When an experiment ends, it is easy to remember thinking the outcome was obvious. Put your expectation in writing before you send the email, launch the feature or run the test.
1. Make a forecast you can check
For [decision], write one observable outcome, a deadline, my current expectation and why. Include the cost, the smallest useful signal, and what would make the result inconclusive. Do not make up a probability if I haven't provided one.
2. Record the outcome without rewriting
Using [original forecast and observed evidence], add an outcome row without editing the original. Separate what happened, how it was measured, what was outside our control, and what is still unknown.
3. Change one decision
Review [ledger rows]. Find a repeated source of overconfidence or underconfidence only if supported. Suggest one change to the next experiment. Show the exact past row that makes the change worthwhile. Keep small samples in perspective.
Try it on one example
Example: before sending five tailored pitches, predict how many qualified conversations will happen by Friday. Record the actual replies and separately note how many people had time to respond. The goal is to improve your next choice.
Inspired by exploring Midas’s prediction-oriented approach at https://midas-trader.com/pricing. This workflow applies to business decisions, not trading.