A useful decision review asks two related questions: what happened, and was the choice reasonable given the information available then? Combining them into a single verdict can hide the lesson.
Start with an original entry
Imagine that a team chooses a two-week pilot of a new reporting workflow. The old process is slow, a small test is affordable and the team has a workable fallback. Their original prediction is that at least four of five pilot participants will submit complete reports by the second Friday.
They record two uncertainties: whether the instructions are clear and whether one data source will be available. They plan to review after the pilot and stop early if the fallback becomes necessary.
Separate observations from explanations
At review, three participants submitted complete reports. One misunderstood a field, and another could not access a required source after permissions changed. These are observations to investigate. “The pilot was a bad idea” is a broader interpretation.
| Question | Example response |
|---|---|
| What happened? | Three of five complete reports; one instruction problem; one access problem. |
| What was knowable? | The team could have tested the instructions with a new user. The later permission change was not announced before launch. |
| What was reasonable? | A limited pilot and a fallback kept the exposure manageable. |
| What should change? | Test the confusing field before expanding and check access again before the next cycle. |
Look for the useful correction
The missed target matters. It does not, by itself, establish that the decision to pilot was wrong. The pilot may have revealed exactly the problems a small test was meant to find. At the same time, an available instruction check may have been worth doing earlier.
Good review preserves both observations: the controlled test was proportionate, and one preparation step should improve. It avoids excusing every problem as luck or treating every failure as proof of poor judgment.
Append the review rather than replacing the entry
Keep the original prediction and reasons. Add what changed, what evidence emerged and what you will do differently. If an earlier statement needs a correction, label the correction with its date instead of silently editing the record.
Use repeated reviews to find patterns
A single case offers a possible lesson. Across several comparable decisions, look for recurring omissions, overly optimistic timelines or assumptions that turn out to be dependable. Be cautious with small samples and choices that differ substantially.
When you recorded a probability, assess it across many clearly defined outcomes. One event occurring does not prove a 70% forecast was accurate; one event failing to occur does not prove it was wrong.
For the underlying distinction, read Outcome Bias at Intuition Management. Explore a worked comparison for the stage before choosing.