During World War II, the U.S. military studied bombers that came back from missions and mapped where they'd been hit. The obvious plan was to add armor where the bullet holes were thickest.
The statistician Abraham Wald, working with the Statistical Research Group at Columbia, pointed out the flaw. The planes in the sample were the ones that made it home. Holes in the wings and fuselage showed where a plane could take damage and survive. The places with no holes, like the engines, were where hits brought planes down. Those were the planes nobody got to study. Armor belonged where the returning planes were clean.
That's survivorship bias: drawing conclusions only from what made it through the filter.
The business book problem
Most popular leadership advice is built on the planes that came home. Study ten successful founders, find what they have in common (they woke up at 5 a.m., took big risks, trusted their gut) and turn it into a formula.
What's missing is everyone who did the same things and failed. Lots of people take big risks and trust their gut. We just don't write books about the ones it didn't work for. Without them, you can't tell a cause of success from a trait that's merely common, or from plain luck.
This is why I'm skeptical of a lot of what passes for leadership wisdom. It describes survivors. It doesn't test what separates them from everyone else.
Where it shows up at work
- Copying "best practices." A practice that worked at a famous company may have depended on its market timing, money or talent. Copy the practice without those, and you copy the risk without the upside.
- Hiring for a mold. "Let's hire more people like our top performers" sounds sensible. But if you never look at people who fit the same mold and struggled, you don't know whether the mold predicts anything.
- Promotion decisions. Organizations often promote the people who are visible and confident, then point to the ones who succeeded as proof the system works. The ones who derailed quietly fall out of the story.
- AI models. Train a model only on wins (closed deals, successful hires, top campaigns) and it can't learn what failure looks like. It may also learn the filter itself, like who got hired in the past, instead of what predicts success.
How to look for the missing planes
- Ask who's not in the sample. Before acting on any success story, ask: who did the same thing and failed? If nobody knows, you don't have evidence yet.
- Run post-mortems on failures, not just wins. Failed projects and failed hires carry information that success stories hide: bad assumptions, poor data, misaligned incentives.
- Validate against outcomes, including the bad ones. In selection, this is basic practice. You check whether a predictor separates people who succeeded from people who didn't, not just whether successful people have it.
- Reinforce the weak spots. In leadership development, look for the gaps that sink otherwise strong leaders under pressure, not just the strengths that got them promoted. Derailment research exists for exactly this reason.
- Make it safe to report what went wrong. If people hide failures, your data will only ever contain survivors.
The point
Visible success hides what actually matters. The most useful question in any room full of success stories is the one Wald asked: what are we not seeing?
If you'd like help building hiring or promotion decisions on the full picture, wins and losses both, let's talk.