पाठशाला Pathshala · दल Dal, The team · Lesson 29 · Scale
The culture document that survives a hundred people
At a hundred people most of the company has never watched a founder make a hard call. Values written as trade-offs with examples and anti-examples let a new manager in Pune make the call the founder would.
Pathshala, The Founder Library · 11 October 2026 · 7 min read
At twenty people culture travels by watching: everyone has seen a founder handle an angry customer, a missed deadline and a bad hire. At a hundred most people have not, and the managers making daily calls in a new office have never been in the room. A culture document is how a founder’s judgement reaches the decisions the founder will never see.
Most culture documents fail at exactly this. This lesson covers why, how to write values as trade-offs with examples and anti-examples, how to test the document on the people it is for, how to make it live in hiring, reviews and exits, and how to revise it without rewriting it. The [tolerance lesson](/library/culture-is-what-you-tolerate) covers the behaviours a company enforces; this one covers writing them down so they scale.
Why most values statements fail
Integrity, customer focus, ownership, excellence, teamwork. Every company has some version of the list and none of them decides anything, because no one is against any of the words. Ben Horowitz, speaking at the Computer History Museum, put the problem plainly: culture is about action, not words; it is not the mission statement. His example of failure is a slogan like “don’t be evil”, which tells nobody what to do in a complicated situation. His example of success is the samurai code, which lasted centuries because it set specific expectations and enforced them.
A value earns its place only if it changes a decision. The test is simple: would a reasonable company choose the opposite? “We value quality” fails, because no company values poor quality. “We ship on Thursday even if a feature is not finished, and cut scope rather than move the date” passes, because plenty of good companies would move the date. Molly Graham’s warning explains the urgency: the first hundred people hired define the next two hundred, and she advises writing the values down early, before growth makes it hard.
Five to seven values, each a trade-off
Write each value as a choice between two good things. Amazon’s leadership principles are the most copied example because several of them cost something. Frugality says there are no extra points for growing headcount, budget size or fixed expense. Have backbone; disagree and commit says leaders are obliged to challenge decisions they disagree with and, once a decision is made, to commit wholly. Each tells a manager what to do when two reasonable instincts conflict.
Keep the list to five to seven. Netflix’s culture memo names eight, each with a handful of behaviours under it; Amazon has sixteen principles, which works for a company that hires against them every day and is too many for one of a hundred. Fewer values used constantly beat more values remembered by nobody. Then rank them. GitLab’s values page states that results for customers sit at the top of its hierarchy, because they drive the performance that enables everything else. A ranking tells a manager what to do when two values point in different directions, which is precisely when they need the document.
Examples and anti-examples
Under each value write five things. The name, short and specific to you. One sentence saying what the company does, ideally with the words “even when”. The trade-off, as “we choose this over that”. Two real examples from the company’s own history, with enough detail that a new joiner can picture them. And one anti-example: behaviour that looks like the value but is not, because that is where most misreadings happen.
A worked value for a business software company. Name: Say no to revenue that breaks the product. Sentence: we turn down deals that need custom work outside the roadmap, even at quarter end. Trade-off: we choose a product every customer can use over a contract one customer wants. Example one: in 2024 the company declined a ₹1.2 crore contract from a large retailer that needed a separate reporting module, and built the general version six months later for every customer. Example two: a sales manager walked away from a renewal that required a dedicated support engineer, and the founders said so in the next all-hands. Anti-example: refusing a reasonable configuration a customer is entitled to under their plan and calling it product discipline. The value is about custom work, not about saying no to customers.
Now picture the reader the document is for. A manager in Pune, hired four months ago to run a twelve-person sales team, has a customer asking on the last day of the quarter for a feature that is not on the roadmap in exchange for signing. Without the document they guess, or call a founder who is on a flight. With it they read the value, the 2024 example and the anti-example, and make the call the founder would. That is the whole job of a culture document.
A value that no reasonable company would reject decides nothing. Write each one as a trade-off, with an example from your own history and an anti-example beside it.
The Pune test
Test the draft on the people it is for, not the people who wrote it. Pick three managers hired in the last six months who have spent little time with the founders, ideally in a different city. Give them five real decisions from the past year, described as they looked at the time, with the outcomes removed: a pricing exception, a hire, a customer escalation, a missed deadline, a conflict between two teams. Ask each to decide using only the document, and to say which value they used. Compare their answers with what the founders actually did.
If they reach the founders’ answer in four of five cases, the document works. If they do not, find out why: a value that is too vague, a missing value the founders were using without writing it down, or a ranking that is wrong. Sometimes the managers’ answer is better than the founders’, which is useful too: it means the document describes the company the founders want, and their own past decisions fell short of it. Rewrite and test again with different decisions.
Making it live: hiring, reviews, promotions and exits
A document that sits in a wiki changes nothing. It has to appear in the four places where the company decides who it is. In hiring: one interview per loop assesses the values, with a question per value asking what the candidate did in a situation where it applied. In reviews: each person is assessed on how they worked as well as what they delivered, against the values by name. In promotions: no one is promoted into management who has a pattern of anti-examples, however strong their numbers; the [managers of managers lesson](/library/managers-of-managers-second-layer) explains why the second layer matters most. And in exits: Netflix’s keeper test, asking whether a manager would fight to keep someone if they wanted to leave, is a values question as much as a performance one.
The founders are held to it first and in public. Every all-hands, tell one story of the document being used, ideally one where it cost the company something. When a founder breaks a value, say so. Horowitz’s point is that the reason behind a rule matters more than the rule itself in conveying culture; stories carry reasons in a way lists cannot.
Revising without rewriting
The document will need to change; a company at three hundred people is not the company at a hundred. Netflix’s memo says so directly: as the business grows and evolves, its culture and the document will too. GitLab calls its values a living document, refined based on lessons learned, and lets anyone propose a change through its handbook, with the chief people officer as the owner. Copy the mechanism, not the scale: anyone can propose a change in writing, the head of people owns the document, and the founders decide once a year.
Change examples often and values rarely. New examples keep the document current and show it is being used; changing a value says the company has changed its mind about who it is, and should be rare enough to be an event, explained at an all-hands with the reason. A value that has not produced a single new example in a year is either not being used or not needed. The [writing culture lesson](/library/writing-culture-decisions-in-documents) covers keeping documents like this alive.
The annual culture document review
Once a year, in the same month: the head of people collects every proposed change and every new example gathered from all-hands stories, reviews and exit interviews. Run the Pune test again with five new decisions and three managers hired in the last year. Replace stale examples with recent ones, check the ranking against the last year’s hardest calls, and retire or rewrite any value that produced no examples. Publish the new version with a one-page note on what changed and why. Every quarter in between, read the exit interviews and the review comments for the values most often cited and most often missed, and bring one story of each to the next all-hands.
The worked value is illustrative; your values should come from your own hardest decisions, not from this page or anyone else’s.