Teams often remember that a decision was made but disagree about what it included or why. Months later, a reasonable choice can look arbitrary because the constraint, rejected options, and owner are gone. People reopen settled questions or quietly follow different interpretations.
A decision log preserves the minimum context needed for consistent action. It is not a transcript and should not become a justification archive. A useful entry states the decision, owner, date, reason, important alternatives, consequences, and conditions that should trigger review.
Record decisions that change action
Log choices that set direction, allocate meaningful resources, accept risk, change scope, or establish a rule others must follow. Do not record every routine preference. The test is whether someone could make the wrong move later because the choice or its boundary was forgotten.
Write the decision as a complete sentence with a verb: We will pilot the new intake process with the service team for four weeks. Avoid topic labels such as intake process decision. A reader should understand the chosen action without reconstructing the meeting.
Preserve the reason without writing a transcript
Capture the two or three factors that carried the decision: customer impact, safety, cost, reversibility, available evidence, or a deadline. Mention significant options that were considered and why they were not chosen. This makes future review fair because it evaluates the choice against conditions known at the time.
Separate evidence from opinion. Link to the relevant estimate, test, or policy when available. If the choice relies on a judgment under uncertainty, say so. A transparent assumption is not weakness; it tells future readers what new evidence could legitimately reopen the decision.
Name ownership and consequences
Identify the decision owner, not just the meeting attendees. Then list immediate actions, people affected, and any commitment created. A decision with no execution owner is only a discussion outcome. The log should connect direction to the first visible movement.
Record dissent when it involves a material unresolved risk, not every preference. State the concern neutrally and who accepted it. This protects candor without undermining the final direction. Once recorded, everyone should execute the decision unless new evidence triggers the agreed review path.
Add an expiration or review trigger
Some decisions remain valid indefinitely; many depend on conditions. Add a review date or trigger such as volume exceeding a threshold, a regulation changing, or the pilot producing results. This prevents temporary choices from becoming permanent simply because nobody remembers to reconsider them.
During review, append rather than rewrite history. Mark the decision affirmed, modified, or replaced and link the new entry. An auditable chain shows learning over time and prevents revised context from making the original choice appear different from what it was. Periodically archive trivial entries and repair broken evidence links so the log remains searchable, credible, and useful to people who were not present.
Write a seven-field decision entry
- Choose a recent decision that changed scope, resources, risk, or a rule, and state the selected action in one complete sentence.
- Add the owner, date, two decisive factors, key rejected alternative, and a link to the strongest supporting evidence.
- List the immediate consequence and first action owner, plus any material dissent or accepted risk in neutral language.
- Set a review date or observable trigger, store the entry where affected people work, and share the direct link.
Common questions
Who should write the decision entry?
Anyone can draft it, but the decision owner should confirm accuracy. The entry should be published quickly enough that misunderstandings can be corrected before execution diverges.
Can a decision log make teams too bureaucratic?
Yes, if every preference is logged or entries become essays. Limit it to consequential choices and use a compact template that takes only a few minutes.