The principle, and the source
Culture is not one solid block. Defaults change in weeks, incentives in quarters, underlying beliefs only as a byproduct of the first two staying consistent for a long time. Speeches move the third layer roughly zero — they don't even touch the first.
"Culture is a pattern of shared basic assumptions learned by a group as it solved its problems… taught to new members as the correct way to perceive, think, and feel in relation to those problems."
— Edgar H. Schein, Organizational Culture and Leadership, 4th ed. (2010), p.18
The layers and their levers
Three layers · which hand moves which
App layer · defaultsmeeting format, review SLA, on-call rules, how people talk in channels2–8 weeks
lever: change the process default directly
Kernel · incentivespromotion bar, perf weighting, who gets praised, who gets kept2–4 quarters
lever: change calibration and the promo bar
Firmware · assumptions"is it safe to say this?" "does failure go on my record?"1–3 years
no direct lever: rewritten only by consistency in the layers above
Work left to right. Skipping left and shouting at the right is what a values poster is.
Scenario
Context: your team won't report bad news; slips always surface in the final week. Your boss says: "Go build a culture of candor."
✗ The usual move
At the all-hands: "We need to be open and candid — bring me problems any time, I won't blame anyone." Then "Transparency" goes into the team wiki. Three months later, the next slip still surfaces in the final week.
✓ Better: work layer by layer
Layer 1 (within two weeks): a fixed 10 minutes at the end of the weekly — one line each: "red / yellow / green + the one thing I'm most worried about," format locked. I go first, and I have to name my own worry.
Layer 2 (one to two quarters): set one line in calibration and write it into the perf template — "surfaced a risk 4+ weeks early with an option" counts as impact; "surfaced it in the final week" costs a notch even if the thing shipped.
Layer 3: leave it alone. It starts moving the first time someone flags a risk early and I reallocate resources instead of assigning blame.
Checklist
- My most recent "culture change" — which layer did it act on? If it was only talk, it didn't reach layer one.
- What have I actually changed in layer two? Can I name the specific promo-bar or perf-weighting change?
- Does my time scale match the layer? Demanding belief change in three months guarantees disappointment — and a wrong diagnosis ("this team just doesn't get it").
Common mistakes
- Mistaking artifacts for culture: you changed the meeting format and announce "our culture has changed." That's a trace culture left behind, not culture.
- Giving a layer-one order in layer-three language: "have more ownership" is unexecutable; "for 48 hours after launch you're on call, with authority to roll back without asking" is executable.
This week's rep + a question
Do: take one culture problem you want fixed and translate it into two things — one default you can change this week, one incentive line you can change this quarter.
Reflect: the strongest underlying assumption on your team right now — which specific event under a previous leader taught it?
The principle, and the source
Rituals — repeated actions with a fixed time, fixed format, and a fixed host — execute the culture once a week. Symbols — one-off, high-contrast events — tell people where the boundary is. Both are cheaper than speeches, and both work.
"…a meeting is nothing less than the medium through which managerial work is performed."
— Andy Grove, High Output Management (1983), Ch.4 "Meetings—The Medium of Managerial Work"
Scenario
Context: you want "write it down before you build it" to become the team default, instead of something you chase every time.
✗ The usual move
A long email on why design docs matter, template linked, closing with "please make this a habit." New RFCs two weeks later: zero.
✓ Ritual (executes weekly)
"Thursdays at 15:00, Design Half-Hour: if you have an RFC you get 5 minutes; if you don't, don't come. I host it; when I travel Alex hosts — this meeting is never cancelled. Afterwards I post in the channel: this week's RFC links, plus whose decision got changed by the discussion."
✓ Symbol (one-off, very loud)
An engineer's RFC once made the team drop the approach you were championing. You tell that story at the all-hands: "Here's what I planned to do; the data in section three of his doc convinced me, and we switched." One of these beats ten emails.
Checklist: will this ritual survive?
- Is it ≤30 minutes? Does it have a fixed host and a backup? (A ritual with no backup doesn't survive its second quarter.)
- Does it contain a moment where someone is named specifically? Rituals without naming decay into status broadcasts.
- Does it still happen when I'm not there? Yes = it's in the OS. No = it's still just my personal habit.
- Can I name one symbolic event I created last quarter? If not, the team reads culture only off my calendar.
Common mistakes + a note on gender
- Rituals that cost too much: a two-hour "culture workshop" once a month loses to ten minutes every week. Frequency beats ceremony.
- Only positive symbols: if crossing the line is never handled publicly, the line doesn't exist. Negative symbols must address behavior, not the person, and the person must know in advance — otherwise you broadcast fear, not a standard.
- The work of running rituals defaults to the same person: booking the room, sending reminders, writing notes. When this "office housework" lands persistently on the women or the most junior person on the team, the ritual itself broadcasts a culture you don't want. Fix: rotate it, and count it as real contribution at perf time.
This week's rep + a question
Do: pick one behavior you want and design it into a ≤30-minute weekly ritual; name the host and the backup on the spot.
Reflect: which of your team's existing rituals is already dead but still on the calendar? What is it broadcasting now?
The principle, and the source
Hiring decides what gets installed; promotion decides what the system optimizes for. The team doesn't know your candidates — it knows the person you promoted. Promotion is the strongest cultural signal of the three, and the only one no amount of messaging can walk back.
"Employers sought candidates who were not only competent but also culturally similar to themselves in terms of leisure pursuits, experiences, and self-presentation styles."
— Lauren A. Rivera, "Hiring as Cultural Matching," American Sociological Review 77(6), 2012
Scenario A: turning "fit" into "add"
✗ Interview feedback
"Great conversation — felt like one of us." Zero information. All it measured was similarity.
✓ Interview feedback
"My round tested candor under conflict. I asked how he handled the last time he and his TL disagreed: he privately asked where their data came from, found a constraint he'd missed, then volunteered in the team meeting that he'd been wrong. He adds something we're missing: nobody on this team volunteers that they were wrong."
The formula: which behavior I tested + the evidence + what he adds that we lack.
Scenario B: the question after promotions land
Context: results are out, and a report asks you in a 1:1: "Why did K get promoted? He drove two people out last quarter."
✗ Don't say
"His business impact really was large; we're addressing the people issues with him privately." Translated honestly: results can buy line-crossing. From here on, every behavioral standard you state is discounted.
✓ Try
"You're right, and this is one I failed to hold. I judged the project couldn't survive without him — that judgment was wrong, and it cost two people."
"What's already changed: starting this cycle, behavioral issues raised in skip-level feedback get their own column in calibration and can't be offset by impact. K's next promotion has to clear that column. This applies to everyone, including me."
Checklist
- In my last three interview debriefs, how many said "we clicked" rather than "here's the behavior I tested and the evidence"?
- The last person I promoted — can the team say which behavior he stands for? If not, that promotion carried no information.
- Is there someone on my team with strong results who crosses the line, still here? If so, in my system impact outranks behavior.
- Who tests culture in the loop? If every interviewer does, nobody does.
Common mistakes + a note on gender
- Leaving the culture interview to "gut feel": without fixed dimensions, what interviewers measure is similarity.
- Talking culture at hiring but not at promotion: the team can't see the first and sees all of the second.
- Handling line-crossers only in private: if the team doesn't know, it didn't happen — a cultural signal exists only when it's seen.
A note on gender
Empirically, "culture fit" is close to equivalent to "similar to the interviewer" (Rivera 2012). For women and people from non-dominant backgrounds the damage runs twice: the reason for rejection can never be stated clearly ("just didn't feel like a fit" — you can neither appeal it nor improve on it), and once inside, they're more likely to be expected to fill the culture gap — running events, onboarding newcomers, sitting on the DEI committee — work with no slot in the perf template. Two operational fixes: rewrite the culture interview as 2–3 fixed behavioral dimensions, one per interviewer, with evidence required in the writeup; and if you genuinely need someone to fill a culture gap, run it as a project — in their goals, with time allocated, counted as impact.
This week's rep + a question
Do: convert your team's culture interview into 2 fixed behavioral dimensions and require specific evidence in every writeup.
Reflect: the last time you called a candidate "not quite a fit" — if you had to write down the behavioral evidence, could you?
The principle, and the source
Every "this one's special, let's skip it" is a loan. Borrowing isn't the mistake; not booking it and not repaying it is. The interest takes a very concrete form: next time, someone skips it without asking.
"Shipping first time code is like going into debt… The danger occurs when the debt is not repaid. Every minute spent on not-quite-right code counts as interest on that debt."
— Ward Cunningham, "The WyCash Portfolio Management System," OOPSLA '92 Experience Report
Scenario
Context: two weeks before peak season, a change needs to skip design review and go straight out. The owner comes to you for approval.
✗ The usual move
"Special case — ship it, be careful next time." No payer, no due date. Three months later the review process actually runs at 40%, and nobody remembers which exception started the slide.
✓ Better: say all of it, then post it in the channel
"Approved to skip, and I'm booking it: I'm the borrower, the due date is Aug 15, and repayment is a retroactive RFC plus a Design Half-Hour talk on what made it safe to skip this time, and under what conditions it wouldn't be."
"This waiver covers this one change only; everything else goes through review as usual."
Checklist: interest is invisible, these four are countable
- Exception count: how many skips did I approve last quarter? No number = no ledger.
- Request rate: do people still come and ask, or do they just skip? Nobody asking = the rule is dead and only the corpse is still on the wiki.
- Explanation half-life: ask someone three months in why the rule exists — can they answer? A rule nobody can justify is the first one sacrificed under pressure.
- Is there a rule I break myself? (A leader's exceptions carry ten times the interest rate.)
Common mistakes
- Treating debt as sin: a team with no culture debt usually isn't a healthy culture — it's a team not doing anything hard. Borrow, but book it.
- Silent repeal: if you don't want a rule, repeal it out loud. An unenforced rule left on the books devalues every other rule — the highest-interest form of culture debt there is.
- Repaying only after an incident: that's not repayment, that's collection — and what the team learns is "it only counts when something breaks."
This week's rep + a question
Do: list every exception you approved this quarter and attach a payer and a due date to each; for the ones you can't repay, formally announce the rule change.
Reflect: which rule on your team is everyone breaking while nobody dares say it's already dead?
Going Deeper
Inside a sub-team of twelve, can you build a local culture different from the company's? Where's the boundary?
You can, but only in layer one and part of layer two. Defaults — review practice, meeting format, on-call rules — are fully yours; incentives are only partly yours, because promotions clear calibration and the company bar wins in the end. The real danger is letting the team believe layer three differs too ("we don't play that game here"), then watching it shatter in some promotion cycle and taking your credibility to zero in one shot. Safer: commit only to the layers you can deliver, and state the boundary — "inside the team, this is how we work; at company level the bar is that one, and I'll translate for you."
Is culture's malleability overrated? What actually happens when a new leader arrives to "reshape the culture"?
New leaders typically hammer layer one in the first 90 days — meetings, process, vocabulary — because it's fast and visible. But the layer-three assumption ("this place reorgs every 18 months, so don't invest long") not only survives, it gets reinforced by the churn itself. The only thing that rewrites layer three is one concrete event that contradicts the history: publicly protecting someone who followed the new rules and took a short-term loss for it. Without that, every layer-one change reads as one more round of noise at the next reorg.
Does "hiring is culture" break down on a team of only senior people whose values are long since set?
It doesn't break down, but the lever moves from shaping to selecting and making explicit. Senior people won't change their values because of your rituals, but they will follow rules that are clearly stated and consistently enforced. So the weight goes on layer-one explicitness (write the defaults down, and write down why) and layer-two consistency, not layer-three assimilation. Conversely, the culture they bring becomes an input to your layer three — that's what culture add actually means, and where the risk sits: you have to tell which imports are additions and which are bad assumptions.
Where does the analogy between culture debt and technical debt fail?
Three places. Who pays the interest: technical debt is paid by machines and whoever comes next; culture debt is paid by specific people, usually not the borrower — typically the most rule-abiding person leaves first. Whether it can be rewritten: technical debt can be cleared in one rewrite; culture debt only amortizes through time and consistency, with a lag (you change the bar, trust takes quarters to return). Measurability: technical debt has a visible balance; culture debt only has proxies — exception counts, regretted attrition, the same name recurring in skip-level feedback. Which means you will systematically underestimate it.