Book Summary
Daniel Kahneman’s Noise — written with Olivier Sibony and Cass R. Sunstein — is about the error nobody talks about. A decade after Thinking, Fast and Slow taught the world about bias, the systematic tilt in human judgment, this 2021 follow-up takes on its neglected twin: noise, the random, unwanted variability in judgments that should be identical. Bias is the arrow cluster landing left of the bullseye; noise is the arrows scattered everywhere. The book’s unsettling claim is that noise is everywhere, costs as much as bias, and is almost entirely invisible to the people producing it.
The evidence is drawn from professions that cannot afford inconsistency. Judges handed the same case file give wildly different sentences — and the same judge gives different sentences before lunch than after. Underwriters at the same insurance company, shown the same risk, set premiums differing by more than half. Fingerprint examiners, shown prints they’d already judged, contradict their own earlier conclusions a significant fraction of the time. Radiologists disagree about the same scans. In study after study, the variability within a single professional’s judgments rivals the variability between professionals. Wherever human judgment operates at scale — courts, medicine, hiring, lending — noise is quietly taxing every decision.
The authors distinguish three flavors. Level noise is simple disagreement between judges. Pattern noise is the subtler kind: the same judge is harsher on some types of cases than colleagues are — one underwriter is tough on tech startups but lenient on restaurants, another the reverse. And occasion noise is the judge in all of us: mood, weather, fatigue, the case that came before. The point isn’t that professionals are incompetent; it’s that the human judgment machine was never built for consistency, and every institution that depends on it is running noisier than it believes.
The second half of the book is the fix: decision hygiene. The authors prescribe noise audits (measure the variability in your own judgments before theorizing about it), structured and sequenced evaluations (form independent judgments before discussing, to prevent anchoring), aggregation (average independent judgments rather than debating toward consensus), and — the most controversial recommendation — replacing human judgment with rules and algorithms wherever the decision is repeatable. Paul Meehl’s decades-old finding gets its due: simple statistical models beat expert judgment in prediction after prediction. For investors, Noise is a manual for the least glamorous and most profitable upgrade available: making your own process less random.
Who is Daniel Kahneman?
Daniel Kahneman (1934–2024) was the psychologist who rewired economics. Working with Amos Tversky, he demonstrated that human decision-making systematically violates the rational-actor model — prospect theory, loss aversion, anchoring, the planning fallacy — work that earned him the 2002 Nobel Prize in Economics (Tversky had died in 1996 and Nobels aren’t awarded posthumously). His 2011 bestseller Thinking, Fast and Slow became one of the most influential psychology books ever written, and Noise is its direct sequel — the error term he felt he’d underemphasized the first time around.
His coauthors bring the applied side. Olivier Sibony is a former McKinsey senior partner who spent a career watching corporate decisions go wrong in predictable ways; Cass Sunstein is the legal scholar and former Obama administration official who co-wrote Nudge and has spent decades designing choice architecture for governments. The trio spans the lab, the boardroom, and the statehouse — which explains both the book’s range and, occasionally, its committee feel.
Kahneman died in March 2024 in Switzerland, having chosen assisted dying — a fact his family disclosed afterward, and a final decision entirely consistent with a life spent thinking clearly about hard trade-offs. Noise stands as the capstone of his applied work: less famous than its predecessor, more directly useful.
Lessons From Noise
Your process is noisier than you think — audit it before defending it. The book’s central practical move is the noise audit: take a set of past decisions, strip the identifying details, and have the same people re-judge them. The gap between the two rounds is your noise, and it is always larger than expected. Investors can run a crude version today: revisit your old buy/sell theses blind and see how often present-you disagrees with past-you on the same facts. The disagreement is data.
Sequence judgments to defeat anchoring. When a committee discusses a case before voting, the first speaker’s view contaminates everyone else’s — that’s occasion noise manufactured by process. The fix is mechanical: independent written judgments first, discussion second. For the solo investor, the equivalent is writing down your thesis and price target before reading anyone else’s analysis, checking the price, or talking to friends. Information order is a choice; choose the one that protects independence.
Aggregate; don’t debate toward consensus. Averaging independent judgments consistently beats the judgment of the group after discussion — discussion mostly amplifies the loudest voice. This is the statistical case for investment clubs and committees done right (independent theses, then average the conclusions), and the case against the guru model. Diversity of independent views isn’t just polite; it’s mathematically superior.
Use rules for repeatable decisions. Meehl’s finding — simple models beat experts at prediction — is the book’s most actionable provocation. Rebalancing rules, position-size limits, buy/sell checklists: these aren’t training wheels, they’re noise reduction. The expert adds value in the rare, genuinely novel situation; everywhere else, the checklist outperforms the gut. The resistance to this idea is itself a bias worth noticing.
Distinguish the two errors in yourself. Bias makes you wrong in a consistent direction; noise makes you wrong at random. Most investors obsess over bias — am I too optimistic? too anchored? — while ignoring that their judgments would scatter even with bias removed. Noise‘s contribution is making the second error visible. You can’t fix what you can’t see, and until this book, almost nobody was looking.
Criticisms of the Book
It’s long and repetitive. Three eminent authors turn out to write like a committee: the core distinction (bias vs. noise) is established early and then re-established, with new examples, for hundreds of pages. Readers of Thinking, Fast and Slow will feel the déjà vu particularly — much of the psychology here is familiar, and the genuinely new material (the noise studies, decision hygiene) could have been a tight 200 pages.
The tone occasionally slips into management consulting. Chapters on “decision hygiene” for organizations read like McKinsey decks rendered as prose, and the recommendations — structured interviews, independent assessments — will feel obvious to readers who already run disciplined processes. The book is most valuable to the precise audience least likely to read it: confident experts who don’t think they need it.
The book underplays when noise is useful. Variability isn’t always a flaw: in creative work, in exploration, in entrepreneurship, a little randomness is how systems discover things rules would never find. The authors acknowledge this in passing, but the frame — noise as pure cost — is too narrow. Evolution runs on noise; so do markets. The question isn’t how to eliminate variability but where it belongs.
Who is This Book For?
Investors first, without question. Noise is the rare book that attacks the actual machinery of investment decisions rather than the theories about markets. If you run a checklist, sit on an investment committee, or have ever wondered why your own past decisions look so foreign in hindsight, this book will pay for itself many times over. Read it after Thinking, Fast and Slow, not before.
Managers, hiring committees, doctors, lawyers — anyone whose job is judgment at scale. The noise audit concept alone justifies the purchase for organizations that make repeated high-stakes calls. Casual readers should be warned: this is the drier, more technical sibling. It’s worth it, but it’s work.
And a note for the skeptics: the most common objection — “but my judgment is the product” — is precisely what the noise research dissolves. Your judgment is the product, which is why its variability is your problem. Professionals in every field studied believed their decisions were consistent right up until the audit proved otherwise. Confidence is not evidence of consistency; it’s often evidence of not having checked. Start with your ten most consequential decisions of the last year and re-judge them blind. The audit is cheap. The noise is expensive.
Final Thoughts
Noise didn’t get the cultural reception of its predecessor, which is a shame — it’s the more useful book. Bias makes for better cocktail conversation; noise makes for better decisions. Kahneman spent his career showing that the mind is a flawed instrument, and his final major work is essentially an owner’s manual for the flaw nobody wanted to admit.
The investor’s version of the book’s message fits in a sentence: make your process boring. Independent judgments, written theses, checklists, rules, aggregation — the unglamorous machinery of consistency. Excitement in process is noise wearing a costume. In a field where the winners are so often just the least random, that’s not small advice. An essential companion to the behavioral canon — slightly overlong, entirely worth it.
