What should a decision log include?

Why keep a decision log at all?

Because memory edits past decisions to match how they turned out. If a call worked, you remember seeing it coming. If it failed, you remember having doubts all along. Both versions feel like memory, and neither is a record.

Psychologists call this hindsight bias. In Baruch Fischhoff’s 1975 experiments, people who were told how an event turned out judged that outcome more likely, and were “largely unaware of the effect that outcome knowledge had on their perceptions.” The paper concludes that “this lack of awareness can seriously restrict one’s ability to judge or learn from the past.”

For a manager, the lesson memory teaches is always the flattering one, so confidence grows while accuracy stays flat. A note written before the outcome is evidence. A note written after is already contaminated.

Which fields does a decision log need?

Seven fields when you decide, four when you revisit. The test for every field: would it help you grade the call in three months? If not, cut it. The “what” is the least useful part, because you will remember it. The “why” is what evaporates.

What does a filled-in entry look like?

Here is a generic example. Swap the brackets for your own details. Writing it takes about two minutes; the revisit takes about three.

Which decisions belong in the log?

Only the ones worth grading: calls where you committed to something, the stakes were real, and the outcome will tell you whether your reasoning held. Skip the routine, reversible calls you make twenty times a day. If you are unsure which is which, the undo test for reversible and irreversible decisions sorts them quickly.

Keep each entry to about two minutes. A log usually dies by being too heavy: twelve fields and a scoring rubric, filled in beautifully for two weeks, then abandoned. The opposite failure is an entry like “Picked vendor B,” which has nothing to grade.

Forgot to log one? Write it anyway and mark it “RETRO.” The options and reasoning are often recoverable, but leave retro entries out when you check your confidence, because hindsight has already been at work. If more than a third of your entries are retro, move the log to wherever you are when you decide, such as a phone note or the calendar invite for the meeting.

How do you review a decision log later?

Twice. Each entry gets its own revisit on the date you set, then you close the file. Every quarter, take thirty minutes alone with the entries whose revisit dates have passed and ask six questions:

1. Which calls landed close to what I expected, and what do they have in common?
2. Which missed, and in which direction? Underestimated time, trusted a polished analysis, decided while rushed?
3. What pattern shows up across the misses? One bad call is noise; three of the same shape is a pattern.
4. Did calls marked “sure” land more often than “coin flip” calls? If not, give your certainty less weight.
5. Which decisions recurred? Each cluster is a candidate for a short template.
6. What is the one thing I’ll change? One, not ten.

Why separate a bad decision from a bad outcome?

The “sound given what I knew?” line is the most important one in the revisit. A sound call can lose and a careless one can get lucky. Grade only by outcome and you will learn to stop deciding after a sound loss, and to stay careless after a lucky win. In five studies by Baron and Hershey (1988), people rated a decision maker’s thinking as better when the outcome was favorable than when it was unfavorable, even though they knew they had all the information the decision maker had. More on this in how to decide when you can’t be sure.

Write the revisit in a flat tone, like a short report on your own process. If it reads like a confession, you have missed the point.

Read together, entries show when you decide badly, such as at the end of a long day or on someone else’s urgent timeline. They also answer “how long do projects like this usually take us?” in your own hand. That is the outside view described by Kahneman and Lovallo: start from how similar cases went, not from the details of your plan.

Should you share your decision log?

Keep the raw log private. If someone else will read it, you will shade the entries: cleaner reasoning, softer expectations, no embarrassing assumptions. Then the log lies in the same direction your memory does. Share what it produces instead, such as a hiring template or a lesson like “estimates that depend on a client’s staff run long.” A shared team record of group decisions is useful too, but keep it separate; its job is to stop the team relitigating settled questions.