"Culture Eats Strategy for Breakfast"

What Bock argues

The chapter's claim is that culture is not a poster in the break room; it is three things you can actually run an organization on — mission, transparency, and voice — and that the default setting on the second one should be cranked far higher than most leaders are comfortable with. Share the roadmap. Show the OKRs. Let people see the numbers, the decisions, and the reasoning behind them. The argument is that openness is not a perk you grant when things are going well; it is the operating system that makes mission legible and voice real. Strategy is what you intend; culture is what actually happens when no one is watching — and an open culture is the one that keeps the two from drifting apart.

It is a good instinct. What Bock states as a conviction, the research states as a mechanism — and the mechanism is more useful than the conviction, because a mechanism tells you what to measure.

What the research actually says (and where 2015 needs an update)

Transparency "works" is too vague to act on. Here is the precise version, and it has been sitting in the social sciences for decades.

People do not primarily judge a workplace decision by whether the outcome went their way. They judge it by whether the process that produced it was fair — and they judge the process hardest exactly when they cannot see it. That is the fairness-heuristic finding (Lind; Van den Bos): under uncertainty, when information about how a decision was made is missing, people reach for fairness as the heuristic that tells them how much to trust the authority and how much of themselves to invest. Absent real information, they do not suspend judgment. They manufacture it — and the manufactured version is rarely generous. Rumors fill the vacuum is not folk wisdom; it is a measured property of how humans process authority.

The structure underneath it is organizational justice, which the field separates into components — most usefully the four-factor model (Colquitt, 2001): distributive (was the outcome fair), procedural (was the process fair), interpersonal, and informational (were you told the truth, and enough of it). And here is the load-bearing precision the Guide must carry, because it is the part marketing always blurs: procedural and informational justice are what drive the organization-directed outcomes — commitment, trust, willingness to speak up, discretionary effort — not distributive justice. Transparency does not make people love their compensation. It makes them trust the institution that set it, and keep spending the effort no one can compel. That is precisely what a leader wants from "culture," and precisely what a closed process destroys first.

The third leg, voice, only becomes real through a fourth construct Bock gestures at without naming: psychological safety — Amy Edmondson's finding (1999) that a team's shared belief that it is safe to speak up, admit error, and ask the dumb question predicts learning and performance. "Give people voice" is an empty instruction in a team where speaking up is punished. Psychological safety is the precondition that turns the org chart's invitation to speak into people actually speaking.

Where 2015 needs an update: Bock wrote before the black box arrived. The fairness-heuristic result is now the most important thing in the room, because AI decisions about people are, by construction, the largest information vacuum an organization has ever pointed at its highest-stakes calls. That is the argument of Show Your Work: a legible model is not a nicety; it is the operational form of "high transparency," and — because an incumbent built opaque cannot retrofit legibility — it is the moat. Bock's "default to transparency" becomes, in the AI era, transparent by construction or not at all.

How you run it

You measure the two halves and you read the gap.

• Felt fairness: the Colquitt four-factor justice battery, run as a pulse, scored by component — so procedural and informational justice are visible separately from distributive. (The components move differently and demand different fixes; a single "fairness score" hides the diagnosis.)
• Voice precondition: Edmondson's 7-item psychological-safety scale on the same pulse.
• Measured openness: a process-side index of how legible decisions actually are — are OKRs published, are decision criteria written down, can an employee trace a call that affected them. This is the measured half, computed from how the organization actually runs, not from how it feels.

Then pair them. The single most valuable signal is not either number. It is the gap between measured openness and felt fairness, read per team and per manager (min-N gated, always). A team with genuinely open processes but low felt-fairness is telling you the openness isn't reaching them — a communication or trust problem. Low on both is a process problem. High felt-fairness on thin actual openness is a goodwill reserve you are spending without noticing. The gap is the diagnosis; the level alone is just a mood.

The analysis you can execute

This is near-zero net-new engineering, because the spoke exists. Route the Colquitt + Edmondson pulse through the procedural-justice / felt-fairness spoke (PAT-PROCEDURAL-JUSTICE) plus a psychological-safety pulse on the survey orchestrator; segment by team/manager through the privacy gate (min-N); and render the measured-vs-felt gap as the headline. The same paired-measurement machinery the Show Your Work essay describes is the Ch 2 analysis — built once, used by the essay, the book chapter, and the toolbox.

The AI-era turn

Every black-box AI decision about a person is an information vacuum with the authority of a system behind it, and a procedurally-sensitive species fills vacuums with suspicion. So "default to high transparency" stops being a values statement and becomes an engineering requirement: build the systems that touch people so their reasoning can be seen and questioned. Legibility is the one move a better-funded, more-data incumbent structurally cannot copy, because their model was built opaque. Transparency was always the culture lever. In the AI era it is also the competitive one.

What to do Monday

1. Run the Colquitt four-factor battery + Edmondson's psych-safety scale as one short pulse on two or three teams. Score procedural and informational justice separately.
2. Build the crude measured-openness index for those same teams (are the OKRs, criteria, and decision trails actually visible?).
3. Put the two side by side and find the gap. Fix the half the gap points to — and resist the reflex to fix felt-fairness with more communication when the real problem is an unseeable process.
4. For any AI you are about to point at a people decision, ask the Ch-2 question before the accuracy one: can the person it lands on see how it decided? If not, you are manufacturing the vacuum.

Cross-refs: content/magazine/show-your-work.md (the spine); Book 1 Unreliable Ch 6/8 (reliability without validity — agreement is not truth); CHAPTER-MAP.md Ch 10 (procedural justice in pay reuses this same machinery).