Survivorship Bias

INTERACTIVEN054
Make a Prediction

This bomber made it home from a mission. Each red dot is a bullet hole. You can add armor to only one area. Where?

Click an area of the plane.

Your Reason

You chose to armor the . What was your main reason?

Look at What’s Missing

You only saw planes that came back. Many didn’t.

Planes that returned

We can count their holes.

Planes that didn’t

?

Their holes never reach the data.

Holes on a returning plane show where it can be hit and still get home.

Reconsider

Knowing these are only the survivors, where would you add the armor now?

Click an area of the plane.

Compare

First choice

After seeing what’s missing

Few holes on the engines doesn’t mean engines rarely got hit. It may mean the planes hit there never came back.

The holes in this interactive are illustrative, not historical counts.

What Really Happened

During World War II, statistician Abraham Wald worked on this problem for the Statistical Research Group at Columbia University.

The tempting idea

Armor the spots where returning planes show the most damage.

Wald’s insight

The data came only from survivors. The missing planes had to be counted too, by asking where hits were rare among the planes that made it back.

The most important evidence was where the returning planes were not hit.

Sources: Abraham Wald, “A Method of Estimating Plane Vulnerability Based on Damage of Survivors,” Statistical Research Group, Columbia University, 1943 (reprinted by the Center for Naval Analyses, 1980); Marc Mangel and Francisco Samaniego, “Abraham Wald’s Work on Aircraft Survivability,” Journal of the American Statistical Association, 1984.

Apply It

A company wants to know why people quit. It interviews only employees who have worked there for 5 or more years. What is missing?

The people who quit are the ones who could explain why, and none of them are in the room.

Apply It Again

A business book studies 50 hugely successful companies and finds that most founders took big risks early. Before deciding risk causes success, which group do you most need to see?

If risky founders also fail often, the book only met the ones who survived.

When data tells a clean story, ask who never made it into the data.