Survivorship Bias: The Plane, the Graveyard, and Why Success Advice Lies

Survivorship bias is the habit of drawing lessons only from the people, companies, or planes that made it — and quietly deleting everyone who didn't from the sample. The result looks like wisdom. It is usually a filtered fairy tale. Armor the wrong part of the bomber, copy the wrong founder ritual, buy the mutual fund that "beat the market for twenty years," and you are making decisions on a cemetery you cannot see.

This is a money-nerd page about that bias: the World War II airplane story (including what the popular version gets right and what it mythologizes), why startup and self-help advice is structurally dishonest, how markets sell you the winners, and how to catch yourself doing it. Related reading on this site: The Dead Don't Write Memoirs, Your Brain's Cheat Codes, and fat tails.

The plane that wasn't there

The meme version goes like this. In World War II, Allied analysts looked at bombers that returned from missions over Europe. The bullet holes clustered on the fuselage and wings. The obvious move — armor those spots — was wrong. Statistician Abraham Wald, working with Columbia's Statistical Research Group, pointed out that the planes hit in the engines and other critical systems were the ones that never came home. The empty regions on the damage maps were the vulnerable ones. Armor where the holes aren't.¹

That story is directionally right and historically a little too neat. Wald really did work on aircraft survivability for the Statistical Research Group. His memoranda, later reprinted by the Center for Naval Analyses, treat the hard problem: you only observe damage on survivors, so you have to estimate vulnerability for aircraft that never returned. His methods were used in World War II and later in Korea and Vietnam. A 1984 exposition by Mangel and Samaniego in the Journal of the American Statistical Association walks through the math without the internet's cartoon version.¹

What the meme compresses is the actual assignment. Wald was not a lone genius scribbling "armor the engines" on a napkin while generals gasped. He was estimating survival probabilities under missing data, with careful assumptions, in a group that also included some of the century's best applied statisticians. The popular telling still earns its keep as a teaching story because the selection problem is real: the sample of things that survive is not a random sample of the things that tried.

If you only study winners, you will armor the fuselage forever.

The startup graveyard

Modern business advice is a survivorship museum. Founders who IPO write books about waking up at 4:30 a.m., journaling, cold plunging, and "first principles." Founders whose companies quietly dissolve do not get book deals. Venture portfolios publish the logos of the companies that worked. They do not publish a wall of the ninety-something percent that returned nothing.

That is not a conspiracy. It is how attention markets work. Survivors have time, money, and a story. The dead have none of those. So the public corpus of "how to build a company" is written almost entirely by people for whom the path happened to work — which means every ritual that correlated with survival gets sold as a cause.

Correlation is not causation, and selection makes the confusion worse. Maybe the cold plunge helped. Maybe the founder who had time for a cold plunge already had a team, a runway, and a market. Copying the plunge without the runway is cargo-cult entrepreneurship. The missing data are the founders who plunged, journaled, and still ran out of money in month fourteen.

The same pattern shows up in "I dropped out of college and got rich" narratives. You do not hear from the much larger set of people who dropped out and did not get rich, because that story does not travel. Selection on the dependent variable — studying only the successes — is one of the oldest ways to fool yourself with data.²

The mutual-fund time machine

Open a brokerage ad. You will see a chart of a fund that crushed its benchmark for the last decade. What you will not see, in the same frame, is the graveyard of funds that were launched, underperformed, and were quietly merged or closed so they would stop spoiling the firm's average.

This is survivorship bias with a fee schedule. Academic work on mutual-fund performance has long warned that ignoring dead funds inflates apparent industry returns. When researchers put the closed funds back into the sample, the average look much less magical.³ The advertisement is not lying about the surviving fund's past. It is lying by omission about the process that produced the chart you are staring at.

Index funds partly dodge this trap by refusing to pretend stock-picking skill is the product. Active-management marketing leans into the trap, because a curated list of survivors is a better sales pitch than a full census of attempts.

Self-help is a survivor's memoir

Self-help is the soft-focus cousin of the same bias. The person who "manifested" a career breakthrough writes the book. The people who manifested and got nothing do not. Every success framework that starts from a famous outcome and works backward will overfit noise: habits, mantras, morning routines, "energy."

That does not mean habits are useless. It means the evidence base is poisoned unless someone counts the failures. When a guru says "every billionaire does X," ask for the base rate: how many non-billionaires also do X? How many people did X and stayed broke? Without the denominator, the claim is a toast, not a finding.

Our sister essay The Dead Don't Write Memoirs is about the same censoring mechanism in personal narrative. Memoirs are written by people who lived long enough, stayed lucid enough, and found a publisher. History that only reads memoirs will sound more intentional than life is.

Why your brain loves this trap

Survivorship bias flatters two mental habits that usually serve you and occasionally wreck you.

First, storytelling. Humans prefer causal stories to messy distributions. A clean arc — "they armored the wrong place, then a statistician saved them" — sticks harder than "conditional survival probabilities under truncated samples." Your brain will happily promote a tidy anecdote into a law of nature.

Second, availability. The winners are visible. Failures are obscure, private, or never recorded. Whatever is easy to recall feels common. That is why lottery winners are interviewed and lottery losers are not a beat. For more on how your wetware shortcuts reality, see Your Brain's Cheat Codes.

Add fat-tailed outcomes — markets, startups, creative careers — and the distortion gets worse. In fat-tailed domains, a handful of extreme survivors dominate the averages, so learning only from them is especially dangerous. That is the crossover with our fat-tails essay: when the mean is driven by rare monsters, a sample of survivors is not a gentle bias. It is a different universe pretending to be the map.

How to catch yourself

You do not need a Ph.D. in selection models. You need a few rude questions:

• Who is missing from this sample? If the advice comes only from winners, you are looking at bullet holes on the planes that returned.
• What is the base rate? How often does this strategy fail when lots of people try it?
• Is the outcome fat-tailed? If one winner can outweigh a hundred losers, studying the winner teaches you almost nothing about what to expect.
• Would this story have been published if it failed? If the answer is no, treat it as entertainment until someone brings the failures.
• Am I confusing a filter with a cause? "Successful people network" may mean networking helps, or that people who were going to succeed also networked, or both. The slogan will not tell you which.

In investing, prefer structures that do not require you to trust a curated survivor list: broad indexes, costs you can measure, claims that survive contact with the full fund graveyard. In careers, treat founder memoirs as travel writing from a country most tourists never reach. In life advice, borrow experiments, not destinies.

Markets, feeds, and the highlight reel

Social feeds are industrial-scale survivorship engines. You see the apartment renovation that finished, not the three that stalled. You see the trading screenshot from the top, not the account that blew up. Crypto Twitter in a bull market is almost pure selection: every timeline is a museum of people who have not yet been forced offline by a liquidation.

The fix is not to become a nihilist. It is to change what counts as evidence. A single survivor story can inspire. It cannot estimate a probability. For that you need denominators: how many wallets, how many launches, how many people tried the same trade with the same thesis. Without a denominator, "I knew a guy who 100x'd" is ambient noise wearing a suit.

Careers work the same way inside companies. Exit interviews and "how I got promoted" threads over-represent people who stayed on a path that worked. The people who were managed out, or who left after a reorg ate their team, rarely get a keynote. If your model of advancement is built only from internal winners, you will overfit to politics that happened to favor them.

What Wald actually bought you

The durable gift of the airplane story is not a vibe. It is a method: when your data are filtered by survival, the absences are information. Empty cells on the damage map were not proof that engines were safe. They were a clue that engine hits were lethal.

Translate that out of aviation. Empty cells in your information diet — the startups that died, the funds that closed, the books that were never written — are often the most important part of the map. Success advice that never visits the graveyard is not humble. It is incomplete.

Armor the engines. Count the dead. And when a chart, a podcast, or a LinkedIn thread shows you only the planes that came home, ask what the sky looked like for everyone else.

Want more money-nerd pages that refuse the survivor-only sample? Start at the MoneySexNerd shop for the books and guides, or keep reading the money cluster from memoirs and cheat codes.

Sources

1. Abraham Wald's Statistical Research Group memoranda on aircraft vulnerability (WWII), reprinted by the Center for Naval Analyses; Marc Mangel & Francisco J. Samaniego, "Abraham Wald's Work on Aircraft Survivability," Journal of the American Statistical Association 79, no. 386 (1984): 259–267, doi:10.1080/01621459.1984.10478038; accessible PDF exposition at UCSC (Mangel/Wald PDF). Popular retellings (e.g. AMS Feature Column, "The Legend of Abraham Wald," 2016) should be read as legend-plus-math, not as a primary documentary of a single meeting.
2. On selection on the dependent variable as a research design failure, see classic methodological discussions in political science and econometrics (e.g. Barbara Geddes, "How the Cases You Choose Affect the Answers You Get," 1990; and standard warnings in econometrics textbooks against sampling only on success).
3. Mutual-fund survivorship bias: see e.g. Stephen J. Brown, William Goetzmann, Roger G. Ibbotson & Stephen A. Ross, "Survivorship Bias in Performance Studies," Review of Financial Studies 5, no. 4 (1992): 553–580, and the subsequent literature that reconstructs returns including dead funds.