Concept
The decision review
A decision review, held on a date set when the decision was made, records the outcome first and then, separately, judges the process given what was known at the time: which error got through, how well calibrated the probability was, and what to change. The lesson is about the process, never about luck.
Reviews go wrong in two ways: they never happen, because nothing prompts them, or they happen as resulting, judging the decision by how it turned out. A review date set when you decide (a tripwire) fixes the first. Annie Duke's four boxes, which cross a good or bad decision with a good or bad outcome, fix the second. The order matters. First the outcome in plain terms. Then the process: given what I knew, were the options wide enough, was the assumption tested, was the probability honest, did the premortem see this coming. Then the lesson, which must be about something you could have done differently with the information you had. A good decision with a bad outcome produces no change to the process; a bad decision with a good outcome must, even though nothing hurt.
A real-life example
Nine months after opening, a second bakery branch is breaking even but the first shop's reviews have slipped, one of the risks the owners named in a premortem before opening. The review changes staffing, not the decision to open, and the tripwire did its job.
How to use it
- 1Review on the date you set when you decided, not when you feel like it.
- 2Write the outcome first, then judge the process separately, then the lesson.
- 3Change the process only for decisions that were poor given what you knew.