Your brain constructs stories to explain events. Most of those stories are wrong.
In 1943, American bombers returned from combat covered in bullet holes. More holes in the fuselage, fewer in the engines. The military concluded: reinforce the areas with the most holes.
Statistician Abraham Wald saw it differently. The returning planes were survivors. The holes showed where a plane could take damage and still come back. Planes hit in the engines never returned. His advice: armor the engines.
This is survivorship bias -- and it is just one of several cognitive biases that make business decision-making systematically unreliable.
The Narrative Fallacy
Daniel Kahneman coined WYSIATI -- "What You See Is All There Is" -- to describe how System 1 thinking forms judgments based solely on available information, without considering what might be missing. Your confidence in a judgment does not track evidence quality. It tracks story coherence.
Nassim Taleb named the complementary concept: the narrative fallacy. Our tendency to construct causal stories around past events, imposing false order and causality on complex sequences. This is strongest for rare events, which is precisely the category that includes most business successes and failures.
Here is a demonstration of how powerful this is. In one study, participants rated "an earthquake in California causing a flood that kills over 1,000 people" as MORE likely than "a flood in the US that kills over 1,000 people." The vivid causal story felt more probable -- even though it is a strict subset of the broader event.
Seeing Causation Where None Exists
Matute et al. (2015) synthesized 20 years of experiments on what they call "illusions of causality" -- when people believe a causal connection exists between unrelated events.
Their findings are unsettling:
- Even when the actual statistical contingency between cause and outcome is zero, participants systematically rate the causal relationship as positive
- When outcomes occur frequently (high base rate), the illusion gets stronger -- regardless of whether the supposed cause has any relationship to the outcome
- When the supposed cause is present frequently, the illusion intensifies further
- People disproportionately attend to co-occurrence (both cause and outcome present) and underweight cases where the cause is present but the outcome is absent
Translate this to business: entrepreneurs are surrounded by high base-rate outcomes ("companies that do X tend to grow" -- but companies in general tend to grow during bull markets). If both the practice and the growth are common, people perceive a causal link even when none exists. This is why business advice is riddled with false causal claims.
The Halo Effect in Business
Phil Rosenzweig's The Halo Effect (2007, named Business Book of the Year at the Frankfurt Book Fair) documented this systematically.
When a company performs well, observers attribute success to clear strategy, strong values, brilliant leadership, and outstanding execution. When performance drops, the same company is suddenly seen as arrogant with risky strategy and stifling culture.
The attributes did not change. The performance changed, and the attributions followed.
His example of Cisco Systems is instructive. During rapid growth in the late 1990s: praised for brilliant acquisitions, masterful strategy, superb customer focus. After the tech bubble burst: accused of arrogance and reckless expansion. Cisco had not fundamentally changed. But the narratives about it reversed completely based on outcomes.
Rosenzweig identified nine "business delusions" including the Delusion of Correlation and Causality, the Delusion of Single Explanations, and the Delusion of Rigorous Research. His critique of business bestsellers -- Built to Last, Good to Great, In Search of Excellence -- is that they relied on post-hoc attributions based on already-known performance, not genuine causal analysis.
Hindsight Makes It Worse
Hindsight bias -- the tendency to view past events as more predictable than they were -- has been studied for 50 years.
Guilbault et al. (2004) meta-analyzed 95 studies with 252 independent effect sizes. The overall effect: Md = .39 (95% CI: .36 to .42). A consistent, reliable bias.
The most troubling finding: manipulations designed to reduce hindsight bias did NOT produce significantly lower effect sizes. The bias resists correction.
Roese & Vohs (2012) identified three escalating levels: memory distortion ("I said it would happen"), inevitability ("It had to happen"), and foreseeability ("I knew it would happen"). Each level compounds the narrative that past outcomes were predictable -- which inflates confidence in future predictions.
Even Experts See Patterns That Are Not There
Chapman & Chapman (1969) demonstrated that experienced clinicians perceived correlations between diagnostic indicators that did not exist in the data. Training designed to reduce this illusory correlation had minimal effect. Expertise did not protect against the bias.
Gilovich et al. (1985) showed a similar pattern with the "hot hand" in basketball. Despite universal belief among players, coaches, and fans that players go on shooting "streaks," the statistical analysis of actual game data showed success rates consistent with independent random trials. The streaks were illusory.
People expect an alternation rate of about 0.7 in random sequences -- when the true rate is 0.5. This causes genuinely random patterns to look non-random. In business, this means a streak of successes from a particular practice feels meaningful even when it is just noise.
Why We Are Wired This Way
Foster & Kokko (2009) provided the evolutionary explanation. They built a mathematical model showing that natural selection favors cognition that commits frequent false positives in causal assessment -- as long as the occasional correct detection carries a large fitness benefit.
When the cost of missing a real tiger (false negative) exceeds the cost of fleeing from a shadow (false positive), evolution selects for shadow-fleeing. Our brains evolved to over-detect patterns and causal relationships because in ancestral environments, that asymmetric cost structure made it adaptive.
In business, this same mechanism causes us to see "success formulas" where there are only coincidences.
Confirmation Bias Locks It In
Nickerson (1998) wrote the definitive review of confirmation bias: the seeking or interpreting of evidence in ways partial to existing beliefs.
Once you construct a narrative about why something works, you systematically seek evidence confirming that narrative and ignore or discount disconfirming evidence. The narrative fallacy is self-reinforcing. The story you build becomes the lens through which you filter all future information.
This creates a particularly dangerous cycle for entrepreneurs. Form a theory about what drives your business. Seek confirming evidence (easy to find). Ignore disconfirming evidence (feels irrelevant). Conclude your theory is validated. Make bigger bets based on it.
Cooper, Woo & Dunkelberg (1988) surveyed 2,994 entrepreneurs. 81% believed their chances of success were at least 70%. A third rated their chances at 100%. Meanwhile, fewer than half of firms survive past five years.
Rationalization Is Automatic
Jarcho, Berkman & Lieberman (2011) published the first fMRI study of rationalization during decision-making. The finding: brain activity associated with post-hoc rationalization occurs at the moment of the decision itself, not during slow deliberation afterward.
By the time entrepreneurs explain their reasoning to others -- or to themselves -- the post-hoc narrative is already constructed. This is not dishonesty. It is an automatic cognitive process.
Eyster, Li & Ridout (2021) formalized this in economic terms: ex post rationalization predicts sunk-cost effects. When an agent has made an ex post mistake, they rationalize by doubling down on the original course of action.
What Actually Helps
The debiasing evidence is mixed but not hopeless:
- Consider alternative explanations. Roese & Vohs (2012) found that actively generating alternative causal stories reduces hindsight bias. Before attributing a success or failure to one cause, force yourself to generate three alternative explanations.
- Demand base rates. Matute et al.'s work shows that causal illusions are strongest when you ignore how often the outcome happens anyway. Before crediting a practice with an outcome, ask: how often does this outcome occur without the practice?
- Seek disconfirming evidence deliberately. The Nickerson review makes clear that information search defaults to confirmation. Override it by specifically asking: what evidence would prove me wrong?
- Study the failures, not the successes. Denrell (2003) showed that learning from survivors creates systematic bias. The real information is in the companies that tried the same things and failed -- but they are invisible because they did not survive into the sample.
None of these strategies eliminate narrative bias. The Guilbault meta-analysis showed that interventions to reduce hindsight bias had disappointingly small effects. But awareness combined with structured debiasing practices can reduce the damage.
The irony of this post is not lost on me. I have just told you a story about why stories are unreliable. Taleb acknowledged the same contradiction: the most effective way to communicate the dangers of narrative is through narrative.
The difference is whether you treat the story as evidence or as a hypothesis to be tested.