Judgment Playbook

Decision journal

Record the call, reasoning, assumptions, prediction, and review date before outcome rewrites memory.

5 minute readtaught in course chapter 10
CommitmentUncertainty

What it is

A decision journal is a dated written record of a decision, made before the outcome is known. One entry holds the choice, the alternatives you rejected, your reasoning, what you expect and how sure you are, the assumptions the choice rests on, and the date you will reopen it. Its purpose is to let a decision be judged later on the quality of its reasoning rather than on the luck of its result. It is not a diary. It is a dated, testable record of a call.

Where it comes from

The term has no single author. The failure it addresses was documented by Baruch Fischhoff in 1975, in "Hindsight ≠ Foresight" in the Journal of Experimental Psychology: Human Perception and Performance. Telling people how something turned out raised their sense of how predictable it had been, and they did not notice the change. Fischhoff argued that this lack of awareness "can seriously restrict one's ability to judge or learn from the past."

The prescription reached management writing in 1989. J. Edward Russo and Paul J. H. Schoemaker put "Not Keeping Track" ninth among the ten traps in Decision Traps: trusting experience to hand over its own lessons, and so keeping no systematic record of what past decisions produced.

The name "decision journal" spread later, mostly through Shane Parrish's Farnam Street, whose template asks for the frame, the alternatives, the expected outcome with probabilities, and the writer's physical and mental state at the time. A much-repeated story has Daniel Kahneman telling an investor to buy a cheap notebook and write down what he expected and why. That remark circulates second-hand, through Michael Mauboussin, rather than through Kahneman's published work; treat it as folklore rather than as a citation.

What it corrects

Take a decision made under real uncertainty, whose result arrives months or years later, shaped along the way by events nobody controlled. When the result lands, the reasoning that produced it is gone. Memory offers a replacement, edited by the outcome. After a success the deciding argument looks obvious and the doubts have disappeared. After a failure the warning signs look as though they sat in plain view.

Ordinary care does not fix this, because the person remembering cannot feel the edit happen. That is Fischhoff's finding: subjects told an outcome still believed they were reporting what they had thought beforehand. Honesty, experience and seniority are no defence against an error you cannot detect from the inside. Only a record written before the outcome, and left alone afterwards, holds the earlier belief in place.

How it works

  1. Write the decision in one sentence, with today's date: what you are choosing, and what you are giving up.
  2. List the live alternatives and why you rejected each one.
  3. State the prediction. What you expect to observe, by when, and your confidence as a number.
  4. Name the two or three assumptions the choice depends on, and mark the one you are least sure of.
  5. Set the review: a date, or an observable event that pulls the entry forward.
  6. At review, read your entry before you look at the outcome. Then score the reasoning and the result separately, and leave the entry unedited.

Worked example

Philip Tetlock's forecasting research is a decision journal run at scale. Over roughly two decades he recruited 284 people whose jobs involved commenting or advising on political and economic trends, and asked them for dated probability estimates on questions inside and outside their own fields. By 2003 the file held 82,361 forecasts. He also recorded how each forecaster explained a miss, which made the excuse checkable against the original number. Scored and published as Expert Political Judgment in 2005, the record showed the experts barely outperforming informed non-experts, and neither group beating simple statistical rules. A reputation assembled from remembered hits would never have produced that verdict. The dated file did, because nothing in it could be revised after the world answered.

In a Business Case Weekly case

In the Nike case, the fork sits in June 2020, when the board is asked to approve Consumer Direct Acceleration: cut wholesale accounts, sell direct, and accept margin arithmetic that makes the move look obvious. A journal entry does not settle the vote. It fixes in writing which claim about consumer demand the plan depends on, what the director expects Nike's position on other retailers' shelves to look like two years out, and which quarterly number, measured outside Nike's own channels, would say the claim had failed. Written that way, what happens by 2024 is evidence about the reasoning instead of a story told about it.

In your answer

  • "I decide X today, against alternative Y, which I reject because …"
  • "This rests on the assumption that …, and I am least confident about …"
  • "By I expect to see …, and I hold that at roughly % confidence."
  • "I will reopen this call when …, and reverse it if …"

Common misuse

The version that does no work is the diary: a full account of what was decided and why it felt right, carrying no prediction and no review date. It reads well and teaches nothing, because it leaves the outcome nothing to contradict. The second failure is grading by result. The entry is reopened, the outcome was good, the entry is marked correct, and a lucky call is filed as a sound one. A decision can be well reasoned and unlucky, and keeping those two verdicts apart is the point.

The test: point at the sentence in the entry that the outcome could have proved false, and at the date on which someone was going to check. If either is missing, you have a diary.

References

Kind vs wicked learning environments

Next method

Distinguish feedback that teaches from feedback that rewards luck, delay, or hindsight.

Open next tool