Decision-Making and Judgment
The best answer I have ever heard in an assessment was a defense of a failure. A country-manager candidate was walking us through an expansion she had led, a market entry that had lost money and been unwound after two years, and instead of the usual repentance, she said something that made the client's HR director visibly flinch: "That was a good decision that went wrong, and given the same information I would make it again." Then she proved it. She reconstructed what was known at the time, what was unknowable, what the option set had actually been, the asymmetry between the cost of trying and the cost of ceding the market to the competitor who did try, and she finished with the one sentence candidates almost never say: "We also got unlucky, and I can show you where."
The room went cold, and the room was wrong. The client wanted to score the outcome; she was showing us the deciding, and the deciding was excellent. We argued for her, she was hired, and she has since made the client a great deal of money by making, presumably, more good decisions, some of which will also go wrong. I begin with her because this cluster's central discipline, separating the quality of a decision from the quality of its outcome, is one that assessors fail as often as candidates, and a chapter that teaches you to catch the resulting fallacy in a candidate's stories had better first install it in your own scoring.
What it is, and what it predicts
Define the cluster as the capability to run the act of deciding well under uncertainty — where Chapter 9 assessed whether commitments land, this assesses how the commitments were chosen. Five collectable components. Framing: turning a mess into a decidable question — what kind of decision is this, what actually hangs on it, is this door one-way or two-way. Speed calibration: matching deliberation to stakes and reversibility — knowing when deciding beats analyzing, and when it does not; the mark of the real practitioner is variance, fast where fast is right and slow where slow is, rather than a uniform personal tempo worn as a brand. Working under uncertainty: carrying probabilities honestly — scenarios, pre-mortems, tripwires, the explicit what would change my mind. Updating: revising on contrary evidence without either of the twin failures — rigidity (escalation of commitment, the best-documented decision pathology in the managerial literature) or capitulation (the leader who folds at the first frown). And decision conduct: how the deciding is run socially — whether dissent is genuinely invited or ceremonially staged, whether the option set is real or a ratification parade, and whether, at the end, the call is visibly owned — because a recommendation is a claim someone answers for, and a decision whose author cannot be located is this cluster's deepest failure, at any altitude.
Chapter 8 flagged the border honestly: this cluster leans against raw cognitive capability, and the book keeps it in Part III because at executive level the differentiator is decision conduct. Everyone on your shortlist can think; not everyone can decide, in company, under uncertainty, out loud, and own it. What the cluster predicts: judgment-and-decision competence tracks managerial performance in the JDM research tradition; escalation of commitment has a documented body count; and the forecasting literature, Tetlock's calibration work above all, shows that updating discipline separates good judges from confident ones, a finding with obvious executive translation even though its subjects were forecasters. Evidence grade: components well-supported, largely from laboratory and non-executive field settings; the executive packaging is this book's synthesis.
And now this cluster's epistemic hazard, the fifth in Part III's series, and the one that dictates its method. After attribution, the inherited machine, the rigged sample, and the borrowed screenplay: the self-theory. When you ask a candidate how they decide, you do not retrieve the deciding; you retrieve their theory of themselves as a decider, and the theory is corrupted at the source, because hindsight rewrites deliberation. The memory research is unambiguous: people cannot accurately reconstruct their own pre-outcome reasoning; the doubt, the near-miss alternatives, the fog, memory edits them into a clean line pointing at what happened. The candidate is not lying. The candidate no longer has access to the truth you are asking for. Which is why this is the cluster the evidence flags for a different method: you do not ask about deciding; you watch it.
How it fakes
The self-theory recital, first and everywhere: data-driven, first principles, one-way doors, disagree-and-commit — the decision-culture vocabulary is as free as every other vocabulary in this book, and a candidate can hold the complete Bezos lexicon without ever having carried an irreversible call. Outcome laundering: good outcomes retold as good decisions and bad outcomes as bad ones — the resulting fallacy, performed fluently because the candidate's own memory performs it first. The decisiveness show: speed theater for the interviewer who wants boldness — and its mirror, the deliberation show for the interviewer who radiates rigor; candidates read which decision aesthetic the room rewards and perform it, which is Chapter 5's criteria-reading operating on its favorite terrain. And hindsight smoothing: the moment of decision rendered cleaner than it was lived — doubt edited out, the fog dissolved, the choice obvious in the telling. Smoothness, as in Chapter 13, is the tell; real decisions have fog in them, and the fog is exactly what you will probe for.
The probe architecture — and where it ends
The four layers still run, but read them here as triangulation in support of observation, because the interview alone cannot carry this cluster.
Layer 1 — standardized stimulus: "Take me to a consequential decision you made with genuinely incomplete information. I want three lists: what you knew, what you didn't know, and what you decided you couldn't wait to learn." The three-list structure is the anti-smoothing device — it forces the fog back into the story.
Layer 2 — planned probes: What were the real options — was there a living option set, or a preferred answer gathering escorts? Who disagreed, and what happened to them and to their argument? What did you believe would happen — in odds, if you can? What was your tripwire — what event would have made you reverse? How did you announce the call, and what did you tell the people whose option lost?
Layer 3 — verification probes: Is there paper — a decision memo, a pre-mortem, a board minute? (Real deciders leave artifacts; the presence of a written what-would-change-my-mind is one of the strongest single documents a search file can contain.) Who was in the room? And the hook that matters most: name the strongest dissenter — would they take my call?
Layer 4 — the signature pair, the purest in Part III because it tests the cluster's core in two questions. First: "A good decision that produced a bad outcome — walk me through it, and defend the decision." Then: "A bad decision that produced a good outcome — where you got lucky, and know it." The pair is the process–outcome separation, administered directly. Candidates who cannot inhabit the first question — who keep apologizing for the outcome while you are asking about the deciding — are showing you that their self-evaluation runs on results, which predicts both overconfidence after wins and over-correction after losses. Candidates who answer the second with precision, volunteering the luck the way our country manager did, are demonstrating the rarest calibration there is: an honest ledger kept against oneself. Add the updating probes — a decision you reversed, and what reversing cost you; something you stayed with a year too long — and the interview has done what an interview can do here.
Then hand the cluster to Chapter 6's methods, because the evidence insists on it. This is the simulation-flagged cluster: the strategy case with the mid-exercise premise shift, the prioritization in-basket, the board simulation's adversarial Q&A — designed, anchored, and scored per Chapter 6. What the exercises let you watch, which no narrative can retrieve: framing behavior — what the candidate does in the first ten minutes with a mess (the first five information requests are the diagnostic reflex, live); the update moment — the premise shift lands, and you observe, in real time, integration versus rigidity versus collapse, the exact behavior Layer 4 could only ask about; decision timing — whether a decision arrives at all, because analysis paralysis cannot hide in a timed exercise the way it hides in a narrated career; and how uncertainty is carried in the recommendation — honest caveats attached to a real commitment, versus the hedge-everything memo that commits to nothing, versus false certainty. One live sample of these four behaviors outweighs an afternoon of decision stories, and that sentence is this chapter's method in miniature.
In the room, and only outside it
Beyond the exercises, two live samples survive honestly in the interview itself. The live update: while discussing one of the candidate's own past decisions, introduce a contrary consideration — "suppose I told you the Gulf distributor data had said the opposite; does that move you?" — and watch the machinery run on their most defended territory. And the language of likelihood: listen for whether probability talk occurs naturally — I thought it was maybe two chances in three; the downside was capped; we'd know within a quarter — or whether the world arrives in certainties. Nobody fakes probabilistic habits for a full afternoon.
What only the outside holds: whether dissent was ever real (the named dissenter, reached through Chapter 7's second ring, is this cluster's counterparty); the artifacts, which either exist or were never written; and the track record of outcomes, which you will read, please, under the chapter's own discipline. An executive career contains few genuinely large decisions; small samples are luck's playground; and an assessor who scores the outcome column without reconstructing the decision column is committing, in the file, the exact fallacy this cluster exists to detect in candidates. Our country manager's unwound market entry belongs in her evidence for, not against, and it took reconstruction, not results, to know that.
Red flags — with their innocent explanations
Nothing ever reversed. Innocent: a short observation window, or genuinely strong calls throughout a lucky decade. But across a full senior career, no reversals means no updating or no honesty about it, tripwires that never existed, or fired unheard. Everything was obvious in retrospect. Innocent: hindsight compression happens to every memory, including yours; probe for the lived doubt before scoring, because the flag is when Layer 2's fog-probes find no fog anywhere. Framework fluency without artifacts. Innocent: some fine deciders theorize badly and decide beautifully, and the exercises adjudicate. Uniform tempo. Innocent: turnaround careers legitimately skew fast. But calibration means variance; a candidate whose every story resolves at the same speed has a style, not a judgment. Outcomes externalized. Bad results consistently attributed to execution, markets, or successors, innocent once or twice, because those things are real; as a pattern, it is the ledger kept against everyone but oneself, and it routes forward to Part IV.
The reference question
Through the two-level method, second ring anchored on the named dissenter. To the dissenter: "You argued against [the decision]. How was your dissent actually handled — and did the process change your mind about anything, even if not the call?" (A dissenter who lost the argument and still respects the deciding is this cluster's version of Chapter 13's right-resister signal.) Frequency, to a former report: "How often did you see them change course on evidence — and how did they treat whoever brought the evidence?" Calibration, with this cluster's distinctive fork: "If you had one irreversible decision to make next year, would you want them in the room — deciding, or advising?" The fork is the finding: markets are full of brilliant advisors who should never hold the pen, and references know exactly which kind they worked for, even when the CV does not.
Scorecard anchors
2 — Below the bar. Self-theory without artifacts; outcomes laundered into process judgments both ways; the good-decision-bad-outcome question produces apology instead of defense; no reversals, no named dissenters, no fog anywhere in the telling; in exercise, the premise shift produces rigidity, collapse, or a recommendation that never commits.
3 — At the bar. Verified consequential decisions with living option sets; at least one competent defense of a well-made loss; dissent demonstrably heard, confirmed from the dissent's side; probability language present; in exercise, the shift is integrated without drama and a real recommendation arrives on time, uncertainty attached honestly.
4 — Distinguishing strength. The process–outcome separation runs unprompted in both directions — including the volunteered lucky win; verified tempo variance across episodes, each speed defensible; artifacts exist (the written pre-mortem, the tripwire that fired and was obeyed); dissenters sought out, by their own testimony; and in exercise, the rare mark — framing that improves the exercise itself, the candidate finding the sharper question inside the one you set. The deciding, watched live, that you would want in the room when the door is one-way.
The seam with Part IV
Beneath the conduct sits the person's relation to being wrong, and this cluster grazes it constantly without being able to score it. The executive who cannot inhabit the lucky-win question is often not evading you; something in them cannot afford the sentence. The decide-first-justify-later pattern, decision-based evidence-making, is rarely a reasoning defect; it is a certainty need, somewhere the reasoning never reaches. Risk appetite itself splits the same way: is the boldness situational judgment or identity's signature, applied to every situation regardless? These are Chapter 20's decisional-evidence questions in their natural habitat; indeed Part IV will reuse this chapter's episodes wholesale, reading the same decisions for what they optimized rather than how they were made. Flag, hand forward, and keep the registers separate: this chapter scores whether they can run the deciding; Part IV asks what the deciding is in service of.
Where the rules run out
The honesty paragraph. The component evidence is strong and mostly not from the C-suite: JDM research is lab-heavy, escalation studies skew managerial, and the calibration findings come from forecasters, not CFOs; the executive translation is reasoned, and marked. The simulation requirement rests on the self-theory hazard plus Chapter 6's incremental-validity evidence, not on a dedicated executive-decision-assessment literature, which barely exists. Outcome track records at this altitude are noise-dominated small samples, and the chapter's own discipline binds the assessor scoring them. The anchors are unvalidated; Chapter 25's log, which is, notice, exactly this cluster's tripwire-and-update discipline applied to the assessor's own decisions, is the long-run judge. It would be strange if it were otherwise: a chapter about judgment that exempted its own judgments from the standard would deserve neither.
Notes and sources
Evidence grades: [M] meta-analytic/systematic; [L] peer-reviewed primary; [S] credible practitioner/survey; [T] flagged synthesis/inference.
- Hindsight bias and the reconstruction of past reasoning: Fischhoff's hindsight research tradition and successors. [L/M]
- Outcome bias in evaluating decisions: Baron & Hershey (1988) and the resulting-fallacy literature. [L]
- Escalation of commitment: Staw's research program and meta-analytic successors. [M]
- Calibration and updating discipline: Tetlock, Expert Political Judgment (2005); Tetlock & Gardner, Superforecasting (2015) — forecaster samples, executive translation flagged. [L/S]
- Simulation necessity and incremental validity for enacted judgment: Chapter 6's sources (Lievens & Patterson, 2011; AC guidelines; premise-shift design per the underlying research templates). [M/L]
- Criteria-reading and decision-aesthetic performance: Kleinmann's ATIC research; Levashina & Campion — Chapter 5's sources. [L/M]
- Decision conduct, dissent, and voice: Detert & Burris; Morrison — shared with Chapter 12. [L/M]
- Speed–reversibility calibration vocabulary: practitioner decision culture (used as object of analysis). [S]
- The three-list stimulus, the signature pair as administered instrument, the deciding-vs-advising fork, five-part packaging: this book's synthesis from practice. [T]