Why do smart founders repeat the same mistake?
Founders repeat mistakes because a startup is a poor teacher, not because information is scarce: feedback arrives late and from many causes at once, and reading about a mistake rarely changes what someone does on Tuesday morning.
Kind vs wicked learning environments
Hogarth and colleagues call a setting “kind” when feedback is quick, accurate and tied to the action, and “wicked” when it is missing or misleading.
A founder makes one pricing decision a quarter, sees the result months later, and can credit it to ten causes.
Experience there can teach the wrong lesson with full confidence. That startups are wicked settings is an argument from the research, not a measurement of Indian founders.
Knowing is not doing: The Lomé, Togo trial
In a randomised trial with 1,500 micro-entrepreneurs in Lomé, Togo (500 per group), traditional business training produced a statistically insignificant 11% rise in profits (Campos et al., Science, 2017).
Training in personal initiative, which teaches action habits, raised profits about 30%.
Treat that as a lead, not proof: the study is about micro-businesses, not funded startups.
The narrative loop and the four interruptions
Each step of the loop gets a break. Each break is a rule in the system, not advice.
→ The story confirms prior belief → (back to Step 1: Decide with false certainty)
Write the expectation first (Mistake Ledger)
Write what you expect and by when, before the outcome.
Fix the test before the work (Kill rule)
The kill rule or exit test is written before the work starts.
Ask for evidence, not a feeling (Pre-registered analysis)
The result is read against the pre-registered analysis, not a feeling.
Bring in outside eyes (Independent reviewer)
The evidence reviewer checks every memo before the test and every result after it. On Startup Indus, the weekly circle does the same job.