Team Operations
Decision Journals: Record the Reasoning Behind Important Choices
Keep a useful record of what you knew and expected before a decision, then review the reasoning separately from the eventual outcome.
What you will learn
- Record evidence and assumptions before knowing the outcome
- Review a decision process separately from whether the result was favorable
- Use differences between expectations and observations to improve the next choice
A decision journal records the reasoning behind a meaningful choice before its outcome is known. It helps you compare what you expected with what later happened without quietly rewriting the original story.
Use it for decisions worth learning from, not every small task. A journal is most useful when there are credible alternatives, material uncertainty and a future point at which you can inspect the result.
What is a decision journal?
Farnam Street's decision-journal guide describes writing down your thinking before memory has a chance to reshape it. The practical value is a contemporaneous record: what was known, what was assumed and why a particular option was chosen.
It is not a justification document written after success, nor a mechanism for blaming someone after an uncertain choice turns out badly. A reasonable process can produce an unfavorable outcome. A poor process can occasionally get lucky.
Capture the choice in a compact record
This original fictional example concerns whether a small service team should test a second delivery partner. It is not a recommendation about a real supplier.
| Record field | Fictional partner-test decision |
|---|---|
| Decision and owner | Operations lead decides whether to run a limited trial with a second partner |
| Known facts | The current partner missed two documented handoff dates; causes remain under review |
| Alternatives | Improve the current handoff, run a small second-partner trial or postpone new commitments |
| Key uncertainty | Whether delays reflect partner capacity or incomplete internal inputs |
| Chosen action | Run one bounded trial using the same approved input checklist |
| Expected observation | Compare handoff completeness and reviewed output; do not infer long-term reliability from one job |
| Revisit trigger | Review after the trial or earlier if an access or quality issue appears |
Link to evidence available at the time. Separate a fact from an interpretation: two missed dates are documented; “the partner is unreliable” is a broader conclusion that may not yet be justified.
Keep the original record intact
Once the decision is made, preserve its version and date. Add later observations in a new review section rather than editing the old assumptions to match what you now know.
Record material changes that occur during execution. If the trial receives a different input checklist, the comparison no longer tests exactly what was planned. That does not make the exercise worthless, but it changes what you can conclude.
Keep sensitive information limited to appropriate readers. The journal should support learning, not become an uncontrolled collection of personal opinions about teammates or customers.
Separate process quality from outcome
Use this original four-case exercise when reviewing a choice. The labels describe the review question, not a score for the person involved.
- Careful process, favorable result: the limited trial used the agreed inputs and produced a useful output. Continue learning; one success does not establish a universal rule.
- Careful process, unfavorable result: an unexpected outage interrupted an otherwise appropriate test. Review resilience without pretending the outage was predictable.
- Weak process, favorable result: the team skipped the access review but encountered no incident. The favorable result does not make the skipped safeguard sensible.
- Weak process, unfavorable result: unclear requirements caused rework. Improve the requirement check rather than merely avoiding that partner forever.
Ask which evidence would have changed the choice at the time. Then ask whether that evidence was reasonably available. This keeps the review from demanding impossible foresight.
Turn the review into a specific learning
Prefer a bounded revision such as “confirm the input owner before a trial” to “communicate better.” Identify where the revised check belongs and who owns it. Avoid creating a heavy approval process for every future decision because one test failed.
The customer survey bias lesson helps examine another source of overconfidence: treating the evidence you collect as more representative than it is. The project management guide provides broader context for turning a reviewed choice into coordinated work.
Copyable AI prompt
Try this with Aurora AI in Finta, Codex, or Claude. Share only authorized decision material.
Help me document [decision] using [evidence available now], [alternatives] and [constraints]. Separate facts, assumptions, uncertainties, decision ownership, chosen action and a future review trigger. Do not invent probabilities or evidence. For a later review, compare the preserved original record with [observations] and distinguish process quality from outcome luck. Suggest one bounded learning. Do not rewrite the original rationale, approve spending or execute the decision.Preserve the reasoning while it is fresh
Explore Finta's Decision Journal Keeper for organizing the record. The important habit is simple: capture the reasoning before the result makes it seem obvious.