Chapter 8
The values on the wall vs. the values in the room
On February 19, 2017, a former Uber engineer named Susan Fowler published a blog post titled "Reflecting on One Very, Very Strange Year at Uber." In flat, meticulous prose, she described propositions from a manager on her first day, an HR department that protected the harasser because he was "a high performer," and an engineering organization where the share of women fell, during her tenure, from 25 percent to 3 percent.14 The post detonated. Within months came a cascade — investigations, executive departures, revelations of software built to deceive regulators — and by June, CEO Travis Kalanick had resigned under shareholder pressure.2 The remarkable thing, in hindsight, is that Uber had a famous, explicit culture: fourteen values, proudly recited, including "always be hustlin'" and "toe-stepping" — the latter meant, officially, to encourage challenging ideas regardless of seniority. As Kalanick's successor Dara Khosrowshahi later admitted, "too often it was used as an excuse for being an asshole."4 Khosrowshahi's rebuild began with the values themselves — rewritten with input and voting from thousands of employees, topped by a norm with no loopholes: "We do the right thing. Period."3 Six years later, Uber posted its first annual profit.2
Netflix, meanwhile, publishes its culture the way other companies publish products. The Netflix Culture Memo — revised most recently in June 2024 under the title "The Best Work of Our Lives" — states, in public, things most companies would never say out loud: Netflix models itself on "a professional sports team, not a family"; it aims to employ only high performers; managers apply a "keeper test" to every employee — *would I fight to keep this person? Knowing what I know now, would I hire them again?* — and if the answer is no, "it's fairer to everyone to part ways quickly."56 The memo demands "extraordinary candor," instructs leaders to "farm for dissent," and concedes openly that the environment is stressful and not for everyone.57 It is, on paper, a harsher culture than Uber's ever claimed to be. It has also powered two decades of reinvention without an implosion.
Here is the puzzle stated cleanly. Uber's espoused culture and Netflix's espoused culture are not that different — both celebrate aggressiveness, candor, high performance, discomfort with mediocrity. One produced a company-threatening catastrophe; the other produced durable excellence. So the difference cannot lie in the words. A skeptic draws the natural conclusion: culture talk is decoration, and what matters is strategy, incentives, and luck. A believer draws the opposite one: culture is precisely what differed — not the stated values but the operating ones, the things that actually got people promoted, protected, or fired. Which is it? And if culture is real, is it something management controls — or something that happens to them?
The central question: is organizational culture a manageable asset or an emergent accident — and what separates a culture that performs from one that implodes? The tools: Schein's three levels, the field's foundational map; where cultures come from — founders, selection, and socialization; the evidence on whether strong cultures actually pay; person–culture fit and its dark side; how to read a culture accurately, including the espoused–enacted gap that destroyed Uber's first act; and the honest evidence on changing culture. Uber and Netflix then return as the decade's clearest paired test.
Organizational culture is the pattern of shared basic assumptions a group has learned while solving its problems — assumptions that worked well enough to be taught to newcomers as the correct way to perceive, think, and feel. Edgar Schein, the field's founding theorist, mapped culture in three levels of decreasing visibility.8
• Artifacts — what you can see and hear: offices, dress, rituals, slogans, org charts, the fourteen values on the wall. Visible, but easy to misread.
• Espoused values — what the organization says it believes: mission statements, culture memos, leadership speeches.
• Basic underlying assumptions — what members actually take for granted: who really gets ahead here, what actually gets you fired, what happens to people who bring bad news. Invisible, rarely articulated, and the true engine of behavior.
Schein's decisive insight is that the levels can contradict each other, and when they do, the assumptions win. Uber's espoused value said toe-stepping meant challenging ideas; the operating assumption said high performers are untouchable — and every employee learned which one was real from what HR actually did.14 The espoused–enacted gap is therefore the single most diagnostic measurement in this chapter: a culture is not what a company says; it is what its members have learned to expect. The framework's limit is that it describes culture better than it predicts performance — for that, the field needed the evidence in Section 2.3.
Cultures are not designed on whiteboards; they accrete. Three engines do most of the work. Founder imprinting: early leaders' assumptions get embedded through what Schein calls primary mechanisms — what leaders pay attention to and measure, how they react to crises, whom they reward, promote, and fire.8 Note what is absent from that list: speeches. Attraction–selection–attrition: organizations grow homogeneous because similar people apply, similar people get hired, and dissimilar people leave — culture partly selects its members rather than shaping them. Socialization: newcomers learn the assumptions through structured tactics (bootcamps, mentors, war stories) and, more powerfully, through watching what actually happens — Van Maanen and Schein's classic analysis shows how socialization design shapes whether newcomers become custodians of the culture or innovators within it.12 The managerial punchline: by the time a culture is visible enough to discuss, it is already self-reinforcing through hiring, promotion, and exit. That is why Khosrowshahi's rebuild ran through the engines — new norms co-written and voted on by employees (socialization by authorship), executive departures (attrition applied upward), and public accountability norms — rather than through posters.3 The caution: these engines also explain why cultures drift toward comfortable homogeneity, the seed of Section 2.4's dark side.
A strong culture is one where values are widely shared and intensely held, producing consistent behavior without close supervision. Kotter and Heskett's influential study of over 200 firms found that companies whose cultures emphasized all key stakeholders and adaptability dramatically outperformed over a decade, while strong-but-rigid cultures could entrench decline — strength magnified whatever the culture contained.10 Jesper Sørensen's more rigorous follow-up sharpened the finding into this chapter's most useful contingency: strong cultures deliver more reliable, less variable performance — a genuine advantage in stable environments — but that same consistency becomes a liability in volatile environments, where the strong-culture firms' performance advantage evaporates because shared assumptions filter out exactly the discrepant signals that turbulent markets send.11
The evidence thus refuses both the skeptic and the cheerleader. Culture demonstrably matters — it is a coordination and control technology, substituting shared assumptions for rules and supervision (which is why strong-culture firms can decentralize). But strength is not goodness: a strong culture is an amplifier, and what it amplifies can be quality obsession, ethical corner-cutting, or obsolete beliefs held with perfect unanimity. Netflix's memo shows a company that has internalized Sørensen's contingency almost verbatim: its principles hard-wire adaptability itself — "uncomfortably exciting," "great and always better" — into the shared values, an attempt to build a strong culture whose content is change.611 Caveats for honest reading: culture–performance studies are largely correlational (successful firms can afford nice cultures), culture is measured imperfectly through surveys, and reverse causation is never fully excluded.
Person–organization fit is the match between an individual's values and the organization's. O'Reilly, Chatman, and Caldwell's Organizational Culture Profile research established the core findings: employees whose values match their organization's are more satisfied, more committed, and substantially less likely to leave — effects visible years after hiring.9 Fit is real and consequential, which is why Netflix's radical move is publishing its profile in advance: the memo functions as a self-selection instrument, warning off candidates who would be miserable and attracting those who thrive on candor and pressure — fit management before the first interview.56 The dark side follows directly from the same mechanism. "Culture fit" hiring degrades easily into similarity hiring — selecting for demographic and stylistic sameness rather than value alignment — which compounds the ASA homogenization of Section 2.2, shrinks the diversity of thought that Weeks 5 and 7 showed decision quality depends on, and can lawlessly encode bias. The discipline that preserves fit's benefits without its pathology: specify the values required, in behavioral terms, and hire for those while deliberately varying everything else — the principle sometimes summarized as hiring for culture add, not culture clone.
Because assumptions are invisible and espousals are cheap, reading a culture accurately — as an analyst, a new hire, or a due-diligence investor — requires triangulating behavioral evidence against the official story. The questions that work are Schein's embedding mechanisms run in reverse:8 What do leaders measure and attend to? (Uber measured growth; harassment complaints did not move the metrics that mattered.1) Who gets promoted, protected, and fired — and for what? (The protected "high performer" taught Fowler's cohort the real values in one HR meeting.1) What happens to bad news and dissent? (Netflix's "farming for dissent" is an enacted mechanism — informed captains must actively collect disagreement before deciding — not a slogan.7) What do the stories veterans tell newcomers celebrate? Every organization's war stories encode its assumptions more honestly than its website. The method's limits: single observations mislead (every culture has hypocrites and exceptions), subcultures diverge — engineering, sales, and operations often inhabit different assumption-worlds under one logo — and outsiders systematically over-read artifacts (ping-pong tables) and under-read consequences (who just got promoted). Weight consequences over artifacts, patterns over incidents, and the treatment of the powerless over the treatment of guests.
The honest answer: yes, slowly, expensively, and only through the same engines that built it. Documented change levers, in rough order of power: leadership succession and visible consequence changes (new answers to "who gets ahead here" — Uber's executive turnover and Kalanick's exit did more than any memo2); crisis (Schein: cultures learn when their assumptions visibly fail — Fowler's post and the investor revolt made Uber's assumptions unaffordable18); participative rewriting (norms employees author and ratify carry socialization force that handed-down values lack — Khosrowshahi's voted norms3); alignment of the hard systems — selection, promotion criteria, pay, and metrics — with the espoused values, without which every culture initiative is theater; and time, measured in promotion cycles, not quarters. What does not work is equally documented: values campaigns without consequence changes, which employees accurately decode as decoration and which widen the espoused–enacted gap the campaign was meant to close. The realistic frame for managers: you cannot install a culture, but you continuously teach one — every decision about attention, promotion, and tolerance is a lesson, whether or not you meant to give it.8
The pairing controls for the variable everyone blames first. Both companies espoused aggressive, candid, high-performance cultures; both are technology firms competing on speed. If harsh values cause cultural catastrophe, Netflix — which states the harshest values in corporate America, in public — should have imploded first. It hasn't, which redirects the analysis to where Schein said to look: not at the values espoused, but at the gap between espousal and enactment, and at what each system actually taught its members.
The facts: through 2017, Uber's fourteen values coexisted with the operating assumptions Fowler documented — harassment tolerated for high performers, HR aligned with the powerful, dissent punished, metrics über alles; the share of women in her organization collapsed from 25 to 3 percent during her single year.14 The unraveling ran February to June 2017: her post, the Holder investigation, cascading scandals, and Kalanick's forced resignation.2 Khosrowshahi's rebuild used the full Section 2.6 toolkit: leadership succession with consequence changes at the top; crisis converted into mandate; norms rewritten with employees and ratified by vote ("We build globally, we live locally… We do the right thing. Period."); a deliberately cooperative reversal of the adversarial regulator posture; and years of alignment work before the financial vindication — a 2019 IPO at roughly $82 billion and, in 2023, the company's first annual profit ($1.1 billion operating).23 As a theory test: Uber's first act is the espoused–enacted gap as accelerant — explicit values gave cover while enacted assumptions taught the opposite lesson daily; its second act is the strongest recent field evidence that deliberate culture change is possible, and of what it costs: the CEO, much of the executive team, six years, and a rebuild of every consequence system. The honest caveats: profitability had many parents (divestments, cost discipline, market maturation), attributing it to culture alone is the halo error of Week 6, and Uber's labor-model controversies did not end in 2017 — culture change is not sainthood.
The facts: Netflix's culture memo — iterated publicly since the famous 2009 culture deck, most recently in June 2024 — espouses talent density ("a professional sports team, not a family"), top-of-market pay, extraordinary candor, freedom paired with responsibility, and the keeper test, with underperformers exited quickly and severed generously.56 The 2024 revision is itself evidence of a learning system: it adds guardrails the earlier versions lacked — managers must evaluate whole records rather than single failed bets, stick with people through "short-term bumps," and "no matter how brilliant," nobody stays who fails to treat colleagues with decency; co-CEO Greg Peters publicly acknowledged that employees had over-heard "freedom" and under-heard "responsibility," prompting the rewrite.567 As a theory test: Netflix is the demonstration that a culture's content matters less than its integrity — the harsh values are enacted as written (people really are paid top of market; the keeper test really is applied; candor really does flow upward), so members can navigate it accurately; nobody at Netflix is ambushed by the real rules, which is precisely the trust Uber's gap destroyed. It also embodies Sørensen's contingency solved from the inside: a strong culture whose shared content is adaptability itself.11 The honest costs, which the memo itself concedes: chronic performance anxiety, a system whose fairness depends entirely on manager judgment quality in keeper-test decisions, and a model that presumes a labor market of confident high-performers with options — exportable to far fewer companies than its admirers assume.57
| Uber (2017 / rebuild) | Netflix | |
| Espoused values | Aggressive, candid, meritocratic (14 values) | Harsher on paper: sports team not family, keeper test, only high performers |
| Espoused–enacted gap | Wide: protection of harassing 'high performers' taught the real rules | Narrow: harsh rules applied as published, with pay and severance to match |
| What newcomers learned | Metrics excuse anything; HR serves power | Candor is safe and expected; performance is the deal, decency non-negotiable |
| Change mechanism observed | Crisis + succession + co-written, voted norms + realigned consequences | Continuous public iteration of the memo; 2024 guardrails after 'freedom' over-read |
| Cost of the model | CEO, executive team, six years, reputational scar tissue | Chronic pressure; fairness rides on manager judgment; limited exportability |
Three conclusions, disciplined by the evidence. First, the pairing relocates the causal variable: catastrophe tracked not the harshness of values but the integrity between espoused and enacted levels — Schein's hierarchy vindicated in the field.8 Second, both cases show culture operating through consequences, never announcements: Uber's employees and Netflix's employees alike learned their cultures from who was protected, paid, promoted, and exited — which is why Khosrowshahi changed the consequence systems and why Netflix's memo reads like a consequence manual.35 Third, resist two seductive overreadings: Netflix's model is not a template (it presumes talent-market conditions most firms lack, and its own revisions concede real damage from earlier versions), and Uber's turnaround is not proof that culture change is reliably available — it required a public catastrophe to unfreeze, which is the most expensive tuition there is. The transferable lesson is the audit, not the answer: wherever you work next, read the consequences, not the wall.
Return to the central question: is culture a manageable asset or an emergent accident — and what separates cultures that perform from cultures that implode? Four conclusions hold.
Culture is real, and it lives in the assumptions, not the artifacts. What a company says is data about its aspirations; what it consistently does — measures, promotes, tolerates, fires — is the culture.8 Audit the gap: it is the best single predictor in this chapter of both ethical risk and employee cynicism.15
Strength amplifies; it does not sanctify. Strong cultures buy consistency and coordination, and they pay for it in filtered signals — an advantage in stable worlds and a hazard in volatile ones, unless adaptability itself is the shared content.1011
Manage the engines, not the posters. Founder imprinting, attraction–selection–attrition, and socialization built whatever culture you inherit; selection criteria, promotion decisions, consequence systems, and participative norm-setting are the levers that actually move it — slowly, and only when the hard systems agree with the speeches.3912
Hire for values, vary everything else. Fit predicts retention and commitment; similarity masquerading as fit predicts homogeneity and blind spots. Specify the values behaviorally, publish them honestly enough that misfits self-select out, and treat every hiring, promotion, and firing as what it actually is: the next lesson your culture teaches everyone watching.59
Notes appear as superscript numbers in the text and correspond to the numbered sources above. DOIs are provided where available; classic books are cited to their original publishers.