Why Role Models Can Mislead: Survivorship Bias and Success Stories

Why success stories are so convincing

Whoever runs a business is surrounded by success stories. Biographies of famous founders, conference talks, portraits in the business press, and posts on social networks tell how someone bet everything on one card, acted against all advice, and won in the end. Such stories are instructive and motivating. But they have a property that is easily overlooked: they are told almost exclusively by those for whom things went well.

This essay therefore pursues one question: Why can role models and success stories mislead entrepreneurs, and how can one still learn from them sensibly?

To answer it, I first clarify the concept of survivorship bias and introduce a model that explains why risky strategies in particular look better than they are as a result. Then I weigh what role models still achieve and end with a check that every success story should pass before it is imitated.

What is survivorship bias?

Survivorship bias describes a thinking error: one draws conclusions from cases that survived a selection process and overlooks those that dropped out along the way. Because those who dropped out are missing, characteristics of the survivors appear to be the cause of their success, even though the failures had the same characteristics.

A well-known example from the Second World War is often attributed to the statistician Abraham Wald. The aim was to reinforce aircraft where returning planes had the most bullet holes. The objection: the holes show precisely the places where a hit was survivable. Aircraft hit in other places did not come back and were therefore missing from the count. The story is told in many versions, but its core is exactly the error discussed here.

On this blog the error appears in several places. The article hidden champions names it as an objection to learning from little-known world market leaders: whoever looks only at successful specialists overlooks those who failed with the same strategy. It belongs to a family of thinking errors described in the article cognitive biases of business owners.

The thinking model: Denrell's undersampling of failure

In 2003 the organizational researcher Jerker Denrell showed that the error does not just produce random mistakes but has a particular direction. His starting point: organizations learn from one another by observing and imitating. But the observations are a biased sample. First, the companies one can see are the survivors of a selection process that has eliminated a large share. Second, books and the business press concentrate on successful companies. Failure is therefore undersampled in what one sees.

The most important conclusion concerns risk. Risky approaches lead more often to very good and to very bad results. They are therefore overrepresented among the best and among the worst. If one sees only the best, because the worst have disappeared, risk appears to be associated with success, even if there is no association in the group as a whole. Denrell concludes that observing existing companies systematically makes unreliable, all-in approaches look better than reliable and diversified ones.

An everyday analogy: whoever asks only the guests at the casino bar in the evening who have just won and are buying a round gets a skewed picture of roulette. The losers went home long ago.

What this means for entrepreneurs' success stories

For owners, Denrell's model has an uncomfortable consequence. Many of the pieces of advice spread as lessons from success stories have exactly the properties the bias favors: betting everything on one card, acting against the advice of others, borrowing quickly and heavily, doing without control and planning. Among the survivors such stories are common. Among the failures they would probably be just as common, but no one asks there.

A finding from research on learning from others' failures fits this. Kim and Miner (2007) examined whether banks in the United States learn from the failures and near-failures of other banks. Experience from one's own regional market had a higher value for survival than experience from other regions, and other banks' near-failures were more instructive than actual failures. Whoever deliberately looks at failures nearby thus learns something that is in no success story.

The portraits on this blog, for instance of Adi Dassler or Robert Bosch, are therefore meant as cases, not recipes. The article on Dassler explicitly warns against inferring the benefit of a founders' dispute from the success of two successor companies, because the many businesses that broke apart over such a dispute appear in no success story.

Whoever listens only to winners mainly learns how winners talk about risk after the fact.

Two views: role models inspire, or role models deceive

The first view stresses the benefit of role models. According to Albert Bandura's social learning theory, people learn a great deal by observation. A role model shows that something is possible and strengthens the belief that one can do it oneself. A survey of 494 students from rural regions of China (Du 2024) found that interest in distant entrepreneurial role models predicted the intention to found a business, mediated by confidence in one's own entrepreneurial abilities. Interest in close role models was also associated with the intention to found, but not through this path.

The second view stresses the danger. Role models deliver stories, and stories are told from the end. What looks like foresight in hindsight was often luck, and what looks like courage was recklessness among the failures. In favor of this view are Denrell's model and the experience that recipes for success are rarely transferable, because market, timing, and capital were different.

I consider both views correct, but for different purposes. For encouragement, role models are valuable. They show that a path is passable, and for that the story does not need to be representative. For the decision on a strategy they are dangerous, because here exactly the question counts that a success story does not answer: how often does this go wrong? A qualitative study from Estonia (Demiankova 2024) also reminds us that negative role models teach too. Founders there rejected certain practices precisely because they had seen in others where they led.

The Role Model Check: four questions before imitating

From the model, a check can be derived that I call the Role Model Check. It takes a few minutes and is suitable for any success story that is meant to lead to a decision in your own business.

1. Who failed?

   Ask who did the same thing and failed. If you know no such cases, that does not mean there are none, but that you cannot see them. Ask associations, chambers, or experienced colleagues about failures.

2. How risky was it?

   Check whether the story is about a strategy that can lead to very good or very bad results. Precisely such strategies look best in success stories.

3. What was the environment?

   Separate what the role model could influence from what fell into their lap: timing, market, capital, luck. Only the first part is transferable.

4. What behavior remains?

   Adopt behaviors, not stories. A role model who talks to customers every day provides a checkable habit. A role model who bet everything on one card provides only an outcome.

• I know at least one case in which the same strategy failed.
• I know how much the role model risked and what a failure would mean for my business.
• I have separated what was due to the role model and what to timing and environment.
• I am adopting a habit or a procedure, not a single outcome.
• I have decided how I will recognize after a few months that it is not working for me.

The bigger picture

Survivorship bias affects not only success stories but also one's own business. Whoever asks only the customers who stayed does not learn why others left. Whoever surveys only the employees who are still there does not learn what drove the departed away. In both cases the visible group is a selection of the satisfied. The solution is the same: deliberately look for the missing cases, for instance through conversations with lost customers or former employees.

This is particularly important for learning from failure. How entrepreneurs learn and start again after failing is covered in the analysis learning from failure. Whoever looks for a mentor should also bear in mind that the mentor's experience is a survival story too. The article mentoring for business owners describes what matters in such a relationship.

The limits belong here as well. Denrell's work is a theoretical model with simulation, not a measurement of German businesses. The study on role models concerns students and their intentions, not actual start-ups or their success. And not every success story is biased: some strategies really are better. The bias only says that one cannot tell from the stories alone.

Conclusion: why do role models mislead, and how can one still learn from them?

The question was
role models can mislead entrepreneurs and how one can still learn from them sensibly. The answer lies in the selection. Success stories come from survivors, and the failures are missing. According to Denrell, this makes risky strategies in particular look more successful than they are, because they occur frequently among the best and the worst are no longer visible.

Role models remain valuable if one uses them for what they achieve: encouragement and examples of behavior. Decisions about one's own strategy require the question that no success story answers. For this week, it is enough to ask the first question of the Role Model Check at the next inspiring story: who did the same thing and failed?

A success story is evidence that something can work. It is not evidence that it usually works.