🧬 kbird.ai

Survivorship Bias

3 min read

Survivorship Bias

Type: Selection — Sampling Also Known As: Survivor bias, selection bias

Definition

Focusing only on the “survivors” (successes) while overlooking those that didn’t survive (failures), leading to overly optimistic beliefs about success rates and strategies. We study winners and ignore losers who used the same methods.

“All successful startups worked 80-hour weeks and took big risks.”

Form

  1. A population attempts a task or strategy
  2. Some succeed (survivors), most fail (casualties)
  3. Attention focuses on survivors
  4. Survivor traits are identified as “success factors”
  5. The same traits in failures are ignored
  6. False lessons about success are learned

Examples

Example 1: The WW2 Bomber Problem

The military studied returning bombers to determine where to add armor. Bullet holes clustered in wings and fuselage. The obvious answer? Armor where the holes are.

Abraham Wald’s insight: They needed armor where there WEREN’T holes — those planes didn’t return. The data was from survivors only.

Problem: The missing data (downed planes) was the most important.

Example 2: Startup Success Studies

We study 10 successful startups. They all share: 80-hour weeks, big risks, ignored critics. “This is how you succeed!” But 90% of startups fail. How many failed startups also worked 80-hour weeks and took big risks?

Problem: Failed startups are invisible in success studies.

Example 3: Publishing Industry

“These 12 books were rejected multiple times before becoming bestsellers. Never give up!” True — but how many rejected manuscripts deserved rejection? Thousands of bad manuscripts are rejected; a few good ones are missed.

Problem: We hear about Harry Potter’s 12 rejections, not the thousands of deserved rejections.

Example 4: Music Careers

You hear a hit song on the radio. The artist must be talented and destined for stardom. But those same qualities exist in countless failed artists. You don’t hear their songs.

Problem: We can’t study the qualities of artists who never got a record deal.

Why It Happens

  • Survivors are visible; casualties are invisible
  • Success stories are interesting; failures are boring
  • We want to believe success is controllable
  • Available data comes from survivors
  • It’s easier to study who succeeded than who failed

How to Counter

  1. Ask: “What am I NOT seeing?” For every success, find failures using the same strategy
  2. Seek base rates: What percentage succeed overall?
  3. Study failures: What did unsuccessful attempts have in common?
  4. Random sampling: Don’t let success be the selection criteria
  5. Cemetery analysis: Visit the startup graveyard, not just the unicorns

References

  • Wald, A. (1943). A method of estimating plane vulnerability
  • Denrell, J. (2003). Vicious circles of failure and success
  • Taleb, N.N. (2001). Fooled by Randomness

Part of the Convergence Protocol — Clear thinking for complex times.

Graph View

Backlinks