The Honest-Applicable-Reversible Values Test
Three tests that separate values which drive decisions from values that just describe your self-image.
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 4
- Confidence
- 90%
Most company values fail because they're identity statements — nice descriptions of who you'd like to be that nobody can ever act on. Larson runs every candidate value through three tests: is it honest (do your actual decisions match it), is it applicable (can someone use it to settle a real trade-off), and is it reversible (could a good company credibly hold the opposite). Values failing any of the three should be cut, not workshopped.
Origin
Larson's own three-part test, developed across Stripe, Uber, Calm and Carta. He grounds it in observed opposites — Stripe's 'optimize globally' versus Uber's implicit 'do what's good for your team' — and in a company he worked at whose execs visited the Facebook campus, saw the values on the wall, and wrote them on their own walls. Lenny adds a complementary test from Airbnb: a real value makes clear who does not belong.
Core principles
- 01You cannot steal values. Facebook's values worked for Facebook because they were true of Facebook.
- 02Honest: the value must match the decisions you actually make, not the ones you'd like credit for.
- 03Applicable: it must be usable on a real decision, like whether to introduce a new programming language.
- 04Reversible: if no credible company could hold the opposite, it carries no information.
- 05Identity values aren't wrong, they're just useless — and they crowd out the useful ones.
- 06Values describe who you are, not who you aspire to be.
- 07Cap the list at roughly four. Seventeen values that nobody can recall is the same as none.
How to run it
- 1
Test for honesty
For each candidate value, check it against decisions you have actually made. 'Users first' is a great Amazon value; a lot of companies aren't users first and that's fine — what's not fine is claiming it while your decisions visibly say otherwise.
Pro tip Pull three recent contested decisions and ask which value they demonstrate. That's your honest list, whatever the wall says.
Watch out A dishonest value is worse than no value: it teaches the org that the values document is theatre.
- 2
Test for applicability
Ask whether the value settles a real trade-off. Stripe's 'optimize globally' resolves 'I want to introduce a new language that's better for my team but worse for the org' — you don't do it. Uber's implicit opposite ('do what's good for your team, ignore everyone else') resolves it the other way. Both are applicable; both companies worked.
Pro tip Write the actual sentence someone would say in a meeting to invoke the value. If you can't write it, it isn't applicable.
- 3
Test for reversibility
Ask whether a real company could credibly hold the opposite. 'We're thrifty' reverses to 'we spend as much as we need to get the best value' — both are run by good companies. 'We build good software' and 'we solve customer problems that matter' don't reverse and are therefore worthless. Netflix's 'we're a team, not a family' is a partial reversal of a common identity value.
Pro tip Lenny's addition: a real value makes it clear who doesn't fit. It functions as a hiring filter — there are people you would explicitly not hire because of it.
Watch out 'Integrity' is the archetypal failure. Nobody is the anti-integrity company, so it says nothing about you.
- 4
Cut, don't polish
Any value that fails a test gets deleted, not rewritten. Larson: if you already have seventeen, you need to get to about four, so people can remember them. Don't let the exec team spend time debating identity values because they feel good to debate.
Pro tip Do the deletion pass a year in, once you can see which values you've actually failed to live.
In the wild
Larson worked at a company where the executive team visited the Facebook campus, saw the values written on the wall, and went home and wrote the same values on their own walls.
→ The values did nothing for the company and arguably undermined employees' confidence in the executive team's critical thinking. Facebook's values were meaningful because they were honestly Facebook's.
Lenny recounts Airbnb starting with six values, one of which was 'simplify'. A year or two in, everyone realised they wanted to simplify but weren't actually good at it — it described an aspiration, not the company.
→ Airbnb cut two values including 'simplify', keeping four that honestly described who they were.
Stripe held 'optimize globally' — don't take the local win for your team if it costs the org. Uber's implicit value was the reverse: do what's good for your team and ignore everyone else, because coordinating slows us down.
→ Both values passed the applicability and reversibility tests, and both companies moved fast on them. The lesson is that the right value is company-specific, and reversibility is the proof it carries information.
Common mistakes
Cargo-culting another company's values
Copying values from a company you admire imports statements that were honest for them and are not honest for you. The result is a set of claims your own decisions visibly contradict.
Debating identity values because it feels good
Exec teams enjoy discussing 'we're kind to other people' or 'we're a family'. These feel great and can't be applied to any decision. They're not wrong, they're inert, and the time spent on them is time not spent on values that would settle real trade-offs.
Keeping too many values
If nobody can remember the list, the list can't be used. Larson's bar: get down to about four. Every non-applicable value you keep dilutes the ones that work.
Is it for you?
Best for
Founders, CTOs and org leaders drafting or auditing company/org values who want values that actually settle arguments and filter hires.
Not ideal for
Regulated or safety-critical settings where certain non-reversible commitments (e.g. safety, integrity) must be stated for legal or cultural reasons regardless of information content.
From the transcript
“good values have to be honest”
“these are the three that I found really useful for any value like is it honest is it a applicable and can you reverse it”
“there's certain values that I think of just like identity values and and they're not like they're not wrong to have identity values they just…”
“you can't do is just like steal values”
From the episode
The engineering mindset
Will Larson (Carta, Stripe, Uber, Calm, Digg)