.

Mainlevée définitive (art. 80 LP): with a judgment, the debtor's remaining defenses are nearly nil, essentially proof by document that the debt was paid or deferred after the judgment, or prescription (art. 81 al. 1 LP). The judgment already settled the merits; the machinery just carries it out.

Mainlevée provisoire (art. 82 LP): the interesting middle tier. The creditor needs a reconnaissance de dette, an acknowledgment of debt in an authentic act or a signed private document, for a determinate sum (art. 82 al. 1 LP). A signed contract stating a fixed rent, a signed loan agreement, a signed convention de sortie with an amount on it: all qualify. An unsigned draft, an email thread, a disputed damage estimate: none do. The judge grants provisional lifting of the opposition unless the debtor immediately makes a liberating defense plausible (vraisemblable, not proven, art. 82 al. 2 LP). Summary procedure, document-based, fast.

Notice what the requirement actually tests: signature plus determinate sum is a check on form, not on truth. The system trusts a signed acknowledgment enough to fast-track it provisionally, while routing everything else through the slow path. (If you come from software: it is a schema check, not a truth check.)

5. Action en libération de dette (art. 83 al. 2 LP): the 20-day counterattack

Mainlevée provisoire is provisional for a reason. The debtor gets 20 days from the mainlevée decision to file an action en libération de dette, a full lawsuit asking the court to declare the debt nonexistent, handled in ordinary or simplified procedure (art. 83 al. 2 LP, no conciliation needed, art. 198 let. e CPC). Two things make this action remarkable:

• Roles flip, burden does not. The debtor is formally the plaintiff, but the creditor still bears the burden of proving the debt exists, exactly as art. 8 CC allocates it (see ch. 3). The procedure inverted the seating chart, not the burden of proof.
• Miss the 20 days and the mainlevée becomes definitive (art. 83 al. 3 LP). Another hard expiry clock.

6. Continuation (art. 88 LP)

With no opposition, or opposition cleared, the creditor can request continuation: no earlier than 20 days after notification of the commandement de payer, and the right expires one year after notification (art. 88 al. 1 and 2 LP), the clock pausing while merits litigation runs (art. 88 al. 2 LP). Then the sequence branches on debtor type (art. 38 and 39 LP):

• Saisie (seizure of assets, art. 89 ss LP) for private individuals.
• Faillite (bankruptcy) for entities and sole proprietors registered in the commercial register (art. 39 LP).

Founder note for 2027: the day Ekeko is inscribed in the registre du commerce, an unpaid CHF 800 invoice against it can, in principle, be pushed all the way to a faillite threat. Registration changes what failure looks like, from "they take a laptop" to "they can threaten the company's existence." Price that into how you handle poursuites against the company: never ignore one.

7. The register, and how to clean it (art. 8a LP)

Every poursuite lands in the register, and anyone showing a plausible interest, typically a prospective landlord or lender, can get an extract (art. 8a al. 1 and 2 LP). Since a poursuite requires no evidence, this is a reputational weak point anyone can exploit. The remedy: art. 8a al. 3 let. d LP. If you filed opposition, you can ask the office, from 3 months after notification of the commandement de payer, to stop disclosing the poursuite to third parties. The office gives the creditor 20 days to prove they started a procedure to remove the opposition (art. 79 to 84 LP); if they cannot, the entry goes dark to outsiders. Translation: an opposed poursuite that the creditor never pursued is treated as unsubstantiated noise and hidden from public view. If Gerofinance's poursuite is still visible and they have not sued, this request is a one-page letter. Nulled, withdrawn, or successfully contested poursuites are likewise non-disclosed (art. 8a al. 3 let. a to c LP).

The CPC path: how the merits actually get decided

Whenever a fork above says "full lawsuit," you enter the CPC machinery. Its stages:

Conciliation first (art. 197 CPC)

Almost every claim must first pass through the autorité de conciliation (juge de paix in Vaud; the commission de conciliation for tenancy). Exceptions in art. 198 CPC include summary procedure and the LP-linked actions such as the action en libération de dette. The hearing is cheap, informal, and settles a large share of cases, which is exactly the design: it is a mandatory settlement checkpoint before the expensive path unlocks. Reread it through the expected-value lens of chapter 12: the state forces both parties to price their case once before burning real money.

If no settlement, the authority issues the autorisation de procéder (art. 209 CPC), a ticket to court valid 3 months (art. 209 al. 3 CPC), or only 30 days in tenancy disputes (art. 209 al. 4 CPC). Yet another expiry clock: sit on the ticket and your path closes (the substantive claim survives, but you re-queue at conciliation).

The conciliation authority also has real teeth at the low end:

(The CHF 10,000 figure is the post-2025 revision value, raised from 5,000. Older articles online still show the old number.)

Which procedure runs the case

• Procédure simplifiée (art. 243 al. 1 CPC): patrimonial disputes up to CHF 30,000, plus certain matters regardless of value, including the core tenancy protections (art. 243 al. 2 let. c CPC). Simpler forms, more active judge, oral options.
• Procédure ordinaire: above CHF 30,000. Full written pleadings, formal.
• Procédure sommaire (art. 248 ss CPC): document-only fast track for defined cases, including all mainlevée decisions (art. 251 CPC).

Costs: who really pays

• Advance: the court can demand an advance from the plaintiff, since the 2025 revision capped at half the expected court costs for most procedures (art. 98 al. 1 CPC), with full advances still possible for conciliation, summary procedure, and appeals (art. 98 al. 2 CPC). This halved the entry toll for ordinary suits, a deliberate access-to-justice fix.
• Allocation: loser pays court costs and the winner's dépens (art. 106 al. 1 CPC), scaled on partial wins (art. 106 al. 2 CPC).
• The gap nobody tells you about: dépens are set by cantonal tariff (art. 96 CPC) and routinely cover only part of the winner's actual lawyer bill. Winning fully can still leave you thousands out of pocket. That delta belongs in every settlement calculation.
• Free-ish zones: no dépens are awarded in conciliation (art. 113 al. 1 CPC), and no court costs are charged in conciliation for tenancy disputes and for employment disputes up to CHF 30,000 (art. 113 al. 2 CPC). A tenant can force a landlord to a conciliation hearing for the price of a stamp.

Worked example: Florissant, or why no title meant no path

Replay your deposit dispute step by step through both machineries.

Gerofinance claimed CHF 4,700 for end-of-tenancy damage. Substantively (chapter frames from ch. 2): who wants what from whom based on what? Landlord wants damages from tenant based on art. 267 CO (return of the premises), with art. 267a CO's prompt-notice requirement and the rule that normal wear and tear (usure normale) is covered by the rent, plus art. 8 CC putting the burden of proving excessive damage on the landlord.

Now the procedural overlay, and watch how each fact opens or closes a gate:

1. The convention de sortie was never signed. In substance land, that just means no acknowledgment. In LP land it is decisive: no signed document with a determinate sum means no reconnaissance de dette, so art. 82 LP is unavailable. Mainlevée provisoire, the cheap fast track, is closed.
2. No judgment exists, so art. 80 LP is closed too.
3. Therefore the only path is art. 79 LP: a full action on the merits, meaning tenancy conciliation, then simplified procedure, with the landlord bearing the art. 8 CC burden of proving damage beyond usure normale against an état des lieux that says usure normale. Lawyer hours against a CHF 4,700 upside, with dépens tariffs guaranteeing the recovery never covers the spend.
4. Your opposition (art. 74 LP) cost you ten minutes and forced them onto that path.

The one-line summary of the whole dispute: the moment you did not sign, their claim lost its title; the moment you opposed, their claim lost its path. The substantive question (was it really usure normale?) never even needed to be litigated, because the expected value of the procedure collapsed first. That is "no right without a path" in its sharpest form: whatever right they believed they had was worth, in practice, exactly what the path to enforce it would cost. Legal strength = provability x enforceability x opponent's costs (chapter 12), and enforceability multiplied everything by roughly zero.

And the mirror lesson for Ekeko: make sure your claims always carry a title. A signed contract or engagement letter stating fees as determinate sums is not just evidence, it is an art. 82 LP fast-track ticket. A client who signed "CHF 12,000 payable 30 days after delivery" and does not pay faces poursuite, then mainlevée provisoire in summary procedure, then a 20-day shot clock to sue you with the burden still on proving their own liberation plausible at the mainlevée stage. A client on a handshake faces a polite email. Contract hygiene is enforcement-path design.

Where the machinery picture breaks down

Two honest caveats. First, real machinery runs the same way every time; judges do not. Summary mainlevée is close to mechanical, but merits litigation involves discretion, credibility, and art. 2 CC equity overrides. Second, the sequence has human costs the diagram hides: a poursuite against you is stressful and reputation-relevant even when meritless, and litigation consumes attention for months. The diagram tells you your options; it does not tell you the load they place on you.

The move in one line

Before valuing any claim, yours or theirs, trace its enforcement path: no title means the slow road, and every step on that road has a deadline that can delete the right.

Drills

1. A client signed Ekeko's engagement letter: "fees: CHF 15,000, due 30 days after final delivery." Delivery accepted, 60 days pass, no payment, and after your poursuite the client files opposition. What is your next move, which article, and what must the client show to survive it?

Answer: Request mainlevée provisoire under art. 82 al. 1 LP: the signed letter with a determinate sum is a reconnaissance de dette. Summary procedure (art. 251 CPC), document-based. The client must immediately make a liberating defense plausible (vraisemblable), for example documented non-delivery or payment (art. 82 al. 2 LP). If mainlevée is granted, the client has 20 days to file an action en libération de dette (art. 83 al. 2 LP), where the burden of proving the debt still sits with you under art. 8 CC.

2. Someone files a baseless CHF 9,000 poursuite against you. You oppose within 10 days and hear nothing more. What happens to your enforcement register entry, and what can you do about it?

Answer: The entry stays visible to third parties with a plausible interest (art. 8a al. 1 and 2 LP). From 3 months after notification of the commandement de payer, request non-disclosure under art. 8a al. 3 let. d LP: the office gives the creditor 20 days to prove they started a procedure to remove your opposition (art. 79 to 84 LP); failing that, the poursuite is no longer disclosed to third parties. The creditor's right to continue the poursuite itself expires one year after notification (art. 88 al. 2 LP).

3. You want to sue a supplier for CHF 8,000 in damages. Map the procedural sequence: mandatory first step, what the authority could do there, which procedure applies in court, and what you must budget beyond your own lawyer.

Answer: Conciliation first (art. 197 CPC). There the authority can broker a settlement, submit a proposition de décision since the value is under CHF 10,000 (art. 210 al. 1 let. c CPC, binding if no party objects within 20 days, art. 211 CPC), but cannot decide outright since the value exceeds CHF 2,000 (art. 212 CPC). Failing settlement you get an autorisation de procéder valid 3 months (art. 209 al. 3 CPC), then procédure simplifiée since CHF 8,000 is under 30,000 (art. 243 al. 1 CPC). Budget: a court-cost advance of up to half the presumed costs (art. 98 al. 1 CPC), refundable from the loser if you win (art. 106 CPC), plus the gap between your actual lawyer bill and the tariff-based dépens (art. 96 CPC), which you eat even in victory.

Exercises

Self-test

Fieldwork

• Audit · ~30 minGo through your client paperwork and sort it by art. 82 LP: which documents are signed and state a determinate sum (a reconnaissance de dette, fast-track ticket), and which are just correspondence? Then fix the pipeline: engagement letters with fixed amounts and signatures, acceptance emails that name the milestone price. Contract hygiene is enforcement-path design.
• Registry · ~15 minRequest your own extract from the registre des poursuites (your local office des poursuites; online in most cantons, small fee). Check for stale entries: any opposed poursuite the creditor never pursued can be hidden from third parties on request after 3 months (art. 8a al. 3 let. d LP) — a one-page letter that cleans the record your next landlord will read.

What this chapter gives you. Every rule that can touch your contract lives at some level of a hierarchy of norms: constitution, federal statute, ordinance, cantonal law, and finally the contract itself. Picture it as a stack of levels, each outranking the one below. This chapter gives you the order of precedence within that stack, the two tie-breaking maxims (lex specialis and lex posterior), the RS numbering system as the master catalogue of Swiss law, and the federalism twist that matters most in practice: the substance of contract law is 100% federal, but the courts that apply it and the officers who enforce it are partly cantonal, with Vaud running its own quirky arrangement. When Ekeko signs its first consulting contract in 2027, every clause in it will sit at the bottom of this stack, and knowing what sits above it tells you which clauses are real and which are decoration.

The stack, top to bottom

Think of it as layered levels of authority: a rule set at a higher level wins over a conflicting rule below it, and lower levels only fill gaps the higher ones leave open. (If you come from software: this is precedence in a layered config system — higher layers override, lower layers fill defaults.)

Two constitutional rules pin this hierarchy in place. Art. 49 al. 1 Cst states the force dérogatoire du droit fédéral: federal law breaks conflicting cantonal law, full stop. And art. 190 Cst tells the Federal Tribunal that federal statutes and international law are determinative for it: even if a federal statute arguably violates the Constitution, the court must still apply the statute. So for statutes the stack is enforced downward against cantons and contracts, but not fully enforced upward against the Constitution. That is a known inconsistency in the architecture, accepted deliberately: Parliament plus referendum is trusted over judicial review. Say what you will, it is documented, intended behavior, not an oversight nobody got around to fixing.

The contract layer deserves one precision. A contract is not law, it is an act of private autonomy that the law above it validates and fills in. Freedom of contract (art. 19 al. 1 CO) means you can set most terms yourself, but only within the limits carved by mandatory law (art. 19 al. 2 and art. 20 al. 1 CO: illegal, impossible, or immoral content is void). Whether a given statutory rule yields to your clause or overrides it is exactly the dispositive / semi-mandatory / mandatory tagging from ch. 4. The stack tells you that statute beats contract when they conflict; the tags tell you whether there is a conflict at all.

RS numbers: the law's master catalogue

Switzerland maintains the Recueil systématique (RS, in German SR): the systematic compilation of all law in force, published at fedlex.admin.ch. Treat it as a catalogue where the number itself tells you the topic:

• First digit = domain. 1 is the state and its structure (the Constitution is RS 101). 2 is private law, civil procedure, and enforcement. 3 is criminal law. 4 is education and culture. 8 is health and social security, and so on.
• Longer numbers mean narrower topics. RS 210 is the Civil Code (CC). RS 220 is the Code of Obligations (CO). RS 272 is the Civil Procedure Code (CPC). RS 281.1 is the debt enforcement and bankruptcy act (LP), which you met the hard way during the deposit dispute. More digits, more specific: RS 221.411 is the commercial register ordinance you will touch when registering Ekeko.
• Prefix 0 = international. Any RS number starting with 0 is a treaty: the CISG is RS 0.221.211.1, and notice how its suffix mirrors the domestic private-law numbering, the same topic in the international shelf of the catalogue.
• Fedlex serves the current consolidated text. Statutes get amended constantly; fedlex always shows the merged, currently in force text, with a version history. When you cite "art. 257e CO", you implicitly mean the version currently in force. If a dispute is about events in the past, check whether the article changed since, because the applicable version is usually the one in force at the relevant time (transitional rules, tit. fin. CC, govern the changeover).

Habit worth installing: when anyone cites a law by name, resolve it to its RS number and open the consolidated text on fedlex. Secondary sources drift out of date; the official catalogue does not.

Conflict resolution within the stack

The vertical rule is easy: higher layer wins (art. 49 al. 1 Cst for federal over cantonal; art. 19 and 20 CO for mandatory statute over contract). The interesting cases are horizontal: two norms at the same level that both claim to apply. Two classic maxims act as tie-breakers. Neither is written in a statute; they are interpretation principles courts apply as background logic.

Lex specialis derogat generali. The more specific norm displaces the more general one for the cases it covers. This is the specialized override beating the general rule, the exact mechanism behind chapter 5's design: the CO general part (art. 1 to 183 CO) provides defaults, and the named-contract rules override them where they speak. Art. 97 CO gives the general damages rule for non-performance; art. 259e CO gives the tenant-specific damages rule for defects of the rented thing; within tenancy, the specific rule is applied first, and the general one fills what it leaves open.

Lex posterior derogat priori. Between equally ranked and equally specific norms, the later one wins, on the theory that the newer enactment expresses the legislator's current intent. The most recent word prevails, but the maxim is used sparingly, because usually the newer statute explicitly repeals or amends the older one in its final provisions, so the conflict never reaches the tie-breaker.

When the two maxims collide (older-but-specific versus newer-but-general), there is no mechanical rule: courts resolve it by asking what the newer legislator plausibly intended, which is interpretation, see ch. 11. The honest summary: within a level, precedence is resolved by reasoning about purpose, not by a lookup table. That is where the layered-levels picture breaks.

Federalism: the substance is federal, the machinery that runs it is partly cantonal

Here is the split that confuses newcomers and that you have already lived through once:

• Substantive private law is exclusively federal (art. 122 al. 1 Cst). The CC and CO apply identically in Vaud, Zurich, and Ticino. There is no "Vaud contract law". Since 2011 civil procedure is also federal: one CPC for the whole country (art. 122 al. 1 Cst, RS 272), and debt enforcement was federal long before that (LP, RS 281.1).
• Court organization stays cantonal (art. 122 al. 2 Cst, art. 3 CPC). Which courts exist, how they are named, how many instances below the Federal Tribunal, which judges sit where: each canton builds its own machinery to apply the shared federal rules. Cantons also appoint the enforcement apparatus: the office des poursuites that served your commandement de payer is a cantonal office executing federal LP procedure.

So the same rulebook is administered by 26 differently organized cantons. (If you come from software: one program, deployed on 26 differently configured hosts.) Concretely in Vaud:

Note what is cantonal here and what is not. The requirement of conciliation before suing is federal (art. 197 CPC), and the requirement of a joint commission with landlord and tenant representatives for lease disputes is federal too (art. 200 al. 1 CPC). But the fact that in Vaud this commission sits at the préfecture, a very Vaudois institution, is pure cantonal organization. Your deposit dispute was federal substance (art. 267 CO restitution, art. 8 CC burden, see ch. 3), federal procedure (CPC conciliation, LP opposition), running on Vaud's machinery.

Practical consequence for Ekeko: your contracts can and should include a forum clause (élection de for, art. 17 CPC allows it for pecuniary disputes between businesses). The clause picks the place; the canton's organization then determines which court at that place hears it. You choose the region, the canton chooses the courtroom.

One more federalism residue: a few pockets of private-law-adjacent regulation are cantonal because the CC delegates them (reserved cantonal law, art. 5 CC), for instance parts of neighbor law and public notarial organization. For contract work you will rarely meet them, but it explains why notary fees for founding a company differ by canton while the company law itself (art. 620 ss CO) does not.

The international layer

Treaties sit above statutes in principle (art. 5 al. 4 Cst; the Federal Tribunal generally lets international law prevail in conflict, with a debated carve-out for statutes Parliament passed deliberately against a treaty). For a contract practitioner, two imported rulebooks matter:

CISG (Vienna Sales Convention, RS 0.221.211.1). If you sell goods to a buyer in another contracting state, the CISG applies by default and silently replaces the CO's sales rules. It applies unless you opt out, not the other way round (art. 6 CISG). Swiss practice opts out almost reflexively ("the application of the CISG is excluded"), often from habit rather than analysis. For Ekeko this is mostly a non-issue, services are outside the CISG's scope (art. 3 al. 2 CISG excludes contracts where the preponderant part is labour or services), but if you ever ship a product with hardware, decide consciously instead of pasting the ritual clause.

Choice of law and forum across borders. For international contracts, the LDIP (private international law act, RS 291) decides which country's law applies absent a choice, and lets parties choose (art. 116 LDIP). Cross-border enforcement of judgments within Europe runs on the Lugano Convention (RS 0.275.12). If a foreign client ever refuses to accept Swiss law and Lausanne courts, that is a priced concession, not a formality: it degrades the enforceability factor in the settlement EV of ch. 12.

Tracing any rule to its level

When you hit a rule in the wild ("you must", "you cannot", "the deadline is"), run this checklist:

1. Is it in the contract? Then ask: which statutory rule sits behind it, and is that rule dispositive (clause valid), semi-mandatory (valid only if it favors the protected party), or mandatory (clause void)? See ch. 4.
2. Is it a statute? Which one? Resolve to an RS number, open fedlex, read the current article. Check whether a more specific statute overrides it for your case (lex specialis).
3. Is it an ordinance? Then it must rest on a delegation in a statute; an ordinance cannot create obligations the statute does not authorize. If a claimed rule appears only in an ordinance and stretches beyond its statute, it is attackable.
4. Is it cantonal? Then it can only govern what federal law left to cantons, mostly organization, procedure residue, and taxes. A cantonal rule purporting to change contract substance loses under art. 49 al. 1 Cst.
5. Is it a treaty? Check scope and whether the contract opted out where opting out is allowed (CISG yes, art. 6 CISG; Lugano no).
6. Is it none of the above? Then it is practice, a template habit, or someone's bluff. Common in negotiations: "standard terms require" often traces to nothing above layer 6, which means it is fully negotiable.

The single most useful output of this trace is negotiating information: a demand grounded in mandatory federal law is a wall, a demand grounded in a dispositive default is a door, and a demand grounded in nothing is fog.

Where the analogy breaks

The stack is not a clean chain of overrides. Three honest caveats. First, art. 190 Cst means the top layer is not fully enforced against federal statutes, as noted above. Second, within a layer, conflicts resolve by purposive reasoning, not mechanically. Third, art. 1 CC adds a fallback ladder inside the statute layer: where the statute is silent, the judge applies customary law, and failing that decides according to the rule they would adopt as legislator. The system ships with an explicit instruction for what to do when its own rules run out. (If you come from software: a documented handler for undefined behavior — more than most codebases can say.)

The move in one line

Before arguing about any rule, locate its layer: what sits above it beats it, what sits below it fills its gaps, and what traces to no layer at all is just negotiation fog.

Drills

1. Your Ekeko client's procurement team inserts: "Cantonal law of Geneva governs the interpretation of this agreement." What is wrong with this clause, and what did they probably mean?

Answer: There is no cantonal contract law to choose: substantive private law is exclusively federal (art. 122 al. 1 Cst), so "Geneva law" for a domestic contract resolves to the same CC and CO as everywhere else. They probably meant a forum clause selecting Geneva courts (élection de for, art. 17 CPC), which is a real and valid choice for pecuniary disputes. Fix the clause to say what it means: Swiss law governs, courts of Geneva (or better for you, Lausanne) have jurisdiction.

2. A supplier in Germany sells you sensor hardware for a client project. Your contract is silent on applicable law. Which sales rules govern, and at which layer of the stack do they sit?

Answer: The CISG (RS 0.221.211.1) governs by default: both Switzerland and Germany are contracting states, the contract is for goods, and the CISG applies unless excluded (art. 6 CISG), so silence means it applies. It sits at the treaty layer, above the CO, and displaces the CO's sales provisions (art. 184 ss CO) within its scope. If you wanted CO rules instead, you needed an explicit exclusion clause.

3. Vaud requires tenancy conciliation to happen before a commission seated at the préfecture. Your friend claims this proves procedure is cantonal. Untangle which parts of that setup are federal and which are cantonal.

Answer: Federal: the obligation to attempt conciliation before suing (art. 197 CPC), the requirement that the lease conciliation authority be a joint commission with landlord and tenant representatives (art. 200 al. 1 CPC), and the whole procedure the commission follows (CPC, RS 272). Cantonal: which body plays that role, where it sits, and who chairs it (art. 122 al. 2 Cst, art. 3 CPC), hence the préfecture and the préfet in Vaud, and the Tribunal des baux as the specialized court behind it. Same federal procedure, Vaud-specific machinery.

Exercises

Self-test

Fieldwork

• Fedlex · ~20 minFrom memory, write the RS numbers for: Cst., CC, CO, CPC, LP. Verify on Fedlex, then add LTVA and LPD to the list by searching. Keep the seven-line table where you draft — resolving a name to an RS number and opening the current consolidated text is the single habit this chapter wants installed.
• Fedlex versions · ~20 minOpen art. 98 CPC on Fedlex and find its version history: locate the 1 January 2025 change that capped the court-cost advance at half. Practice answering “which version was in force on date X?” — the skill that matters whenever a dispute is about events in the past and the article has since changed.

What this chapter gives you. A statute is a highly compressed summary of life. Articles are short, life is long, and every real case forces a judge to unfold open terms like "normal wear" or "just cause" into a concrete yes or no. This chapter shows you how that unfolding works: the four interpretation methods the Tribunal fédéral actually uses, the fallback protocol when the statute has a genuine gap (art. 1 CC), the explicit delegation to equity (art. 4 CC), and then the layer of settled answers, Swiss precedent, which is persuasive rather than binding, yet followed with near-perfect discipline. (If you come from software: precedent works like a cache — reused for speed and consistency, invalidated only deliberately.) You will learn to read a citation like ATF 139 III 297 or 4A_545/2011, find the decision on bger.ch, use the commentaires as the field's reference literature, and understand honestly how this differs from common law stare decisis. This is Frame D of Part II: after the hierarchy of norms told you which text applies (see ch. 10), this chapter tells you what that text means when it does not say.

Why interpretation is not optional

You met open terms the moment you read your first article as a checklist (Chapter 1). "Usure normale", "justes motifs", "sans délai", "bonne foi": these are not defects, they are deliberate openness. The legislator states the rule in outline and leaves the filling-in to the judiciary, because enumerating every case would be impossible and instantly outdated. So every element of a Tatbestand that is not a number or a date eventually needs interpreting.

The Tribunal fédéral refuses to rank its interpretation methods. It calls its own approach a pragmatic pluralism of methods (pluralisme pragmatique des méthodes, the classic reference is ATF 123 III 292): no method has priority, and even a clear literal text can be set aside if the other methods show it misses the point of the rule. For a systems mind this is uncomfortable, because it means interpretation is not a fixed mechanical procedure. It is closer to a panel of four advisers, each answering from a different angle, with the judge weighing the votes. The honest consequence: you can predict outcomes well, but never with mechanical certainty.

The four methods

Two practical notes. First, the literal method runs on three equally authoritative texts (German, French, Italian), so a divergence between language versions is itself an argument, and lawyers genuinely check all three. Second, the teleological method usually wins ties in modern case law, because the court says the goal is the objectively right decision in the normative system, oriented to the ratio legis, not text-worship.

Worked example: usure normale

You have live experience here. Art. 267 al. 1 CO says the tenant must return the premises in the state resulting from a use conforming to the contract (usage conforme au contrat). Nothing in the CO defines "normal wear". Run the four methods:

1. Literal. "Use conforming to the contract" implies that living in an apartment, as contracted, degrades it. Walls dull, carpets flatten. The text itself concedes that some degradation is the tenant's contractual right, not damage. But the text gives no threshold: where does living end and damaging begin? The words are clear; the dividing line is not.
2. Systematic. Look at the neighbors. Art. 256 al. 1 CO obliges the landlord to deliver and maintain the thing in a state fit for its use: maintenance burden sits with the landlord by default. Art. 259 CO gives the tenant only minor cleaning and repairs (menus travaux). The scheme allocates the cost of keeping the thing usable to the party collecting rent. Reading "usure normale" narrowly against the tenant would contradict the architecture of the surrounding rules.
3. Historical. The tenancy law revision in force since 1990 was explicitly protective of tenants; the legislative materials treat rent as the price that already compensates the landlord for ordinary depreciation. The drafting record says, in effect: do not bill twice.
4. Teleological. Purpose: fair risk allocation. Wear is a predictable, priced-in operating cost; damage is an anomaly attributable to the tenant. Hence the practice courts converge on: amortization tables (tableaux de durée de vie, negotiated by landlord and tenant associations, soft law that courts use as a reference) that give each component a service life. A 20-year-old carpet has residual value zero; the tenant who ruins it owes nothing, because its value was already fully consumed by lawful use.

Four methods, one convergent output: normal wear belongs to the landlord's ledger, and even genuine damage is compensated only at residual value. That is the interpretation your Gerofinance dispute ran on, stacked on top of the art. 8 CC burden allocation from ch. 3: the landlord had to prove damage beyond usure normale, and could not.

When the statute is silent: art. 1 CC

Sometimes interpretation fails honestly: there is no rule to interpret. Art. 1 CC is the famous fallback protocol, and it deserves its reputation:

• Al. 1: the law governs every matter to which its letter or its spirit applies. Interpretation first, always.
• Al. 2: absent an applicable provision, the judge decides according to customary law (droit coutumier), and absent custom, according to the rules the judge would adopt acting as legislator (modo legislatoris).
• Al. 3: in doing so, the judge draws on established doctrine and case law (jurisprudence).

Read that as a ladder of fallbacks. Custom almost never comes into play in practice; the interesting rung is the third: when there is a true gap (lacune), the judge does not simply halt and refuse to decide, the judge writes the missing rule, generally and prospectively, as if drafting a statute. This is not a license for vibes. The judge as legislator must produce a rule that would fit the rest of the law, and al. 3 immediately tethers the exercise to the existing body of learning: doctrine and precedent. Note what al. 3 does not say: it says the judge takes inspiration (s'inspire) from case law. That single verb is the whole Swiss theory of precedent, and we will come back to it.

Delegated equity: art. 4 CC

Distinguish the gap case from a different mechanism. Sometimes the legislator deliberately delegates. Art. 4 CC: where the law reserves the judge's discretion (pouvoir d'appréciation) or directs a decision based on the circumstances or on just cause (justes motifs), the judge decides according to law and equity (règles du droit et de l'équité).

So whenever an article contains "circumstances", "justes motifs", "équitable", that is not sloppiness, it is an explicit standing instruction from the legislator: work this value out case by case, at the moment of decision. Examples you will meet in contract life: reduction of an excessive penalty clause (art. 163 al. 3 CO), termination of durable contracts for just cause, moral damages "in view of the circumstances". Art. 1 CC fills gaps the legislator did not see; art. 4 CC fills blanks the legislator drew on purpose. One is a hole nobody noticed; the other is a space left open by design. (If you come from software: art. 1 CC is a missing function; art. 4 CC is a function pointer the legislator registered on purpose.)

Precedent: the shelf of settled answers

Now the layer of settled answers. When the Tribunal fédéral interprets "usure normale" or fills a gap, the result is expensive to produce: full four-method analysis, five judges, sometimes a public deliberation. Nobody wants to redo that work in every case. So the system puts the answer on the shelf and reuses it.

The Swiss rule: precedent binds no one, formally. A cantonal court in Vaud is not legally obliged to follow ATF case law, and the Tribunal fédéral is not bound by its own prior decisions. The only statutory hook is art. 1 al. 3 CC's "s'inspire". Precedent in Switzerland is a shelf of persuasive answers, not of binding ones.

Why everyone follows it anyway. Three forces, all incentive-shaped rather than rule-shaped:

1. Reversal probability. A lower court that ignores standing TF case law will almost certainly be overturned on appeal. Judges do not enjoy being reversed; parties can predict the reversal and appeal accordingly. Deviation has a near-guaranteed cost and near-zero payoff.
2. Equal treatment and legal certainty. Art. 8 Cst. (equality) and the constitutional value of legal certainty push toward like cases being decided alike. Treating identical facts differently without reason is itself arguable as arbitrariness (art. 9 Cst.).
3. Discipline in overturning settled answers. The TF allows itself to change course (changement de jurisprudence, Praxisänderung) only for serious, objective reasons: better insight into the ratio legis, changed circumstances, evolved legal views. A change applies immediately, including to pending cases, with a carve-out for procedural traps: a change that would forfeit a party's right through, say, a new way of computing appeal deadlines must not arrive without warning (see ATF 135 II 78). In short: settled answers are overturned rarely, openly, and with transition warnings.

The net effect: a system in which reuse is formally optional behaves, at the 95th percentile, exactly like binding precedent. For your settlement EV calculations (see ch. 12), you can treat a clear ATF on point as near-deterministic.

Reading a citation

Two citation formats exist, and they encode different things.

ATF 139 III 297 (in German: BGE) is a published leading decision:

Publication in the ATF collection is itself a signal: the court decided this ruling states a principle worth settling and reusing. Contract law lives overwhelmingly in part III.

4A_545/2011 is a docket number (numéro de dossier), attached to every decision, published or not:

A decision cited only by docket number ("arrêt du TF 4A_545/2011 du 11 janvier 2012") was not selected for the official collection. It is still real case law, freely citable, just a weaker kind of settled answer: less curated, sometimes fact-specific.

Finding decisions. Everything since 2000 is full-text searchable at bger.ch under "Jurisprudence". Search by docket number, by ATF reference, or by keywords, ideally the French or German term of art ("usure normale", not a paraphrase). The statutes themselves live on fedlex.admin.ch in consolidated form. Free, no paywall, and the ATF headnotes (regestes) are translated summaries worth reading first.

Doctrine: the reference literature

Between raw statute and raw case law sits doctrine (la doctrine): academic commentary, above all the article-by-article commentaires. The big ones for your purposes: the Commentaire romand (French), the Basler Kommentar (German), the Berner Kommentar (German, exhaustive), plus treatises like Tercier/Pichonnaz on obligations. A commentaire entry on art. 267 CO will give you the interpretation history, the ATF references, the controversies, and the practice tables, in one place.

Think of doctrine as the field's reference manual and its running record of open disputes, combined. It is not law, it binds nobody, but art. 1 al. 3 CC names it explicitly as a source of inspiration, and Swiss courts cite professors by name in judgments, openly. When doctrine is split, judgments say so ("la doctrine majoritaire considère..."), and the majority view is a genuine predictor of outcomes. If you ever need to research a contract question seriously: statute first, then the commentaire on that article, then the ATFs the commentaire points to. That path is faster and safer than keyword-searching case law cold.

The honest difference from stare decisis

Because the analogy will tempt you, here is where it breaks.

Two consequences matter in practice. First, a Swiss ATF states its principle abstractly, almost like a statute amendment, so you spend less effort on the facts of the precedent and more on the stated rule. Second, because nothing formally binds, argument from purpose can beat argument from precedent: if you can show the settled interpretation no longer serves the ratio legis, you are making a legitimate move, not heresy. It rarely wins, but it is on the board in a way it is not for an English county court.

So the settled-answers picture, honestly audited: Swiss precedent is a body of settled answers with no hard guarantee that they will keep being applied, held coherent by incentives (reversal risk), by constitutional pressure (equality, non-arbitrariness), and by disciplined, announced overturning. It behaves like a strictly binding system almost always, and the "almost" is exactly where good lawyers hunt.

The move in one line

When the text runs out, run the four methods; when the law runs out, the judge writes the missing rule (art. 1 CC); and treat ATF case law as a set of settled answers, only persuasive on paper but near-binding in expectation.

Drills

1. Art. 337 CO lets an employer terminate immediately for "justes motifs". Is a judge deciding whether motifs are "justes" filling a gap under art. 1 al. 2 CC, or doing something else?

Answer: Something else. "Justes motifs" is a deliberate delegation, so the judge decides according to law and equity under art. 4 CC (pouvoir d'appréciation). Art. 1 al. 2 CC only comes into play for a true gap, a question the statute neither answers nor delegates. Here the legislator saw the question and intentionally left the space open: a blank drawn on purpose, not a hole nobody noticed.

2. Decode "ATF 147 III 440 consid. 4.1" and "5A_121/2019". What does each fragment tell you, and which carries more weight?

Answer: ATF 147 III 440: published leading decision, volume 147 (147 + 1874 = year 2021), part III so civil law, starting at page 440, and you are pointed at reasoning block 4.1 (considérant). 5A_121/2019: docket number, second civil law court (family, property, successions, LP), case 121, filed 2019, and citation by docket alone signals it was not selected for the official collection. The ATF carries more weight: publication means the court flagged it as a principle worth settling and reusing.

3. A cantonal judge in Vaud is convinced a standing ATF interpretation of a CO article is wrong. Legally, may the judge deviate? Practically, what happens, and what would it actually take to flip the case law?

Answer: Legally yes: precedent is not formally binding in Switzerland, art. 1 al. 3 CC only requires drawing inspiration from case law. Practically the decision gets appealed and almost certainly reversed, which is why deviation is rare. Flipping the case law requires convincing the Tribunal fédéral itself that serious, objective reasons justify a changement de jurisprudence: better insight into the ratio legis, changed circumstances, or evolved legal views. The new line then applies immediately, even to pending cases, subject to fair warning where a procedural change would forfeit rights (ATF 135 II 78).

Exercises

Self-test

Fieldwork

• bger.ch · ~30 minRetrieve ATF 115 II 464, the decision that makes art. 404 CO mandatory. Read the regeste, then find the considérant that states the rule. Notice the format: an abstract principle, stated almost like a statute amendment, detached from the parties — the Swiss unit of precedent, and the opposite of a common-law ratio welded to its facts.
• Library · ~45 minGet your hands on a commentaire entry for one article you care about — the Commentaire romand on art. 404 CO, say, via a university library or an online excerpt. Observe the layering: text, drafting history, ATF references, doctrinal controversies, practice notes. Then memorize the research path this gives you: statute first, commentaire second, the ATFs it points to third. Faster and safer than keyword-searching case law cold.

What this chapter gives you. Almost no dispute ends with a judge. The overwhelming majority end in a negotiated number, and that number is not random: it is priced by what a court would do, discounted by proof risk, enforcement risk, and everyone's costs. Mnookin and Kornhauser called this "bargaining in the shadow of the law", and it means the CC and CO you learned to read mechanically in Part I are not primarily litigation tools. They are pricing inputs for a market that trades before trial. This chapter gives you the pricing formula, the real Swiss cost stack, the expected-value (EV) arithmetic of a live dispute (your own Florissant deposit fight as the worked example), the moves that shift EV during negotiation, the cases where fighting genuinely beats settling, and the hygiene rules for closing a deal so it stays closed.

The market nobody sees

Courts publish judgments. They do not publish the far larger set of disputes that settled in a lawyer's letter, a conciliation hearing, or a terse email with a bank transfer attached. In Swiss civil practice, most disputes never reach a judgment on the merits: a large share die or settle at the mandatory conciliation stage (art. 197 CPC makes conciliation the default entry point for civil claims), and most of the rest settle before or during trial. Rough order of magnitude, and it varies by domain: think 90 to 95 percent resolved without a merits judgment. The trials you read about in ATF decisions are the tail of the distribution, selected precisely because both sides thought they could win (or one side was irrational, insured, or fighting for something other than money).

Think of the courtroom as the expensive, slow venue of last resort. The shadow market is the informal venue where the actual volume clears, and its prices track the courtroom's rules for a simple reason: either side can always force the matter into court, at a cost. So the informal price never drifts far from what the court would order, minus the cost and risk of going there. Your legal knowledge matters mostly because it changes your price in that market.

The strength formula

Frame E of this book:

Legal strength = provability x enforceability x opponent's costs.

Multiplication, not addition. A zero in any factor zeroes the product.

Note where the analogy is honest and where it breaks: provability and enforceability really are probabilities you can estimate. "Opponent's costs" is not a probability, it is a lever that shifts their EV threshold for settling. The formula is a mnemonic for what to audit, not an equation you evaluate to four decimals.

The Swiss cost stack

EV arithmetic needs real numbers. These are current orders of magnitude for francophone Switzerland; verify tariffs for your canton before relying on them.

Lawyer rates. CHF 300 to 500 per hour for ordinary civil work in Vaud and Geneva; partners at business firms higher. A modest dispute (demand letter, conciliation, one exchange of briefs, one hearing) burns 15 to 40 hours. So a "small" CHF 20,000 fight can easily cost CHF 8,000 to 15,000 in fees per side.

Court advance, avance de frais. The plaintiff fronts court costs before the machine starts. Since the CPC revision in force 1 January 2025, the court may as a rule demand at most half of the expected court costs as advance (art. 98 al. 1 CPC), with exceptions where the full amount can still be required, notably conciliation, summary proceedings, and appeals (art. 98 al. 2 CPC). Expected court costs themselves follow cantonal tariffs (art. 96 CPC) and scale with the amount in dispute: for a CHF 30,000 claim in Vaud, think low four figures.

Frais vs dépens, and the unrecoverable delta. Costs follow the event: the loser pays court costs and owes the winner dépens, party compensation (art. 106 CPC). But dépens are set by cantonal tariff, not by your actual invoice. Winning typically recovers only part of what you paid your lawyer. That unrecoverable delta is a pure deadweight loss you eat even when you win, and it belongs in every EV calculation. In conciliation proceedings no dépens are awarded at all (art. 113 al. 1 CPC): each side eats its own costs there by design.

The conciliation layer. Before most claims reach a court, you pass through the conciliation authority (art. 197 CPC). It is cheap, fast, and for residential and commercial lease disputes it is free of court costs entirely (art. 113 al. 2 let. c CPC). If no deal, the plaintiff gets an autorisation de procéder and must file at court within 3 months, or within 30 days for lease disputes (art. 209 al. 3 and 4 CPC). Miss the window and the authorization lapses: back to the start of the queue.

Time. Conciliation to first-instance judgment: expect 1 to 2 years for a contested ordinary case, plus appeals. For a solo consultant, 30 hours of your own attention on a dispute is 30 hours not billed and not sold. Price your own hours honestly; they are usually the largest hidden line item.

The EV template. For a claimant:

EV(fight) = P(win) x (amount + tariff dépens) − P(lose) x (court costs + opponent's dépens) − own lawyer fees − own time − enforcement risk haircut

Compare against EV(settle) = offer on the table, now, certain, minus nothing. The gap between those two numbers is the entire negotiation.

Worked example: the Florissant table

Your deposit dispute, run through the formula. Facts: landlord's régie claims CHF 4,700 for end-of-lease damage. The convention de sortie was never signed by you. The état des lieux notes normal wear, usure normale, which the tenant never owes: the tenant returns the premises in the state resulting from conforming use (art. 267 al. 1 CO), and defects must be raised by the landlord immediately at handback on pain of forfeiture for discoverable defects (art. 267a CO). The deposit sits in a blocked account and the bank may only release it with both signatures, a judgment, or after one year without legal action by the landlord (art. 257e al. 3 CO). Burden of proving damage beyond normal wear: the landlord, because it is an element of their claim (art. 8 CC).

The landlord's EV table for actually pursuing CHF 4,700:

The régie's rational move is to bluff by invoice and hope you pay from fear. The moment you demonstrated you knew the burden allocation and the deposit mechanics, their bluff EV collapsed too, because bluffing only pays against people who fold. That is why the dispute resolved without a hearing. You did not win by being right. You won by making being right legible, which moved their EV below zero. The law never entered a courtroom; it worked entirely as a shadow price.

Moves that shift EV

Negotiation moves are EV surgery on one of the three factors.

Evidence demands (attacks provability). Asking the other side, in writing, to specify exactly which facts they will prove and with what, forces them to price their own proof risk. "Please indicate which document establishes acceptance of these charges" is a checklist attack straight out of ch. 1. Silence or vagueness in response is itself information, and it is information their own lawyer now has to explain internally.

Insurance locks (shifts costs asymmetrically). Protection juridique insurance (yours or theirs) changes who feels the cost factor. If you hold legal protection insurance, your marginal cost of fighting drops toward zero and your threats become credible: say so early. If they are insured, expect less settlement pressure from fees and more from time and annoyance. For Ekeko: a legal protection policy for business disputes is cheap EV armor; the premium buys credibility in every future negotiation, not just coverage.

Counterclaim options (raises their downside). A live counterclaim, reconvention (art. 224 CPC), converts their lawsuit from a bounded bet into an open one. Even a modest, well-founded counterclaim (defects in their performance, art. 97 CO; unpaid extras, art. 62 CO) means their worst case is no longer "recover nothing" but "recover nothing and pay". Never invent one; a weak counterclaim costs you credibility, which is the currency everything else here is denominated in.

Deadlines (weaponizes the procedural machinery). Procedure is full of clocks that transfer pressure: the 30-day lease filing window after conciliation (art. 209 al. 4 CPC), the one-year deposit release (art. 257e al. 3 CO), prescription running down its expiry clock (art. 127 f. CO, see ch. 6), the 10-day window to oppose a poursuite (art. 74 LP). Starting a poursuite yourself is a legitimate pressure move: it costs little, interrupts prescription (art. 135 ch. 2 CO), and lands the debtor a registry entry visible to future landlords and lenders (art. 8a LP). Conversely, filing an opposition, as you did, costs one signature and forces the creditor to go get a court title (art. 79 to 82 LP). Cheap moves that force expensive responses are the best moves in this market.

Anchoring with the law visible. A settlement offer that shows its arithmetic ("your provable exposure is X, your cost to establish it is Y, here is Z") is harder to dismiss than a bare number, because rejecting it requires disputing the arithmetic. You are handing them the memo their own advisor would have to write.

When fighting is worth it

Settlement is the default, not a law of nature. Fight when:

1. The product of the formula is genuinely high. Strong documents, clear burden position, solvent opponent, and a cost-shifting regime that favors you. Then EV(fight) can honestly exceed any offer.
2. You are a repeat player and this case sets your price. If Ekeko lets one client skip an invoice quietly, the next negotiation with anyone who heard about it starts lower. Repeat players rationally fight some negative-EV cases to maintain the shadow price of all future contracts. This is reputation as a settled answer that gets reused: expensive to establish once, then invoked in every later negotiation for free. (If you come from software: reputation works like a cache — computed once, read many times.)
3. Someone else pays. Legal protection insurance, or a fee-shifted forum where your recoverable dépens approximate real costs.
4. The claim is cheap to run. Small claims in simplified procedure (art. 243 CPC, up to CHF 30,000) with tight documents can be run with light lawyer involvement or none.
5. Principle, priced honestly. Fighting "on principle" is legitimate if you write down the number you are paying for the principle and accept it deliberately. Unpriced principle is just tilt with a file number.

Do not fight when the missing factor is enforceability. A brilliant case against an insolvent counterparty is a donation to your lawyer. Check the registre des poursuites extract (art. 8a LP) before investing, not after.

Settlement hygiene

A settlement is itself a contract, a transaction, formed like any other (art. 1 CO) and binding without any court. Sloppy settlements spawn the sequel dispute. Rules:

• Written, always. Not for validity (art. 11 CO: form is free unless prescribed) but for provability. You learned in ch. 3 what an oral deal is worth when the other side develops amnesia.
• Full and final, pour solde de tout compte. The phrase that closes the ledger: each party declares all claims from the named relationship settled. Scope it consciously. "All claims arising from the lease of [address]" kills everything in that container, including claims you have not discovered yet, so make sure you want the container closed and not just this claim.
• Name the mechanics. Who pays what, to which IBAN, by which date, and what happens on non-payment (the whole original claim revives, not just the discounted one, is a standard and worthwhile clause).
• Clean up the machinery. If a poursuite was started, the settlement must oblige the creditor to withdraw it; a withdrawn poursuite is no longer disclosed to third parties by the office (art. 8a al. 3 let. c LP). If a conciliation or court case is pending, settle it in the procedure where possible: a settlement recorded before the conciliation authority or court has the force of a binding decision (art. 208 al. 2 CPC), which upgrades your contract into an enforceable title for free.
• No-admission clause if you are the payer. "Without any admission of liability" costs nothing and blocks the settlement from being recycled as evidence elsewhere.
• Mind the semi-mandatory floor. Some claims cannot be waived by settlement at all: an employee cannot give up claims arising from mandatory rules during the employment and for one month after it ends (art. 341 al. 1 CO). A "full and final" signed inside that window is partially void, and the waived claims come back. Check ch. 4 before assuming a signature closes anything.

The move in one line

Compute both sides' expected value under the real Swiss cost stack, then negotiate by moving their number, because 9 disputes in 10 are priced in the shadow of the courtroom, not inside it.

Drills

1. Your client owes Ekeko CHF 18,000 on an invoice, disputes nothing in writing, but stalls. List the cheapest EV-shifting move first and cite its article.

Answer: Start a poursuite (LP). It is cheap, requires no proof at that stage, interrupts prescription (art. 135 ch. 2 CO), and creates a registry entry visible to the client's future counterparties (art. 8a LP). If they oppose (art. 74 LP), your signed contract and invoice set you up for provisional mainlevée (art. 82 LP). Note the claim is under CHF 30,000, so any merits phase runs in simplified procedure (art. 243 CPC) after mandatory conciliation (art. 197 CPC).

2. In the Florissant table, which single fact change would have flipped the landlord's EV positive, and through which factor of the strength formula?

Answer: A convention de sortie signed by the tenant acknowledging the CHF 4,700. That is a provability change: it converts the landlord's hardest element (damage beyond usure normale, accepted by the tenant) into a document, roughly inverting P(win) under art. 8 CC. Enforceability and costs were never the landlord's problem; proof was. Moral: never sign the exit inventory under pressure at the door.

3. You settle a dispute with a former employer three weeks after your last day, signing "pour solde de tout compte" and waiving overtime pay. Enforceable?

Answer: Not as to the overtime, if it derives from a mandatory or semi-mandatory rule: art. 341 al. 1 CO bars the employee from waiving such claims during employment and for one month after termination. The settlement stands for genuinely disputed, freely disposable points, but the protected claims revive despite the full-and-final wording. Timing the same settlement five weeks after the end would change the analysis.

Exercises

Self-test

Fieldwork

• Pricing · ~30 minTake one live or past dispute of yours and run the EV template with real numbers — from both chairs, the way the Florissant table does it for the régie. Then write down your settlement floor before any negotiation happens, so no future mood can drift it. If you have no dispute handy, price the Ekeko CHF 13,000 invoice from Chapter 14.
• Market · ~20 minGet two quotes for a business protection juridique policy. Compare the annual premium against the EV shift it buys you in a typical dispute at your scale — not just coverage, but the credibility of every future threat to escalate. Cheap EV armor is still armor.

What this chapter gives you. Everything in Part I and Part II was abstraction. This chapter is the full method run on a real case, end to end: your own case, Florissant 31, Prilly, replayed move by move. (If you come from software: this is the integration test.) A régie claimed CHF 4,700 after a 13.5-month tenancy; the realistic exposure ended at CHF 100, conceded voluntarily for credibility. Nothing exotic happened. Twelve ordinary models, applied in the right order to a stack of PDFs, did all the work. Each section below tags the move being executed, so you can watch the toolkit run on real inputs instead of textbook ones.

The board at move zero

Seven items: two walls at CHF 1,000 each, bathroom joints CHF 1,000, tub re-enameling CHF 1,000, kitchen cabinets CHF 500, shower head CHF 100, one missing key CHF 100. Total « estimatif » CHF 4,700 against a guarantee of CHF 4,812. That coincidence is not a legal argument, but it is a prior worth logging.

Move: the magic question (see ch. 2)

Who wants what from whom based on what? The régie wants CHF 4,700 from the tenant. Based on what norm? There are only two candidate claim foundations, and they route completely differently:

1. Contract: the convention de sortie as a reconnaissance de dette (acknowledgment of debt, art. 17 CO). If it exists, the debt is stipulated; the régie only enforces a signed number.
2. Statute: damages for returning the premises in worse-than-conforming state, art. 267 al. 1 CO in connection with art. 97 CO. If the convention does not exist, the régie must build the claim element by element.

The tablet failure at 17:21 deleted route 1. The « Locataire sortant » signature block holds a typed name and nothing else. There is no reconnaissance de dette. Everything downstream follows from asking the question precisely: the régie is not "owed CHF 4,700 minus objections", it is at zero and must construct a claim under route 2 from scratch. Framing determines who is climbing and who is standing at the top.

Move: decompose art. 267 into elements (see ch. 1)

Art. 267 al. 1 CO: the tenant must return the thing in the state resulting from use conforming to the contract. Read as Tatbestand → Rechtsfolge, the landlord's damage claim reduces to this checklist, every element conjunctive:

One false element and the whole claim yields nothing. Here elements 2 through 5 all fail, and they fail on the régie's own documents. The tub is the elegant one: entry EDL says « Émaillage neuf », the amortization table gives a professional re-enameling 20 years of life, and it needs redoing after 13.5 months of use the exit EDL itself calls normal. That is not tenant damage; that is a workmanship question between the owner and its contractor. Note the trap avoided: letter v1 argued "old tub, fully amortized," which the entry EDL flatly contradicts. Verifying the entry EDL page by page before sending replaced a fatal argument with a better one.

Move: the burden ladder (see ch. 3)

Art. 8 CC puts every element of that checklist on the landlord: prove the damage, prove it was absent at entry, prove it exceeds normal wear, prove the quantum (art. 42 CO). Unprovable = false for whoever needed it, and the régie needed all five.

The EDLs are the burden instruments, the case's only sensors. Two readings matter:

• The exit EDL is signed and therefore opposable to both sides. But read what it actually records: « Usure normale • Traces », « Usure normale • Marques », « Usure normale • Rémaillage à effectuer », ménage « satisfaisant », no amounts, no items charged to the tenant. The tenant's 17:02 signature endorses those factual findings; it cannot be stretched to cover amounts fixed in a separate document generated nineteen minutes later and never signed.
• The entry EDL is the delta baseline. Damage claims are computed as exit minus entry, and the entry snapshot already contains the kitchen cabinet marks (photos 11-15), the joint marks (« Marques sur les joints », photo 25), and a shower head in « usure normale ». Delta on those items: zero.

Then there was the live burden-reversal attempt. When the tenant contested a point at the inspection, the agent's method was, in substance: it is not recorded in the entry EDL, so it cannot be verified, so it is imputed to you, doubt against the tenant. That is art. 8 CC run backwards. A silent entry EDL creates no presumption against the tenant; silence is the landlord's proof gap, not the tenant's. The letter names the method rather than quoting words that were not certain verbatim, a small honesty decision that also happens to be the legally safer one.

The framing is exact here: where data is missing, the law does not split the difference or guess a middle value. It assigns the loss entirely to the party who carried the burden. The régie's evidence base was its own two EDLs, and both testified for the tenant.

Move: fine print vs mandatory law (see the precedence-tags chapter, 04)

Adversarial review of the dossier surfaced a clause buried on page 17 of the exit EDL: signature boilerplate aimed at making the document a reconnaissance de dette for art. 82 LP purposes (provisional lifting of opposition in debt enforcement). Two answers, layered:

1. Factual: an art. 82 LP title must state a determined, unconditional sum. The signed EDL contains no amounts at all; the amounts live only in the unsigned 17:21 document. A clause cannot acknowledge numbers that are not in the instrument.
2. Precedence: art. 267 al. 2 CO declares void any pre-termination agreement obliging the tenant to pay lump sums beyond actual damage. Tenant-protection law in the CO restitution rules is mandatory or semi-mandatory; contract fine print sits at the bottom of the norm stack (see ch. 10) and cannot write itself an exemption. The convention's own clause, « les retenues valent reconnaissance de dettes », would have collided with the same wall even if signed: a reconnaissance de dette is causal (art. 17 CO), so every defense on the underlying debt, usure normale, amortization, absence of actual cost, survives it.

This is the precedence-tag move doing quiet, decisive work: before arguing facts, check whether the clause being waved at you is even allowed to exist.

Move: vices du consentement, declared preemptively (see ch. 6, formation stage)

Belt and suspenders. In case the régie ever alleged some form of consent to the amounts, the letter declares invalidation for defect of consent (vices du consentement, art. 23 ss CO, erreur essentielle) within the one-year window of art. 31 CO. Two drafting details mattered. First, art. 31 CO requires a declaration, not a reservation; a draft that "reserved the right to invalidate" was corrected to an actual declaration, because a reserved right invalidates nothing. Second, the declaration is conditional in form ("should any consent be alleged") so it concedes no consent while still stopping the clock. Formation defects are fought at the formation stage of the lifecycle; waiting until enforcement to raise them is how deadlines eat defenses.

Move: art. 267a CO, the forfeiture gate

Art. 267a CO obliges the landlord, at restitution, to verify the premises and immediately notify the tenant of defects for which the tenant answers; absent timely notice, claims are forfeited except for hidden defects (art. 267a al. 2 and 3 CO). Two uses in this case:

• Backward: the exit EDL cannot simultaneously classify an item as « usure normale » and constitute clear notice of a defect the tenant answers for. A notice must let the tenant identify what is actually imputed to him; "normal wear" is the opposite of that.
• Forward: the flat was re-let the same day, in the recorded state, zero vacancy. Once third parties occupy the premises, any later-discovered defect that a routine inspection should have caught is barred. The claim surface is frozen at 14 Aug 2026 and cannot grow.

A related discipline followed from this: the letter names none of the vague items' possible real targets (holes likely drilled by previous tenants, a pre-broken glued block under the sink). Vague charges stay vague until the régie commits to specifics in writing, and specifics inherit the régie's proof problem, not the tenant's.

Move: the enforcement path (see the substance-vs-procedure chapter, 09)

No right without an enforcement path, and here the path analysis flipped the entire picture of what was actually at risk. The Gmail sweep revealed that the "garantie bancaire" in the lease is not a bank deposit at all: it is a rent-guarantee insurance policy (UNIQA via AutoMate, caution solidaire, art. 496 CO). No blocked capital exists. The real risk is the insurer paying the régie on demand and then recovering from the tenant.

So read the insurer's CGA the way you learned to read statutes: mechanically, condition by condition. The payment gate: UNIQA pays only against (1) a tenant-signed reconnaissance de dette, (2) a commandement de payer not struck by opposition, or (3) a final judgment. Three doors, and the tenant controls two of them directly:

• Door 1 stays shut by the standing rule sign nothing: no re-sent convention, no release, no décompte.
• Door 2 stays shut by opposition totale within 10 days of any commandement de payer (art. 74 LP), the one hard deadline in the entire case, the same LP mechanics you already know from Part II. An opposed poursuite costs the creditor a fee and forces it into court; an unopposed one is a payment title.
• Door 3 requires the régie to litigate through mandatory conciliation and win on the merits, against its own EDLs.

Plus a timer running for the tenant: absent a lawsuit within one year of restitution, the guarantee releases de plein droit (diary entry: ~15 Aug 2027). Substantively the régie's claim was weak; procedurally it had no self-executing path to the money. Both layers checked, as chapter 9 insists they always must be.

Move: the endgame expected value (see the settlement chapter, 12)

Legal strength = provability × enforceability × opponent's costs. Run the numbers from the régie's chair, which is exactly what the adversarial round did: the dossier was staged three times as a disguised régie mandate to outside counsel, in fresh sessions with independent legal research. All three simulated memos converged: expected recovery CHF 400-700, recommendation settle or drop. The régie faces near-zero provability (its own documents say normal wear), blocked enforcement (three-door gate, opposition promised in writing), and real costs (conciliation, then court, for a consumer-sized claim its file cannot carry).

On the tenant's side, one deliberate expected-value (EV) trade: the CHF 100 key was conceded pour solde de tout compte. The memory of the entry agent testing every key was vague; contesting a CHF 100 item on shaky recollection risks the credibility of six solid contestations worth CHF 4,600. Paying 100 to armor 4,600 is not weakness, it is portfolio management. The settlement floor was set explicitly in advance, never pay above ~CHF 300, so that no future negotiation mood could drift it. And the letter's posture matches the EV: ASLOCA membership stated as fact, consultation booked, door to an amiable solution held open, deadline of 31 Aug given. Strength, calmly priced.

What the régie's lawyer would have needed to win

To collect meaningfully more than the key, opposing counsel needed, per item:

1. A signed convention, or failing that a signed instrument stating determined amounts (art. 82 LP). Does not exist; the only signed document contains no amounts.
2. Proof that each state exceeded usure normale (art. 267 al. 1 CO), against an exit EDL that classifies 6 of 7 items as normal wear. That means contradicting the client's own contemporaneous, countersigned report.
3. Proof the damage post-dates 27 Jun 2025, against an entry EDL documenting the kitchen marks, joint marks, worn shower head, and a freshly enameled tub whose premature failure points at the owner's contractor, not the tenant.
4. Timely, specific notice of each defect (art. 267a CO), when the notice given classified the items as wear, and the flat was re-let the same day, freezing the record.
5. Actual invoices and residual-value math under the tableau paritaire (art. 42 CO), instead of seven round estimates summing suspiciously close to the guarantee.
6. A willingness to fund conciliation and litigation for an expected value its own file prices at a few hundred francs.

Each requirement is independent and each was missing. A chain of six conjunctive elements with per-element probability near zero multiplies out to a case no rational lawyer runs.

The five decisions that won it

1. Do not sign, ever again. The tablet failure was luck; converting it into a standing rule (sign nothing, no convention, no release, no décompte) was a decision. It kept door 1 of the insurer's gate permanently shut.
2. Fight on their documents, not on memory. Every load-bearing argument quotes the régie's own EDLs: « usure normale », « satisfaisant », « Émaillage neuf », « Marques sur les joints ». Testimony ages badly; their paper testifying for you does not.
3. Verify before arguing. The page-by-page entry-EDL check killed the false old-tub theory before it was sent and replaced it with the stronger workmanship argument. Three adversarial reviews and three disguised opposing-counsel simulations did the same for the rest. Nothing left the building unchallenged.
4. Map the enforcement path before the merits mattered. Discovering the guarantee was an insurance policy with a three-title payment gate, then notifying the insurer and pre-committing to opposition totale, turned "they can grab the deposit" into "they must win a lawsuit first."
5. Concede CHF 100 on purpose. One small, uncertain item traded away for total credibility on everything else, with a hard settlement floor fixed in advance. EV thinking applied to yourself, not just the opponent.

The move in one line

When the machine of obligations comes for you, run the whole book in order: find the claim foundation, decompose it into elements, hang each element on the right party's burden, check what the fine print is allowed to do, freeze the claim surface, cut the enforcement path, and then price the endgame like the portfolio it is.

Drills

1. The régie re-sends the convention "for regularization" with a friendly note saying the amounts are only estimates anyway. What is the correct response, and which two rules from this chapter does signing it violate?

Answer: Do not sign, and answer nothing that acknowledges the amounts. Signing would create the reconnaissance de dette (art. 17 CO) that opens door 1 of the insurer's payment gate and hands the régie an art. 82 LP title, converting a claim they must prove into a debt you must fight out of. It violates the standing rule "sign nothing" and undoes the enforcement-path analysis of chapter 9. "Only estimates" changes nothing: even a causal reconnaissance keeps all defenses alive, but you would be volunteering the burden flip.

2. A commandement de payer for CHF 4,700 arrives on a Tuesday. List the exact procedural move, its deadline, its legal basis, and what happens to the insurer's gate if you miss it.

Answer: File opposition totale within 10 days of notification (art. 74 LP), by declaring it to the officer at service or at the office des poursuites, ideally in writing with proof. Miss the deadline and the poursuite becomes unopposed, which satisfies door 2 of the UNIQA CGA: the insurer can pay the régie and then recover the full amount from you, converting a near-zero claim into a real debt through pure procedural default. It is the only hard deadline in the case.

3. Suppose the exit EDL had marked the living-room wall « Dégâts » instead of « Usure normale », with a photo. Rebuild the régie's claim through the art. 267 checklist: which elements would now be satisfied, and which would still be open?

Answer: Element 1 (a degradation exists) and arguably element 2 (exceeds normal wear) would now rest on a signed contemporaneous record, so the tenant's own signature would work against him there. But elements 3 through 5 stay open and stay on the landlord under art. 8 CC: prove the damage post-dates the entry EDL of 27 Jun 2025, prove attribution to the tenant, and prove quantum via actual invoices scaled by residual value under the tableau paritaire (art. 42 CO), not a round estimate. A signed « Dégâts » entry weakens the position but does not by itself produce a single franc of liability; the checklist is conjunctive to the end.

Exercises

Self-test

Fieldwork

• Your own files · ~60 minAutopsy a past dispute of your own — any dispute, won or lost — using this chapter’s structure: the board at move zero (a dated timeline table), the magic question, the element table with per-element status, the burden ladder, the enforcement path, the endgame EV. Write it down. The discipline of the write-up is the exercise; most people discover the move they missed.
• Drafting · ~45 minWithout rereading the chapter, draft the skeleton of the Florissant response letter: the foundation question, the element walk built on the régie’s own EDLs, the art. 267a CO forfeiture point, the conditional art. 31 CO declaration, a deadline and an open door. Then compare against the five decisions that won it, and patch what you forgot.

What this chapter gives you. A forward-looking crash investigation. The case is fictional but built from real failure modes: Ekeko signs a CHF 48,000 fixed-scope ML pipeline engagement with a Geneva scale-up, the client kills it at 60%, disputes quality, refuses part of the fee, and claims ownership of all code including Ekeko's pre-existing library. We run the mess through the book's models (qualification, termination, defects, burden, enforcement, IP), then rewind and show the six clauses that would have prevented each failure, ending with a checklist for Ekeko's real template. Everything from Part I fires at once.

The facts, as they arrive

• January: Ekeko Sàrl signs a two-page agreement with Helvetiq Data SA (Geneva) titled "Consulting Agreement". Scope: "design and build an ML feature pipeline and model serving layer". Price: CHF 48,000, invoiced "as the work progresses". No milestones, no acceptance procedure, no IP clause, no termination clause.
• March, roughly 60% delivered: the client emails, "We are stopping here. We will pay for hours used."
• April: Ekeko invoices CHF 31,000 (work done plus two weeks of now-idle reserved capacity). The client pays CHF 18,000, and replies that the delivered pipeline "does not work in production", so nothing more is owed.
• May: the client's lawyer adds a new claim: under the agreement Helvetiq owns "all code produced in connection with the project", including ekeko-featurelib, the internal library Ekeko built over two years and reused here.

Run the magic question from ch. 2: who wants what from whom, based on what? Ekeko wants CHF 13,000 from Helvetiq based on the contract (art. 97 CO plus the fee rules of the applicable contract type). Helvetiq wants a price reduction based on defects, and an IP assignment based on almost nothing, as we will see. Three claims, three foundations, three burdens.

Step one: qualification decides the whole tree

The contract says "Consulting Agreement". The label is irrelevant: qualification follows the actual content of the obligations, not the title the parties typed (interpretation per art. 18 CO, real and common intent over wording). The two candidate specializations of the general rulebook (see ch. 5) are:

• Mandate (mandat, art. 394 ss CO): you owe diligent activity in the client's interest, not a guaranteed result. Think doctor, lawyer, advisor.
• Contract for work (contrat d'entreprise, art. 363 ss CO): you owe a defined result, an ouvrage. Think builder, and per Federal Court practice, the developer of specified software delivered as a finished object.

Swiss courts qualify fixed-scope software development with a defined deliverable as contrat d'entreprise or a mixed contract with entreprise rules for the build; open-ended advisory and iterative work leans mandat. "Design and build an ML pipeline for CHF 48,000" smells strongly of a result. This one qualification flips almost every downstream rule:

That last row is the precedence question from ch. 4, and it is where the honest bad news lives. Hold it for the rewind.

The termination claim

If entreprise: "we are stopping here" is a valid move, the maître can always withdraw before completion, but art. 377 CO prices it: pay for the work done and fully indemnify the contractor, including the profit margin on the cancelled 40%. Ekeko's CHF 31,000 invoice was, if anything, modest. (Claiming Ekeko was late or the work doomed would instead invoke art. 366 CO, withdrawal for delay or anticipated defective performance, with the burden of those elements on the client, art. 8 CC.)

If mandat: art. 404 al. 1 CO lets either side revoke at any moment, and the Federal Court treats this as mandatory law that no clause can neutralize (ATF 115 II 464). The only money for termination itself is art. 404 al. 2 CO, damages for revocation en temps inopportun, at an inopportune moment, and case law limits that to negative interest: costs incurred in reliance, not the lost profit on the unbuilt 40%. Two weeks of reserved, now-idle capacity is a textbook art. 404 al. 2 CO item. The margin on the remaining work is gone.

So the qualification fight is worth real money: roughly the profit share of CHF 17,000. This is why chapter 5's model of general rules and specialized overrides is not academic.

The "does not work" claim

Now run ch. 3. Price reduction is a right-destroying defense against Ekeko's fee claim, so Helvetiq carries the burden of proving the defect, art. 8 CC. "Does not work" is not a fact, it is a conclusion; they must prove specific non-conformities against the agreed spec. And the two-page contract barely has a spec, which cuts both ways: hard for them to prove deviation, hard for Ekeko to prove conformity.

If entreprise, the defect regime adds procedure on top of burden. The client must inspect the work as soon as feasible and give notice of defects, the avis des défauts, promptly (art. 367 al. 1 CO); hidden defects must be notified immediately upon discovery (art. 370 al. 3 CO). Silence plus use equals acceptance, and acceptance discharges the contractor for unnotified defects (art. 370 al. 1 and 2 CO). Helvetiq took delivery of the 60% in March, said nothing concrete, and first alleged "does not work" in April, after the invoice. A court will ask: where is the timely, specific defect notice? Defect claims also prescribe two years from acceptance for a movable work (art. 371 al. 1 CO), though prescription is the least of their problems here.

Proven defects would open the art. 368 CO remedies: repair, price reduction, or refusal of unusable work. Not on the list: "we own everything and pay nothing."

If mandat, there is no formal defect-notice guillotine; instead Helvetiq must prove a breach of Ekeko's duty of diligence (art. 398 al. 2 CO) plus damage plus causation, the full art. 97 CO element checklist, with the fault element presumed against Ekeko but everything else on them. Vague dissatisfaction still loses.

Meanwhile the missing acceptance procedure hurts Ekeko too: with no signed recette (acceptance protocol) per milestone, "what was delivered and was it accepted" becomes a swearing contest. What cannot be proved counts as false for whoever needed the fact, and both sides needed facts they cannot cheaply prove.

The money: enforcement path

Ekeko is owed CHF 13,000 on its own numbers. Substance without procedure is decorative (see ch. 9), so the path is the one you already walked in your deposit dispute, from the creditor's side this time:

1. Mise en demeure: written payment demand with a deadline; default interest at 5% runs from default (arts. 102 and 104 al. 1 CO).
2. Réquisition de poursuite at the Geneva debt collection office; the office serves a commandement de payer (art. 69 LP). This costs little and requires no proof.
3. The client will file opposition within 10 days (art. 74 LP), exactly as you once did.
4. To lift it, Ekeko needs mainlevée provisoire, which requires a signed acknowledgment of debt (art. 82 LP). A two-page contract without a price-per-milestone schedule is a weak reconnaissance de dette for a disputed partial amount, so realistically Ekeko must sue on the merits (action under the CPC, likely the simplified procedure since the amount is below CHF 30,000, art. 243 CPC).
5. Fee claims prescribe: five years for this category of professional work claims (art. 128 ch. 3 CO), so there is time, but interest and legal costs are the real clock.

Settlement expected-value logic (chapter 12's frame): claim strength is provability (medium: no acceptance protocols) times enforceability (good: solvent scale-up) times opponent's costs (high: lawyers cost more than CHF 13,000). Expect roughly 60 to 75 cents on the franc, precisely the tax the missing clauses imposed.

The code: who owns what

Here the client's lawyer is bluffing, and the default rules say so:

• Copyright arises in the author and moves only by contract (transferability, art. 16 LDA). Switzerland has no general work-for-hire doctrine for contractors. Nothing was assigned in writing, so Ekeko still owns the copyright in everything it wrote.
• The one statutory exception is art. 17 LDA: rights in software created by an employee in the course of employment duties belong to the employer. Ekeko is a contractor, a separate Sàrl, not an employee. Art. 17 LDA does not apply, full stop.
• At most, the client can argue an implied license to use the delivered code for the contract's purpose, which follows from the contract's aim (pertinence theory in practice), and even that is contested terrain when the fee is unpaid.
• ekeko-featurelib was created before and outside the project. There is no conceivable foundation for a transfer claim on it. They want an assignment based on a clause that does not exist. Claim denied at the checklist's first element.

So the IP position is Ekeko's best card, and also its scariest exposure in mirror image: a well-lawyered client template would have contained a blanket assignment ("all code produced in connection with the project"), and without a carve-out that wording plausibly captures modifications to the pre-existing library. The missing clause cost Helvetiq this time. Next time, the missing carve-out could cost Ekeko its core asset.

Rewind: the six clauses that would have prevented this

Precedence check first (ch. 4): everything below is dispositive and freely draftable, except where flagged.

1. Qualification recital. "The parties agree that Ekeko owes defined deliverables described in Annex A, accepted per clause 3, and intend this agreement as a contract for work (art. 363 CO) for those deliverables." Courts qualify by substance, not labels, so this recital is evidence of intent, not a switch you flip. But combined with a real deliverable structure it makes the entreprise qualification, and with it art. 377 CO full indemnity, very hard to dislodge.

2. Milestone acceptance. "The work is delivered in milestones per Annex A. Each milestone is accepted when the acceptance tests in Annex B pass, or 10 business days after delivery if the client raises no specific written defects. Acceptance of a milestone makes its invoice due and bars later complaints about that milestone except hidden defects notified immediately upon discovery (art. 370 al. 3 CO)." This converts "does not work" from a swearing contest into a lookup: which milestone, which test, which notice, which date. It is the single highest-value clause in the list.

3. Termination compensation, with honest fragility. "The client may terminate for convenience with 15 days' written notice, paying for all work performed plus 30% of the fee for unstarted milestones as agreed compensation." If the contract is qualified entreprise, this simply concretizes art. 377 CO and is solid. If a court requalifies it as mandat, art. 404 CO is mandatory (ATF 115 II 464) and case law strikes down termination penalties that effectively obstruct free revocation; a moderate compensation framed as covering reliance costs has better odds than a fee-forfeiture, but no drafting fully escapes art. 404 CO. Say this plainly in your own head: this clause is a probability booster, not a guarantee. The recital in clause 1 is what carries it, because it keeps you out of mandat territory altogether.

4. IP assignment with carve-out. "Upon full payment, Ekeko assigns to the client the copyright in the project-specific deliverables listed in Annex A (art. 16 LDA). Excluded from any assignment are Ekeko's pre-existing tools, libraries, and general know-how, including modifications and extensions to them; the client receives a perpetual, non-exclusive, non-transferable license to use these as embedded in the deliverables." Three load-bearing parts: the upon full payment trigger (unpaid client, no IP), the carve-out with its modifications extension (protects ekeko-featurelib even where project work touched it), and the license that makes the carve-out commercially acceptable to the client.

5. Liability cap respecting art. 100 CO. "Ekeko's aggregate liability is capped at the fees paid under this agreement; liability for indirect and consequential damages is excluded. These limits do not apply to unlawful intent or gross negligence (art. 100 al. 1 CO)." A cap that pretends to exclude gross fault is void on that point, and a court finding one void term reads the rest with hostile eyes. Draft the mandatory floor into the clause yourself; it costs nothing and signals competence.

6. Payment schedule with default interest. "Each milestone is invoiced on acceptance, payable within 20 days. Late payment bears interest at 5% (art. 104 al. 1 CO) without further notice, the due date serving as mise en demeure (art. 102 al. 2 CO). Ekeko may suspend work while any accepted invoice is overdue (art. 82 CO by analogy for the exchange of performance)." Bonus: a signed contract with a fixed amount per accepted milestone is a far stronger reconnaissance de dette for mainlevée provisoire (art. 82 LP), which upgrades your enforcement path from full lawsuit to summary proceeding.

Ekeko template checklist, one page

Plus two operational habits no clause replaces: get every acceptance in writing, even a one-line email reply, and keep ekeko-featurelib in its own separate repository with its dated change history intact, because provenance is proof (art. 8 CC applies to IP disputes too).

The move in one line

Qualify the contract before you sign it, because mandat or entreprise is the branch that decides who can walk away, what walking away costs, and who must prove that the code works.

Drills

1. Helvetiq's lawyer argues the contract was a mandate all along, pointing at the "Consulting Agreement" title and the hourly-flavored "as the work progresses" invoicing. What is the strongest single fact in the fictional contract for Ekeko's entreprise qualification, and what does winning that argument change in francs?

Answer: The fixed price of CHF 48,000 attached to a defined deliverable points to a promised result, an ouvrage (art. 363 CO), and courts qualify by real content, not title (art. 18 CO). Winning it puts the termination under art. 377 CO rather than art. 404 CO: work done plus full indemnity including lost profit on the cancelled 40%, instead of at most reliance damages for untimely revocation (art. 404 al. 2 CO). Roughly the margin on CHF 17,000 of cancelled work rides on it.

2. Suppose the roles were reversed: the client's template contained "Client owns all code produced in connection with the project" and Ekeko signed it unmodified, then used and extended ekeko-featurelib during the build. What does Ekeko still own, and what has it plausibly lost?

Answer: Copyright transfers only to the extent agreed (art. 16 LDA), and assignment clauses are interpreted against overreach, so the pre-existing library as it stood before the project stays Ekeko's: it was not "produced in connection with the project". But the extensions and modifications written during the project were, so the client has a plausible claim to those, which can fragment the library's ownership and poison its reuse. That is exactly what the carve-out clause with its "including modifications and extensions" language prevents. Art. 17 LDA never enters the analysis, because Ekeko is a contractor, not an employee.

3. The client ignores Ekeko's CHF 13,000 invoice entirely. Map the cheapest path to actual money, naming each procedural step and where it can die.

Answer: (1) Written mise en demeure with deadline, starting 5% default interest (arts. 102, 104 CO). (2) Réquisition de poursuite, commandement de payer served (art. 69 LP): cheap, no proof needed. (3) Client files opposition within 10 days (art. 74 LP): the path forks. (4) With a signed acceptance protocol plus milestone price, Ekeko seeks mainlevée provisoire on the contract as reconnaissance de dette (art. 82 LP): fast summary proceeding. Without it, mainlevée likely fails for the disputed amount and Ekeko must sue on the merits, simplified procedure below CHF 30,000 (art. 243 CPC): slower, costlier, and the point where settlement expected value usually says take 70 cents. The claim survives five years (art. 128 ch. 3 CO), but the leverage decays much faster.

Exercises

Self-test

Fieldwork

• Drafting · ~90 minWrite your actual template: the six clauses adapted to your service, plus an Annex A (deliverables) and Annex B (acceptance tests) that fit what you really sell. Then run Chapter 4’s clause-audit checklist against your own draft — name the displaced rule and its tier for every clause you wrote. Finish by pricing one paid lawyer review of the result: insurance amortized over every future engagement.
• Repo hygiene · ~20 minPut your pre-existing library and tooling in its own repository with its dated history intact, separate from any client work, and add the carve-out sentence (“including modifications and extensions”) to your template’s IP clause. Provenance is proof — art. 8 CC applies to IP disputes too.

You now run the Swiss machine fluently: find the norm, decompose the elements, allocate the burden, compute the consequence. This chapter deliberately takes that machine away and drops you into the other legal family, where your method is not just unfamiliar but wrong.

What this chapter gives you. The other legal family, and the same contract dispute resolved twice so you can watch the two methods diverge. You will see why an English lawyer answers "which article applies?" with a stack of decided cases; what stare decisis actually binds and how distinguishing escapes it; why English law has no consideration-free contracts, no general good-faith duty, and no sympathy for your art. 2 CC instincts; why Anglo-Saxon contracts are forty pages long and yours are four; and what changes the day you sign a SaaS agreement governed by English or New York law — which, if you work in tech, you will. Chapter 11 showed Swiss precedent as a persuasive cache over authoritative code (see ch. 11); the common law is the inversion — the cache is the source, and statutes are patches on top. An honest-limits section keeps the contrast from hardening into caricature.

Two build systems

Chapter 0 stated the deep split in one paragraph (see ch. 0); here is the mechanism underneath it.

The Swiss codes were designed: drafted as a coherent system, by identifiable authors, adopted on a known date, versioned on Fedlex. When you face a dispute, you resolve it top-down: locate the norm, check its elements, apply its consequence. Cases matter, but as interpretations of the code — expensive reasoning stored for later reuse, formally binding on no one.

The common law was never designed. It accreted: eight centuries of English court decisions, each resolving one concrete dispute, each becoming raw material for the next. There is no master text to decompose; the law of contract is the emergent behavior of thousands of decisions, summarized (unofficially, non-bindingly) in treatises. Statutes exist, but they carve exceptions into the case law rather than replace it — patches, not a rewrite.

This inverts the direction of authority. In Switzerland a decision is authoritative because it correctly applies the code; in England a rule is law because a court of sufficient rank decided it, and stare decisis — "to stand by things decided", the doctrine that precedent formally binds — makes the decision obligatory for every court below and, with narrow exceptions, for the deciding court itself. The binding part is the ratio decidendi: the legal rule necessary to the outcome on those facts. Everything else the judge said is obiter dictum — remarks in passing, persuasive only. So common-law reasoning is a retrieval problem, not a decomposition problem. Set the two procedures side by side:

Note what the English step 3 implies: the rule never fully detaches from its facts. A Swiss ATF states its principle abstractly, almost like a statute amendment; a common-law ratio is bounded by the facts that generated it, which is why distinguishing — arguing your facts differ in a way that matters, so the precedent does not govern — is the central courtroom skill. Where a Swiss lawyer argues "the element is not satisfied," an English lawyer argues "that case is not this case." Same job, entirely different procedure.

One more inversion worth logging: the Swiss system is honest about judicial creativity (a judge facing a gap openly legislates, art. 1 al. 2 CC) and stingy with formal bindingness. The common law is the reverse — precedent binds formally, while judges maintain the fiction of merely "declaring" law that always existed, which is why overruling a precedent is doctrinally fraught.

The same dispute, twice

Concrete case, Ekeko-flavored. Ekeko orders a specialized compute server from a vendor, delivery promised within ten days, needed for a client project. The vendor delivers five weeks late. The client project collapses; Ekeko loses CHF 40,000 in fees and wants the vendor to pay.

The Swiss run. You know this machine (see ch. 2). Who wants what from whom, based on what? Ekeko wants CHF 40,000 in damages from the vendor, based on art. 97 al. 1 CO (non-performance or improper performance of an obligation), routed through the debtor's-default rules of art. 102 ss CO for late performance. Decompose:

Then the reduction layer: art. 43–44 CO let the judge scale damages by degree of fault and cut them for Ekeko's own contributory choices (no buffer in the plan, no substitute rental — the mitigation duty bites). The "too indirect?" question lives inside adequate causation: was this loss, in the ordinary course of things, a foreseeable kind of consequence of late delivery? Every step is an element with an address, and the fight will be about proof and quantum, exactly as Chapter 3 trained you (see ch. 3).

The English run. Now re-govern the contract under English law. The lawyer does not reach for a code — there is none to reach for. Breach is conceded (late is late; fault is irrelevant — common-law contract liability is strict, with no exculpation branch). The entire battle is remoteness of damage: is the CHF 40,000 the kind of loss the law lets you recover? And the answer is not an article. It is a case: Hadley v Baxendale (1854).

The facts, which every common lawyer carries around the way you carry art. 8 CC: a Gloucester mill's crankshaft broke; the mill sent it via a carrier to engineers in Greenwich as a pattern for a new one; the carrier delivered late; the mill stood idle and sued for the lost profits. The Court of Exchequer refused them, and Baron Alderson's judgment laid down the rule that still governs: breach damages are recoverable only if the loss (1) arises naturally, in the usual course of things, from the breach, or (2) was within the reasonable contemplation of both parties at contracting as the probable result of breach — in practice, the special circumstances were communicated. The mill lost because the carrier had no reason to know it would stand idle: for all the carrier knew, there was a spare shaft.

Watch the reasoning style: the English lawyer does not decompose Hadley into elements; she lines up the cases that applied and refined it and argues by resemblance and distinction. The key refinement is Victoria Laundry v Newman Industries (1949): a boiler delivered five months late to a laundry; the Court of Appeal allowed ordinary lost profits — a boiler supplier can contemplate that a laundry launders for money — but refused the exceptionally lucrative government dyeing contracts the seller knew nothing about.

Applied to Ekeko: the vendor's counsel argues limb 1 covers at most some generic loss of use, and that the CHF 40,000 engagement is Victoria Laundry's lost dyeing contract — an unusually valuable opportunity never communicated, hence outside limb 2, hence irrecoverable. Ekeko's counsel distinguishes: unlike the mill and the laundry, Ekeko told the vendor at ordering time that the machine was for a specific client project with a hard start date (if it did — now you see why what you write in the ordering email matters under English law). The judge decides which precedent's facts this case sits closer to; lower courts are bound by the Court of Appeal's ratio in Victoria Laundry, and nobody re-derives the rule from purpose the way a Swiss court might re-run the four methods. The rule is where the cases left it.

Same dispute, same money, and notice what moved: in Zurich the fight is proof and quantum inside a known checklist; in London it is which decided case this one resembles. A substantive delta hides inside the method delta, too: Swiss law filters remote losses through adequate causation and judicial discretion, assessed at trial; English law asks what was in both parties' contemplation at contract formation. Under English law you fix your damages exposure on signing day, by what you disclosed — a single difference that drives page after page of Anglo-Saxon drafting.

Formation, briefly: the smoke ball

Formation diverges too. Art. 1 CO makes a contract from the mutual, concordant manifestation of will — offer and acceptance, nothing more. English law demands a third element with no Swiss counterpart: consideration — each side must give something of value in exchange; a naked promise, however seriously meant, binds no one. (Swiss law happily enforces gratuitous promises; at most a form requirement intervenes, e.g. art. 243 CO for promises of gift.)

The classic formation case is Carlill v Carbolic Smoke Ball Co (1893). A company advertised £100 to anyone who used its smoke ball as directed and still caught influenza, adding that £1,000 sat deposited in the bank to show sincerity. Mrs Carlill used it, caught flu, and claimed. The Court of Appeal built her contract out of case-law atoms: the advertisement was no mere puff (the deposit showed intent) but a unilateral offer to the world, accepted by performing its conditions; her consideration was the inconvenience of using the ball. A Swiss lawyer resolves the same pattern in one move — public promise (art. 8 CO), or simply offer and acceptance by conduct (art. 1 CO) — and never asks the consideration question, because the code never coded it.

The deltas, tabulated

Three rows deserve a second look. Good faith: the absence of an art. 2 CC equivalent is the most disorienting delta. English law prizes certainty — commercial parties may rely on the text they signed, ruthlessly if necessary — so your Chapter 7 instinct ("technically flawless but shocking will get caught") is simply false there; what you signed is what you get. Discovery: being sued in the US can force disclosure of your internal emails, chat logs, and drafts, at staggering cost — the process itself is a weapon, and settlement pricing (see ch. 12) must include the discovery burn on both sides. Costs: the American rule — each side pays its own lawyers, win or lose — reshapes the settlement market: a defendant can be economically right to settle a meritless claim because winning still costs six figures.

Why the contracts are forty pages long

Now you can answer the question every Swiss reader asks when a US SaaS agreement first lands in their inbox: why is this thing so long?

Because the drafter starts with no ready-made layer of defaults. Your four-page Swiss mandate contract is short because the CO supplies thousands of dispositive rules — termination regimes, warranty schemes, notice periods, risk allocations — that apply automatically wherever your text is silent (see ch. 4). A common-law drafter writing under English or New York law has no such fallback layer — or rather one made of scattered case law, expensive to retrieve and fuzzy at the edges. So the contract carries everything on board: every definition, every remedy, every termination trigger, every risk allocation is written out in full, because the document must be self-contained against a background of cases rather than resting on a code.

The same logic explains the famous entire agreement clause ("This Agreement constitutes the entire agreement between the parties and supersedes all prior negotiations, representations and understandings…"). It reinforces the parol evidence rule — the common-law doctrine that a written contract cannot be varied by evidence of prior side-agreements — and kills claims based on things said during negotiation. That email where the sales rep promised 99.99% uptime? The entire agreement clause just deleted it. Under Swiss law, art. 18 CO (real common intent trumps inaccurate wording) and art. 2 CC give such clauses far less bite; under English law they are load-bearing walls.

The practical checklist when you sign under English or New York law:

1. Everything you rely on must be in the document. Pre-contractual assurances die at the entire agreement clause. Negotiate the promise into the text or treat it as nonexistent.
2. Disclose your stakes at formation. Hadley limb 2 means your recoverable damages are capped by what the counterparty could contemplate on signing day. If late delivery kills a CHF 200k project, say so in writing before signing — and expect them to demand a liability cap in response.
3. Read the exclusion clauses as the real law of the contract. "In no event shall either party be liable for indirect, consequential or special damages, or loss of profits" — that boilerplate does Hadley-adjacent work, and courts enforce it between businesses. No art. 2 CC will rescue you from a harsh but clear allocation.
4. Do not assume a fairness backstop. No general good faith, no judicial reduction of your bad bargain, and (post-Makdessi) even fierce liquidated-damages clauses can stand if a legitimate interest supports them. The text is the territory.
5. "Endeavours" language is case-law-defined. "Best endeavours" and "reasonable endeavours" carry different, litigated meanings in English law. Words you would read as synonyms are distinct precedent clusters.

The bridge: choosing your law

None of this is imposed on you; it is chosen, clause by clause. Private international law lets commercial parties pick their contract's governing law and forum (in Switzerland, art. 116 LDIP — Loi fédérale sur le droit international privé, the private-international-law act). The last two clauses of every cross-border contract — governing law and dispute resolution — are therefore not boilerplate but the selection of the entire legal machine the contract will run on. The standard bridge between the families is arbitration: a private tribunal, seated in Geneva, Zurich, London, or Singapore, applying whichever law the parties chose, its award enforceable in some 170 states under the New York Convention of 1958 — coverage no state-court judgment can match. For a Swiss company facing a US counterparty, "Swiss law, arbitration in Geneva" versus "New York law, New York courts" is the difference between disputing on your home machine and on theirs, discovery, jury risk, and the American costs rule included. Price that before signing (see ch. 12 — every term of the strength formula changes with the regime).

Honest limits

The contrast above is real but drawn at maximum voltage; four corrections keep it honest.

Statutes matter in the common law too. English sale-of-goods disputes run substantially on the Sale of Goods Act 1979; US commercial law is heavily codified in the Uniform Commercial Code. A US lawyer reads UCC Article 2 almost like a civilian — but through the case law interpreting it, which keeps the method common-law-shaped.

Good faith is creeping in. The UCC imposes good faith in the performance and enforcement of every contract within it (UCC §1-304 — performance and enforcement, note, not negotiation), and English courts have begun implying good-faith duties into long-term "relational" contracts in first-instance decisions. The doctrinal wall stands, but it leaks — as does consideration: promissory estoppel (Denning's High Trees case, 1947) enforces some promises no consideration supports.

Civil judges make law de facto. Chapter 11 already confessed this: the Swiss "persuasive cache" behaves like binding precedent at the 95th percentile, and whole fields (the details of employment termination, the contours of usure normale) are effectively judge-built. The families differ in the direction of authority and the unit of reasoning more than in who actually writes the operative rules.

Convergence instruments exist. The CISG (the UN sales-law convention) gives Switzerland and the US a shared codified sales law for cross-border goods — the UK, notably, never joined — and international arbitration has evolved a procedural middle style, its documentary disclosure broader than Swiss courts allow and far narrower than US discovery. The core remains split all the same: ask an English lawyer "which article?" and a Swiss lawyer "which case?", and each reaches for the tool the other considers an accessory.

The chapter in one line

Common law resolves disputes by retrieving and distinguishing binding decided cases, Swiss law by decomposing coded articles — so under English or New York law your protection is only what the signed text says against the case-law background: no code defaults, no consideration-free promises, no art. 2 CC safety net.

Drills

1. Ekeko signs a New York-law SaaS agreement after the vendor promised by email that data residency would be Switzerland-only. The signed contract is silent on residency and contains an entire agreement clause; the data ends up in Virginia. Under the Swiss method you would reach for art. 18 CO or art. 2 CC — what happens to those instincts here, and what should Ekeko have done?

Answer: Both instincts fail. The entire agreement clause, backed by the parol evidence rule, exists precisely to exclude pre-contractual assurances, and no general good-faith duty patches the harsh result (misrepresentation claims exist, but as a separate, harder, often contractually-excluded route). A promise not written into the signed document should be treated as nonexistent; Ekeko should have negotiated the residency commitment into the agreement itself — under a case-law regime, the signed text is the whole machine.

2. Re-run the late server dispute and assume Ekeko wrote in the purchase order: "This machine is required for the Helvetia Re engagement starting 1 March; delay will cost us the engagement (approx. CHF 40,000)." How does this sentence change the analysis under English law, and does it change anything under Swiss law?

Answer: Under English law it is potentially outcome-flipping: it moves the CHF 40,000 from Victoria Laundry's irrecoverable "exceptional contract" category into Hadley's second limb — special circumstances communicated at formation, hence within both parties' reasonable contemplation on signing day. Under Swiss law it helps but less dramatically: recoverability runs through adequate causation and the art. 42–44 CO reduction layer at trial, where the disclosure is good evidence of foreseeability rather than the hinge of a formation-time test. Expect a rational English-law vendor to answer such a sentence by demanding a liability cap — the negotiation working as designed.

3. An English colleague reads Chapter 11 and objects: "Your Swiss courts follow ATF precedent 95% of the time and your judges openly legislate under art. 1 CC — so the families are the same, you just won't admit your judges make law." Give the honest two-part answer.

Answer: Concede the convergence: Swiss precedent behaves near-bindingly in expectation, Swiss judges build whole doctrines, and common-law systems carry major codes (Sale of Goods Act, UCC) — neither family runs pure. What still differs is the direction of authority and the unit of reasoning. A Swiss decision draws its legitimacy from correctly interpreting the code, states its rule abstractly, and can be attacked head-on through the ratio legis (see ch. 11); a common-law ratio is itself the law, stays welded to its facts, formally binds lower courts, and is escaped by distinguishing rather than re-deriving purpose. Same amount of judge-made law, opposite architectures — hence the visibly different working methods and contracts.

Exercises

Self-test

Fieldwork

• Audit · ~30 minOpen the terms of service of a SaaS product you actually use. Find three clauses: governing law and forum, the entire agreement clause, and the consequential-damages exclusion. For each, write one line on what it deletes compared to your Swiss instincts (art. 18 CO real intent, art. 2 CC good faith, the CO’s default remedies). Notice which legal machine you are already running on without having chosen it.
• Reading · ~30 minRead a case summary of Hadley v Baxendale (1854) and Carlill v Carbolic Smoke Ball Co (1893) from any reputable source. For each, extract the ratio in one sentence welded to its facts — the mill with no spare crankshaft, the £1,000 bank deposit showing sincerity. That fact-bound unit of authority is exactly what a Swiss considérant is not; feel the difference once and Chapter 15’s table becomes permanent.

What this chapter gives you. The revision history behind the codes. The CC and CO did not appear from nothing in 1912: they are the latest stable version of a system that has been in development for about four thousand years, and every major design decision in them — publication, fixed sanctions, general parts, plain drafting, versioned consolidation — was a fix for a specific failure some earlier era built and regretted. This chapter runs the history as a systems story: why law got written down at all (Hammurabi), how it became a discipline with professionals and literature (Rome), how the great consolidation nearly vanished and then rebooted Europe (Justinian, Bologna), why England split off (1066), why the continent suddenly produced codes around 1800, and how Switzerland — late, fragmented, lucky in its architect — ended up with the citizen-readable codes this book is about. None of these dates will win you a deposit dispute — but knowing why the machine is shaped this way makes every chapter before this one less arbitrary.

Why write the rules down at all

Start with the problem, unchanged since the Bronze Age: a ruler, an official, or a stronger neighbor decides your case however they like, and you have no way to say that is not the rule. The rule lives in someone's head, which means there is no rule, only power with good posture.

The first known fix is publication. Around 2100 BC, the Sumerian Code of Ur-Nammu — the oldest surviving law collection, fragments of clay tablet — already lists offenses with fixed tariffs in silver. But the famous release is Babylon, circa 1754 BC: the Code of Hammurabi, some 282 provisions carved into a two-meter basalt stele, erected in public (the surviving copy, carried off to Susa as loot, was dug up in 1901 and stands in the Louvre). Two decisions on that stone found everything this book describes.

First, the form. Hammurabi's provisions are casuistic: if a man does X, then Y follows. A builder's house collapses and kills the owner: the builder is put to death. Condition, consequence. That is the Tatbestand → Rechtsfolge shape you learned to decompose in chapter 1 (see ch. 1), already stable thirty-seven centuries before art. 97 CO. Law has been if-then from the very first line.

Second, the fixed sanction. Modern readers flinch at lex talionis — the "law of retaliation," eye for an eye — and read savagery. Read it as an engineer instead: it is a cap. In a world of clan feud, the default response to a lost eye was open-ended escalation — an eye, then a life, then a village. Talion publishes a maximum: one eye costs exactly one eye, no more, and everyone can read the tariff in advance. It is calibrated proportionality replacing unbounded revenge, plus the deeper guarantee that the sanction is fixed before the case, not invented after it by whoever is angriest. The calibration was crude — sanctions scaled openly with social class — but the principle, that published rules bind the ruler's own courts, is the one Fedlex still runs on; hold that thought.

Readable by the governed: the Twelve Tables

Jump to Rome, ~450 BC, because the next failure mode is subtler: law can be written and still be hidden. Early Roman law was administered by patrician priests who knew the forms and shared them with no one. The plebeians — the party that kept losing on procedure it could not inspect — went on political strike until the rules were published. The result was the Twelve Tables (traditionally 451–450 BC): the customary law drafted by a commission, posted in the Forum, and drilled into schoolboys for centuries. The originals are lost (Rome was sacked in 390 BC); we know them from quotations.

The content was archaic. The move was permanent: law is legitimate only if the governed can read it. Every later chapter of this story is a rerun of the plebeian demand. When chapter 0 said Swiss law is public, versioned, and addressable (see ch. 0), it was cashing a check written in the Forum.

Rome's real invention: law as a discipline

What made Rome the upstream source of half the planet's legal systems was not the Twelve Tables. It was what Rome built around its law over the next six centuries: a maintenance culture.

Three components. The praetor, the annually elected magistrate who ran the courts, published an edict at the start of his term listing the actions he would grant. Over generations this edict became a correction layer — ius honorarium, magistrate-made law — which the Romans themselves described as existing to aid, supplement, and correct the rigid old ius civile (the citizens' core law). A living system with a formal mechanism for layering fixes on top of a stable core: chapter 10's norm stack has ancestors (see ch. 10).

The jurists: private experts who gave reasoned opinions (responsa) on live cases, argued with each other in writing, and built up a professional literature — Gaius, Papinian, Ulpian, Paul. This is the moment law becomes a discipline rather than a decree. No jurists, no doctrine; no doctrine, nothing for Bologna to teach a millennium later.

And the textbook: around 161 AD, Gaius wrote the Institutes, a beginner's course that organized all private law into persons, things, and actions. That three-part schema is still the load-bearing skeleton of civil codes — the CC's books on persons, family, inheritance, and property are its descendants.

The consolidation release: Justinian

By the 500s AD the Western empire was gone and the law was a thousand years of accumulated statutes, edicts, and juristic writing — mutually contradictory, physically scattered, practically unusable. The Eastern emperor Justinian ordered the great merge. Between 529 and 534, a commission under Tribonian delivered what later ages called the Corpus Iuris Civilis ("body of civil law"): the Codex (imperial legislation, deduplicated and current), the Digest (533 AD — the juristic literature, roughly 1,500 books read, excerpted, and reduced to a twentieth of their bulk in 50 organized books), and the Institutes (a fresh student textbook, built openly on Gaius). Everything not merged was retired: superseded sources lost legal force.

Call it a consolidation release — everything current merged into one authoritative edition — but note where the analogy is dishonest. An engineer consolidates in order to keep building; Justinian tried to freeze the system, banning commentary on the Digest so that no new doctrinal layer could grow. It did not work — it never does — but the intent matters: a final edition, not a foundation.

History's joke was better: in the West the Corpus was almost immediately lost, surviving five centuries as a handful of unread manuscripts — one nearly complete Digest copy carried it across the gap — while Europe ran on fragmented, oral custom, roughly back at the pre-Hammurabi baseline.

The long reboot: Bologna and the ius commune

Around the 1070s the Digest resurfaced in Italy, and within a generation a teacher named Irnerius was lecturing on it at Bologna — conventionally dated ~1088 and counted as Europe's first university. What followed is the strangest episode in this history: an entire continent adopted, as living law, a body of law with no living experts, no empire behind it, and a five-hundred-year gap in its history — because it was simply better than anything then in use: complete, reasoned, and written down.

The glossators annotated the recovered text line by line (Accursius's standard gloss, ~1230, ran to some hundred thousand notes); the commentators — Bartolus above all, 14th century — went further and adapted the Roman material to feudal, commercial, and municipal reality: an inherited system reworked, piece by piece, for a world its authors never saw. Students from all over Europe studied it at Bologna and its imitators alongside canon law, the church's parallel system (systematized by Gratian around 1140), then went home as judges, notaries, and chancellors. Out of this spread the ius commune — the "common law" of continental Europe in the original sense: a shared, university-taught, Latin-language legal layer that local statutes and customs merely overrode in spots. The architecture — a common base layer of defaults, local overrides on top — is the one the continent has run ever since.

The split: England goes its own way

England is the counterfactual that makes the rest legible. In 1066 the Normans conquered a whole kingdom at once and built the most centralized administration in Europe — crucially, before the Bologna revival could be received. By Henry II (reigned 1154–1189), royal justice was a working national system: standardized writs that slotted each grievance into a fixed form of action, itinerant royal judges, juries, and central courts whose accumulated decisions were the law — "common law" meaning common to the realm, one law from a strong crown rather than one law from a shared university text.

So when Roman law arrived, England did not need it: it had a working system and an entrenched profession, trained in the Inns of Court on pleading and precedent rather than at universities on the Digest. The continent adopted the ius commune largely because its thousand fragmented jurisdictions had nothing comparable; England's very earliness made it immune. Software people would call this a fork, though strictly England never branched off the Roman line — it simply never joined it. Either way, the two lineages parted here and never remerged: judge-made law hardened into binding precedent (stare decisis, "to stand by what is decided"), and when the writ system's rigidity produced injustice, England delivered the fix as a separate court — the Chancellor's equity jurisdiction, a discretionary correction layer running beside the common law for centuries (the two administrations were merged only in the 1870s). Switzerland runs the equivalent inline: art. 2 CC is equity as a built-in exception handler rather than a second court (see ch. 7). The full comparison — how the two families decompose one and the same dispute — is chapter 15's job (see ch. 15); this section only explains why there are two families at all.

The age of codification

For five centuries after Bologna, the continent ran the layered system: Roman base, canon-law additions, local custom on top. Then, in one long generation around 1800, states began replacing the whole layered structure with single, systematic, vernacular codes. Why then? Four forces landed at once. Nation-states wanted one law for one territory — a sovereign cannot easily rule through a Latin professor-law it does not control. Enlightenment rationalism held that law could be derived from reason and stated clearly enough for citizens — and, in France especially, aimed to demote the despised pre-revolutionary judges from lawmakers to appliers. Printing and literacy made a mass-distributed statute book practical for the first time. And fragmentation had become intolerable: pre-revolutionary France ran on hundreds of regional customary laws in the north and Roman-derived written law in the south.

The milestones: Prussia's Allgemeines Landrecht (ALR, 1794) — roughly 19,000 paragraphs trying to pre-decide every case, the eternal over-specification failure mode. Then the one that mattered: the Code civil des Français, the Code Napoléon (1804) — compact, confident, readable, and exported at bayonet-point and by prestige across Belgium, the Netherlands, Italy, Spain's orbit, and from there Latin America. Austria followed with the ABGB (1811), concise and natural-law-flavored, still in force today.

Germany is the interesting delay. In 1814 the jurist Thibaut called for a single German civil code on the French model; Savigny's famous rebuttal argued that law grows organically from a people's history and spirit (Volksgeist) and that German legal science was not yet ripe to codify — premature standardization would freeze a half-understood system. Savigny won the argument, and his historical school spent the century doing the science (the Pandectists, mining the Digest with near-mathematical rigor). The result arrived only in 1900: the BGB, the most precisely engineered code ever written — abstract, rigorously organized, with a general part of shared rules that everything else builds on — and famously written for professors, not citizens. Hold both halves of that outcome; Switzerland is about to learn from each.

Switzerland merges late — and merges clean

Switzerland entered this era as a worst case of fragmentation with a federal twist. The 1848 constitution created the federal state but left private law cantonal: into the 1880s, Zurich had its own admired code (Bluntschli's, 1853–55), Bern its own, the western cantons ran adaptations of the Code Napoléon, and parts of the interior ran on uncodified custom. Twenty-five-plus legal systems for three million people.

Unification came in two constitutional grants and two codes. The 1874 constitutional revision gave the federation competence over the law of obligations and commercial law — the traders' law, where fragmentation hurt commerce most — and the first Code of Obligations followed in 1881 (in force 1883). Full competence over all civil law came by constitutional amendment in 1898.

Then the stroke of institutional luck: the federation had already commissioned Eugen Huber, and Huber had done the Savigny homework before drafting. His System und Geschichte des schweizerischen Privatrechts (four volumes, 1886–1893) surveyed and systematized every cantonal legal order — a full audit of everything then in force, before writing the replacement. When he then drafted the Civil Code essentially alone, he was consolidating what Swiss law already was, not imposing an invention, which is a large part of why the merge took: Parliament adopted the CC unanimously in December 1907, in force 1 January 1912, with the CO revised and re-enacted in 1911 as the CC's fifth book (the structural fact you logged in chapter 0).

Huber's second decision is the one you benefit from on every page of this book. Writing a decade after the BGB, with its abstraction on the table as the state of the art, he deliberately went the other way: short articles, ideally three paragraphs, one sentence each, marginal titles, ordinary language — a code a citizen could read, in three languages. Add his other famous choice, art. 1 CC's instruction that the judge fill statutory gaps by the rule they would adopt as legislator — trusting the human at the point of application instead of pretending the ALR's dream of total pre-specification could work — and you have the design philosophy this whole book exploits. The design also passed the harshest test available: in 1926, Atatürk's Turkey adopted the Swiss CC and CO nearly wholesale, and the machine has run there since — same design, different country.

The modern layer: treaties above, versions below

Two developments since then, both already familiar from chapter 10's norm stack. Above the codes, international harmonization: the CISG (Vienna, 1980; binding for Switzerland since 1991) quietly standardizes cross-border sales among some hundred states, and the Lugano system wires Swiss judgments into Europe — law converging by treaty where it once converged by university.

Below, the quieter revolution: the text itself became infrastructure. Every act numbered in the Recueil systématique, every amendment merged into a consolidated current version, full version history retained, three equally authentic languages, all of it publicly addressable — and since 2016 the electronic version, today served by the Fedlex platform, is the legally authoritative one. Cite "art. 267 CO" and you are resolving a stable public reference to an exact, versioned text (see ch. 10 for the registry's layout).

Squint, and Fedlex is Hammurabi's stele with a version history: the same guarantee — the rules are published, fixed in advance, and readable by everyone the ruler's courts will judge — upgraded from basalt to a continuously consolidated public text. Four thousand years of iteration, one invariant: publication is the original defense against arbitrary power.

The changelog

Software teams keep a changelog: a running record of what each version changed and why. Keep one for law — line the eras up, note what each one changed and what problem drove the change — and the pattern is hard to miss: each era's headline feature fixes the previous era's failure mode.

Read down the right column and the why of legal evolution resolves into four recurring fixes. Arbitrariness → publication (Hammurabi, the Twelve Tables, Fedlex — the same fix, three times, at increasing resolution). Fragmentation → consolidation (Justinian, the ius commune, the national codes, Huber's CC, the treaties — five times). Inaccessibility → readable codification (the plebeians' demand, the vernacular codes, Huber's three-paragraph articles). Rigidity → sanctioned flexibility (the praetor's edict, England's equity, art. 1 and art. 2 CC, and the interpretation methods of chapter 11 — every era learns that a fixed text needs a disciplined way to bend, see ch. 11). One honest limit: a changelog records intent, and much of this history was accident — Justinian's consolidation survived by one manuscript; England's entire family exists because its courts happened to mature a century early. The pattern is real, but it was selected, not planned.

That is also the answer to a question this book has begged since chapter 0: why does Swiss law let you read it the way you read code? Because being readable that way is not a coincidence of style. It is the accumulated, deliberately preserved output of forty centuries of fixing the same failure — power that will not show you its rules.

The chapter in one line

Legal history is one long repair cycle — arbitrariness fixed by publication, fragmentation by consolidation, inaccessibility by codification, rigidity by disciplined interpretation — and the CC/CO on Fedlex is simply the current stable version of Hammurabi's original guarantee: the rules, published, where the ruled can read them.

Drills

1. A colleague calls lex talionis proof that ancient law was barbaric. Make the engineering case that in 1754 BC it was a feature, and name the modern principle it prefigures.

Answer: The baseline it replaced was not modern damages law but unbounded clan revenge — an injury answered by escalation without limit. Talion publishes a fixed cap: the sanction for one eye is exactly one eye, known in advance, no more. That is proportionality between wrong and sanction, plus the deeper rule — still the core of legality — that sanctions are fixed by published law before the case, not invented after it by the powerful. The barbarity was in the tariff's currency, not in the mechanism.

2. Your English colleague cites cases at you; you cite articles at her. Trace each habit to its historical root, and explain why England never adopted the Roman base the continent runs on.

Answer: Her habit descends from 1066: a precociously centralized crown built national courts whose accumulated decisions were the law, so law lives in decided cases and stare decisis binds. Yours descends from Bologna: fragmented continental jurisdictions adopted the recovered Roman corpus as a shared university-taught base (ius commune), later rewritten by nation-states as systematic codes — so law lives in enacted articles you decompose into elements. England never took up the Roman base because its own system matured before the Bologna revival arrived, with a profession trained in the Inns of Court, not on the Digest. The working comparison is chapter 15 (see ch. 15).

3. What problem does Fedlex solve that Hammurabi's stele already solved — and what problem does it solve that the stele could not?

Answer: The shared problem is arbitrariness: both publish the rules in fixed form, in advance, where anyone the courts will judge can read them, binding the ruler's own apparatus to its stated law. What the stele could not solve is change: carved text has no mechanism for amendment or consolidation, so a stone code drifts from the law actually applied. Fedlex adds versioning — every amendment merged into an authoritative consolidated current text (electronically authentic since 2016), with the history retained so you can resolve which version was in force at the relevant time (see ch. 10). Publication guards against arbitrary power; versioned publication guards against stale or silently shifting text.

Exercises

Self-test

Fieldwork

• Reading · ~20 minFind a translation of the Code of Hammurabi online and locate the builder provisions (the collapsing-house series, §§ 229 ff.). Rewrite two of them as Tatbestand → Rechtsfolge checklists, connectives and all. Confirm for yourself that law has been if-then since 1754 BC — then notice what is missing that Chapter 3 supplies: nothing on the stele says who must prove the elements.
• Fedlex · ~15 minOpen the CC on Fedlex and find, in the act’s metadata, its adoption date (10 December 1907) and entry into force (1 January 1912). Then reread art. 1 CC one last time, knowing Huber wrote it so that you — not a professor — could read it. The book you just finished is only cashing that check.

What this chapter gives you. The book's master index and operating manual. Appendix A maps every article cited in the book to a one-line statement of what it does and the chapters that use it. Appendix B collects the deadlines and prescription periods into one expiry-clock table, because the clock kills more claims than the merits do. Appendix C is the operating manual: where the authoritative sources live, how to search them, and when to pay a professional. Nothing here is new doctrine, it is the index; the other chapters do the work.

Appendix A: Article reference table

Three codes dominate this book: the Civil Code (Code civil, CC, RS 210), the Code of Obligations (Code des obligations, CO, RS 220), and the two procedure statutes, the Debt Enforcement and Bankruptcy Act (Loi sur la poursuite pour dettes et la faillite, LP, RS 281.1) and the Civil Procedure Code (Code de procédure civile, CPC, RS 272). One-liners are compressed; the article's full element checklist is always richer, so read the source before relying on a row (see ch. 1).

Civil Code (CC)

Code of Obligations, general part

Code of Obligations, special part (named contracts)

LP (debt enforcement)

CPC (civil procedure)

Appendix B: Deadlines and expiry clocks

Two clock families. Procedural deadlines (délais) are hard cutoffs: miss them and the door closes whatever the merits. Prescription is an expiry clock on the claim itself: the right survives, but becomes unenforceable if the debtor pleads the exception, which courts never raise on their own (see ch. 6). That is also why you never volunteer a payment or written acknowledgment on a stale debt: it resets the clock (art. 135 CO).

Clock mechanics, in three rules: interruption resets prescription to zero (art. 135 CO), suspension pauses it (art. 134 CO), the debtor can waive the exception for at most 10 years at a time (art. 141 al. 1 CO), and the periods themselves are off-limits to contract drafting (art. 129 CO, a mandatory tag, see ch. 4).

Appendix C: How to look things up

Statutes: Fedlex

Everything federal lives at fedlex.admin.ch, the official compilation. The key is the RS number (Recueil systématique), a stable identifier that works like a catalogue number:

Practical habits. Search the RS number, not the name, and always open the "current consolidated version": Fedlex also serves historical snapshots, and citing a stale one is quoting an outdated edition of the rules. The French, German and Italian texts are equally authentic (art. 14 LPubl, RS 170.512), so when a French sentence is ambiguous, hold it against the German: a free second telling of the same rule, and where the two versions diverge is exactly where the ambiguity lives (see ch. 11). (If you come from software: two equally authentic language versions are two independent implementations of one spec; diffing them exposes the ambiguity.)

Case law: bger.ch

Federal Supreme Court decisions are free at bger.ch under "Jurisprudence". Two tiers, as in ch. 11: published leading cases are cited ATF (Arrêts du Tribunal fédéral) as volume, part, page, for example ATF 142 III 638, where the part encodes the chamber (III is civil law); everything else goes by docket number, for example 4A_123/2024 (4A is the first civil law division). Search tip: quote exact phrases and query articles the way courts write them ("art. 257e CO"). The ATF collection is the curated set of settled answers that get reused; docket decisions are persuasive but less canonical.

Counterparty checks: Zefix and the poursuites registry

Before signing with a Swiss company, spend five minutes on zefix.ch, the central commercial register index: legal form, seat, signatories, liquidation status, all free. This is where Ekeko will itself appear once registered in Vaud. For solvency, request an extract from the registre des poursuites at the debtor's local office des poursuites; anyone showing a legitimate interest, such as a prospective creditor, can obtain it (art. 8a LP). A client with a page of open poursuites belongs in your pricing, your prepayment terms, or your decision to walk (see ch. 12).

Cantonal layer: Vaud

Cantonal law sits below federal statute in the norm stack (see ch. 10) and lives in the Base législative vaudoise on vd.ch, with lexfind.ch for cross-cantonal search. Useful Vaud addresses: the justice de paix for general conciliation, the commission de conciliation en matière de baux and the Tribunal des baux for leases, the Tribunal de prud'hommes for employment. Tenant-side, ASLOCA Vaud gives members advice and representation for a modest annual fee, usually the highest-expected-value first move in a lease dispute.

When to pay a lawyer

Run the settlement expected-value frame from ch. 12. Self-serve or use a subsidized service (ASLOCA, a union, legal-protection insurance) when the stakes are a few thousand francs, the track is conciliation or simplified procedure (art. 243 CPC) where courts establish facts more actively, and the question is one this book's moves can handle. Pay a lawyer when any of these hold: the amount at stake is a meaningful multiple of the fee, the opponent has counsel and the matter is headed past conciliation, the fight is over legal characterization rather than facts, or an Appendix B deadline is about to fire and you are not sure which one. For Ekeko specifically, one paid review of your standard consulting template is insurance amortized over every future engagement, a better spend than any single dispute. A fixed-fee first consultation through the Ordre des avocats vaudois prices the rest of the decision.

The move in one line

Keep the index outside your head: articles are lookup keys, deadlines are expiry clocks you calendar the day they start, and Fedlex, bger.ch and Zefix are the authoritative sources you consult instead of guessing.

Drills

1. A commandement de payer for CHF 4,700 lands in your mailbox on March 3. Name the two deadlines that start ticking and what happens if each side misses theirs.

Answer: For you, opposition within 10 days of service (art. 74 LP), or the pursuit continues as if the debt were admitted. For the creditor, the right to continue lapses 1 year after service (art. 88 al. 2 LP); and after 3 months of creditor inaction you can ask that the entry be hidden from third-party extracts (art. 8a al. 3 let. d LP).

2. Your lease ended June 30 and the landlord has neither sued nor started a poursuite. When can you demand the deposit directly from the bank, and on what article?

Answer: One year after the end of the lease, so from July 1 of the following year, the tenant can require the bank to release the deposit if the landlord has asserted no claim through the courts or through debt enforcement (art. 257e al. 3 CO). Before that, release needs both parties' agreement, a final judgment, or an unopposed commandement de payer.

3. Ekeko delivers a custom ML pipeline in January 2027 and the client accepts it. In August 2029 they claim a defect. Which appendix rows decide the outcome?

Answer: A delivered build is a movable work under the contrat d'entreprise, so defect claims prescribe 2 years from acceptance (art. 371 CO): August 2029 is out of time, and acceptance already discharged the contractor for apparent defects (art. 370 CO). The client's only angle is a concealed or hidden defect notified immediately on discovery, and proving that sits on the client (art. 8 CC, see ch. 3).

Exercises

Self-test

Fieldwork

• Setup · ~45 minBuild the operating kit this appendix describes: bookmark the Fedlex pages for RS 210, 220, 272 and 281.1, the bger.ch search, and Zefix. Then create a standing deadlines calendar seeded with every Appendix B clock that currently applies to your life — deposit release dates, invoice prescription, any running opposition or objection windows. The clock kills more claims than the merits do; the calendar is the counter-weapon.
• Zefix · ~15 minRun one real counterparty — a client, your régie, a supplier — through zefix.ch: legal form, seat, who signs. Write one line on what you would now check before signing anything above pocket-money size with them. Make the five-minute check a reflex, not a project.