Survivorship Bias

During World War II, engineers studied returning aircraft to decide where to reinforce armor. They mapped bullet holes across planes that had made it back from combat. The pattern seemed obvious: reinforce the areas with the most damage.

Then someone pointed out the mistake.

The data only showed the planes that survived.

This became one of the most famous illustrations of what is now called survivorship bias. The logic sounds simple, but it is easy to miss in practice. The engineers were only looking at aircraft that returned safely. The planes that were shot down never entered the dataset at all. The bullet holes they could see were, paradoxically, the places a plane could be hit and still survive. The truly critical areas were the ones with no surviving data—hits in those regions meant the plane never made it back.

This small correction flips the entire conclusion. Instead of reinforcing the damaged areas, engineers needed to reinforce the areas with no visible damage on surviving aircraft.

The deeper idea extends far beyond aircraft design. Whenever we observe only the “winners” in a system, we risk misunderstanding what produced their success. We see successful startups, published authors, viral videos, or elite athletes, and then try to reverse-engineer their path. But what we do not see are the countless attempts that followed the same strategy and quietly failed before anyone noticed them.

The bias is not in the data itself, but in what is missing from the data.

Survivorship bias works quietly because absence does not draw attention. We notice what exists, not what never had the chance to be seen. This makes the surviving sample feel more representative than it actually is.

The result is a subtle distortion: success stories become overly informative, while failures disappear from view entirely. What remains is a filtered version of reality that feels complete, but is missing most of the system that produced it.

The World War II example is remembered not because it is unusual, but because it is everywhere once you start looking for it. Any time outcomes are visible but failed attempts are hidden, the same pattern can appear.

Sometimes, what seems like evidence is really just the portion that managed to survive.