Survivorship Bias

We only see what survives. Failed companies don't write business books. Dead soldiers don't give interviews. Losing strategies aren't studied. This systematic blindness distorts our conclusions—often catastrophically.

The classic example: WWII bombers returned with bullet holes concentrated in certain areas. The obvious conclusion: reinforce those areas. The correct conclusion (Abraham Wald): reinforce the areas WITHOUT holes. Planes hit there didn't return.

The survivors were the wrong sample.

The Mechanism

Survivorship bias occurs when:

1. A selection process eliminates some cases
2. Only surviving cases are visible/sampled
3. Conclusions are drawn from this biased sample
4. The conclusions are wrong because the dead aren't represented

The bias is invisible—you can't see what's not there. The sample looks complete because the missing cases leave no trace.

Where It Appears

Business

"Study successful companies to learn success." But for every successful company with trait X, how many failed companies also had trait X? We don't know—they're gone. Business books are written by survivors; the graveyard is silent.

Investing

"This fund has beaten the market for 10 years." But how many funds started 10 years ago? The ones that failed closed. The survivors look impressive; the base rate is hidden.

Music/Art

"Old music was better." No—you only hear the old music that survived. The bad stuff was forgotten. You compare the best of the past to the average of the present.

Architecture

"They don't build them like they used to." The buildings that weren't well-built fell down. You see the survivors. Construction quality looks like it declined; actually you're just seeing selection.

Startups

"Dropouts succeed—look at Gates, Zuckerberg, Jobs." For every successful dropout, thousands failed invisibly. You see the winners because they're visible. The losers disappeared into the graveyard.

Why It's Pernicious

Invisible by Design

The bias hides itself. The dead don't complain. The failed don't write books. The losing strategies aren't documented. Absence of evidence feels like evidence of absence.

Narrative-Compatible

Survivors tell coherent stories about their success. These stories become templates. But the same story might have been told by failures who did the same things and lost.

Selection Looks Like Causation

"Successful people share trait X" might mean "trait X causes success" or might mean "many people have trait X, and some of them randomly succeeded and are now visible."

Corrections

Ask About the Dead

Whenever you're drawing conclusions from survivors, ask: what would the failures look like? How many were there? Did they share the same traits?

Base Rates

Before concluding trait X predicts success, ask: what fraction of people with trait X succeed? The visible success rate is meaningless without the base rate.

Seek Failure Data

Actively look for failure cases. They're harder to find but more informative. The graveyard teaches what the survivors can't.

Pre-Mortems

Before acting, imagine failure and trace backwards. What would have caused it? This surfaces risks that survival stories omit.

— Decoded by DECODER