What is Survivorship Bias?

Also Known As: Survival bias

Definition:

Survivorship bias happens when we concentrate on people or companies that were "selected," while overlooking those that were not selected because of their lack of visibility.

Survivorship Bias Examples:

Listening to Startup Survivors: The world of startups is full of stories of who made it, how, and why. What rarely gets shared are the stories of startups that died following the exact same path of success. If you were to study car rideshare startups, to avoid survivorship bias, you should not only study the difference between Uber and Lyft, but also Sidecar, Split, and Bridj.

Market Research: Let’s say that I am doing market research for a car company. I am trying to determine if my product should include a sunroof. When I analyze all car models that were profitable over the past year, I see that only 10% have a sunroof. Based on this, I may make a decision that a sunroof should not be included in the car model we are designing.

However, what if in the unprofitable segment of car models, only 1% have sunroofs? If I know this data, it may make more sense to include a sun roof on the car we are designing. Looking only at the winners has made us blind to the reality of the situation. We must asses both the winners and losers for the most realistic conclusions.

Marketing Tactics: The problem with most marketing tactics written about online is that they are successful. Not many authors choose to write about marketing tactics that failed. The problem with this is that we are constantly chasing the next greatest tactic. This is otherwise known as shiny object syndrome.

Instead of truly understanding the context of the situation, we are looking for a quick fix to our problems. It's akin to a get rich quick scheme, rather than letting our tactics flow from a well-thought out strategy. The latter takes a lot more effort, which makes it a less appealing decision in the short term.

See also: Selection bias, Cherry picking, Econometrics, Fooled by Randomness, Meta-analysis, Multiple comparisons problem, Selection principle, Texas sharpshooter fallacy