Decisions outlive the meeting. Keep their reasons.
People make decisions, then move on. The decision stays and its reasons leave with them. A decision log keeps the why beside the what, for the colleague who wasn’t in the room.

Hindsight rewrites memory. A record doesn’t.
Without the reasons, a team can only keep a decision it no longer understands, or reverse it and bring back the problem it solved.
Psychologist Baruch Fischhoff showed that once people know an outcome, they judge it as more predictable than it was. He called it “creeping determinism”.
Journal of Experimental PsychologyMicrosoft’s architecture guidance: a decision never recorded “will likely be forgotten, leading to repeated debates.” Keep the log append-only, with a confidence level for each decision.
Microsoft LearnMichael Nygard proposed the decision record: title, context, decision, status, consequences. “All consequences should be listed here, not just the ‘positive’ ones.”
Michael NygardThoughtworks placed lightweight decision records in the Adopt ring of its Technology Radar, its strongest recommendation.
ThoughtworksOne row per decision. Newest first.
From a clinic, a school, a service centre and a bank branch: the same fields hold any significant decision.
| ID | Title | Decided | Owner | Decision | Expected outcome | Confidence | Review | Status |
|---|---|---|---|---|---|---|---|---|
| D-005 | Book-ahead appointments for account opening | 2026-10-20 | Branch manager | We will offer book-ahead slots for account opening, with walk-in kept. | Shorter waits for quick requests. | Low | 2027-02-28 | Proposed |
| D-004 | Coordinator role adds discharge calls | 2026-06-12 | Operations manager | We will add discharge calls to the coordinator role from July. | Nurses spend less time on calls by December. | Medium | 2026-12-15 | Accepted · supersedes D-001 |
| D-003 | Service counter opens at 7:30 on Mondays | 2026-05-05 | Service centre manager | We will open the counter at 7:30 every Monday from June. | Average Monday wait falls by the end of September. | High | 2026-10-15 | Accepted · review due |
| D-002 | Two shorter assessments replace the end-of-term exam | 2026-04-22 | Deputy head, academic | We will run two shorter assessments per term in grades 7 to 9 from September. | Teachers act on mid-term results within two weeks. | Medium | 2027-01-31 | Accepted |
| D-001 | Follow-up calls move to a coordinator | 2026-02-10 | Operations manager | We will move follow-up calls to a dedicated coordinator from 1 March. | Completion rises within three months. | Medium | 2026-06-01 | Superseded by D-004 |
The XLSX flags Review due when a review date passes with no outcome written. The files add context, options considered, evidence, what remains uncertain and the outcome at review, plus a one-page record sheet.
Including those rejected, and why. The field most often missing and most often needed.
One sentence in the active voice that begins “We will…”.
Link the survey, report or log itself, so later readers can judge whether it still holds.
What should be true if it works, and by when. Written before you know.
High, medium or low, with one line on what remains uncertain.
Proposed, accepted, superseded or retired. Superseded links to its replacement.
Write it the week you decide.
Choose what counts
Decisions that shape how you work, affect quality or safety, or are hard to reverse. Not every purchase.
Record it that week
A record written the same week is accurate. One written months later is a reconstruction.
Name the rejected options
And why each lost. That is what the next team will want to reopen.
State what you expect
The outcome, the date and your confidence. Hindsight will try to erase all three.
Review on the date
Write what actually happened beside what you expected. Weak evidence, changed conditions or bad luck: say which.
Supersede, never overwrite
A new record replaces the old one and links back to it. The history of a change of direction is often the most valuable part.
Why decision logs die.
The log turns into minutes. Nobody reads it, then nobody writes it.
Without the rejected options, the same debate returns in a year.
By then the outcome is known and the record is shaped by it.
The trail disappears. Write a new record and mark the old one superseded.
Then no one ever learns whether the decision did what it promised.
A log beside the plans and procedures it affects gets read. One in a folder does not.
When the log needs to remember on its own.
A spreadsheet holds rows. It can’t show how far each one can be trusted, or what happened next.
- People ask “why did we decide that?” and the answer is in someone’s inbox.
- Decisions, procedures and lessons live in different files that never meet.
- You can’t tell a ratified decision from someone’s guess.
- The assistants your team uses can’t see why you decided anything.
ROADYOM PRISM
Every decision, with its weight.
Prism holds an organisation’s decisions and knowledge as items with a stated trust level: ratified, evidenced, observed or inferred. Each decision links to its evidence, to the steps that carry it out, and to the outcomes and Pause and Learn lessons that follow. When a decision is superseded, the earlier one stays visible. People make the decisions. Prism keeps their reasons.
Coordinator role adds discharge calls
- Decision
- We will add discharge calls to the coordinator role from July.
- Owner
- Operations manager
- Review
- 2026-12-15
Questions, answered.
Is this the same as meeting minutes?
No. Minutes record what was said. A decision record keeps what was decided, why, what else was considered and what you expected. One page or less.
Who should write the record?
The owner of the decision, or someone they name, the same week. The people consulted should be able to read it and correct the facts.
What about decisions we made years ago?
Start with new decisions. Then reconstruct the few past ones people ask about most, and mark them as reconstructed.
Isn’t this a software engineering practice?
It started there, with the architecture decision record. Nothing in it is specific to software. Admission rules, assessment policies and opening hours have context, options and consequences too.
What does Confidence add?
It records what you believed before you knew the result. At review, a confident decision that failed teaches something different from a cautious one.
Should a decision ever be deleted?
No. Retire it or supersede it. The record of a reversed decision is how the next team avoids repeating it.
Sources (5)
- Hindsight ≠ foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance 1(3) (B. Fischhoff), 1975.
- Documenting Architecture Decisions. Michael Nygard, Relevance (now Cognitect) blog, 2011.
- Architecture decision record (Azure Well-Architected Framework). Microsoft Learn, 2024.
- Decision Log (Code-With Engineering Playbook). Microsoft, 2024.
- Lightweight Architecture Decision Records (Technology Radar). Thoughtworks, 2018.
Decide on purpose. Twice.
Once when you choose. Again when you review. Free, in English and Arabic.