When a sample has been quietly filtered — the failed funds closed, the dead startups forgotten, the bombers that never came back — the winners that remain can't tell you the whole story. Often they tell you the exact opposite of the truth. This is survivorship bias: one of the most common and most expensive analysis traps there is. The fix is a habit of mind — always asking what got filtered out before you copy what's left.
In 1943 the statistician Abraham Wald was asked where to add armour on returning bombers, based on where they came home riddled with bullet holes. His answer inverted the question. Below is the same puzzle — a map of every plane that made it back. Press the controls and see whether you reach his conclusion.
These are the survivors, mapped hit-by-hit.
Survivorship bias is a species of selection bias: your sample has been quietly filtered by who survived. That makes it a cousin of — but not the same as — the more famous "correlation isn't causation". Naming the trap correctly tells you where to look next.
| The trap | What's wrong with the data | The tell | The fix |
|---|---|---|---|
| Survivorship bias | The sample is filtered by survival — the losers are simply gone. | You're studying only what's left standing: winners, survivors, what came back. | Go and find the dead — the funds that closed, the startups that died, the planes that didn't return. |
| Selection bias (the parent) | The sample isn't representative — some process decided who got in. | Your data was gathered in a way that skews who appears in it. | Ask how the sample was assembled, and who could never have been included. |
| Correlation ≠ causation | The data is all there — the link is misread. | Two things move together and you assume one drives the other. | Hunt for a confounder or a reversed arrow. Nothing's missing; the story is wrong. |
Every trait the survivors share — grit, a 5am alarm, a dropped-out degree. The trouble is the losers often share those traits too. Studied alone, survivors overstate the causes of success and quietly hide the role of luck.
The base rate, the failures with the very same traits, the fatal hits no survivor carries. This is where the real answer lives. The whole discipline in one line: study the dead, not just the living.
Here are five of the most famous founders alive, and one thing they share: none of them finished their degree. Draw the obvious lesson — then widen the lens and watch it fall apart.
So the lesson is… drop out?
Here is a stack of claims stated with total confidence. For each one, decide: is the conclusion distorted because you're only seeing the survivors — or is the full population actually in view? Watch for the ones that study the failures on purpose; those are the antidote, not the trap.
The same filtered sample turns up across investing, self-help, product and engineering. In every case the fix is identical: go and find the data that got filtered out.
"The average fund in our category returned 9% a year." Funds that performed badly get quietly closed or merged out of existence — and out of the average.
Billionaires wake at 5am, read 50 books, dropped out, took the big risk. So do that. The advice is reverse-engineered purely from the people it worked out for.
"They shipped with no funding, no marketing, and it went viral — so skip the marketing." The case study is told by the one bet in a thousand that surfaced.
"This service has never fallen over — copy its design." The systems still standing may be well-built, or may simply have been lucky enough to dodge the load that broke the others.
A study picks a dozen standout firms and distils the traits they share into a formula for greatness — with no comparison group at all.
Survivorship bias is easy to nod along to and startlingly hard to catch in the moment — because the missing data is, by definition, the data that isn't in front of you. Teams that build the reflex make better bets: they price in the failures, discount the hero stories, and go hunting for who isn't in the room. That's the work we do with the leaders and teams we coach — turning a sharp idea into a habit that changes decisions.
Talk to us about coachingBefore you copy the winners, ask what got filtered out to leave only them. The answer you need is usually in the gap, not the sample.
The traits survivors share are also shared by plenty who failed. Shared success traits are candidates to test, never proof of cause.
The funds that closed, the startups that died, the people who quit and the planes that didn't return. Go and study the dead, not just the living.