Decision Records: How to Stop Re-deciding Things

Count the meetings last quarter where you discussed something you had already settled. Not revisited deliberately. Discussed again because nobody could find what was decided, or could find it but not the reasoning.

For most teams the number is between three and ten. Each one costs an hour of several people's time and ends, usually, in the same conclusion.

That is what a decision record prevents. It is not documentation. It is the thing that stops you paying twice.

The format

Short enough that people actually write it. Five fields.

Decision. We are moving invoicing to the first of the month rather than on completion.

Date and who decided. 14 March, Ana (Bookkeeping role).

Why. Cash arrives in a predictable window instead of scattered through the month, which is what has been causing the gap in weeks three and four.

What we decided against, and why. Weekly invoicing: more admin than the cash-flow gain justifies at our size. Deposits on everything: would cost us the two retainer clients who have said no to that before.

What would make us revisit. If more than a fifth of clients push back, or if the cash gap persists past June.

Two paragraphs. Ten minutes. Written by whoever made the decision, not by a scribe.

The field everybody leaves out

What we decided against, and why.

Nearly every team records the conclusion and drops the alternatives. That single omission is what makes the record useless six months later, because the question that comes back is never "what did we decide". It is "why aren't we doing the obvious thing?" — and the obvious thing is almost always one of the alternatives you already rejected for a reason you no longer remember.

Without it, a new person suggests weekly invoicing in month two. Nobody can remember why not. You discuss it for an hour and land where you already were. That is the meeting you are trying to stop having.

The second most-dropped field is what would make us revisit. Without it every decision is either permanent or reopened at random, and both are bad. A stated trigger turns "we already decided that" from a door slammed into a condition somebody can check.

What deserves one

Not everything. A record for every decision is how the practice dies in a month.

Write one when:

• It will be questioned later by someone who was not there
• It cost real money or changed direction
• You deliberately rejected an obvious alternative
• It creates an exception to a normal rule
• Reversing it would be expensive

Do not write one when: it is inside someone's role and nobody will ask again, or it is reversible in an afternoon.

The test: would a new person, six months from now, look at this and think we had made a mistake? If yes, write the record. That is exactly the case where the reasoning is the valuable part.

Where they live

Anywhere findable and searchable. The format matters more than the tool.

Two rules that matter more than where:

One place, not per-project. Decisions cross projects, and the person searching does not know which project it belonged to.

Searchable by the words someone would use. Not "Q1 Finance Review 14 March" but "invoicing timing". People search for the subject, never for the meeting.

Where it goes wrong

Writing them as minutes. Minutes record what was said. A decision record records what was concluded and why. Minutes are much longer and much less useful; a team that has minutes usually believes it has decision records and does not.

A backlog project. Nobody is going to reconstruct two years of decisions, and the attempt kills the habit before it starts. Begin with the next decision. The record becomes valuable at about thirty entries, which takes a few months.

Assigning it to one person. The decider writes it. A scribe cannot write the reasoning because they were not the one weighing it.

Never reading them. If nobody ever cites one, they will stop being written. Make citing them normal: "we looked at this in March, here is the record, has the trigger changed?"

Perfect prose. These are working notes. Bullet points and a plain sentence beat a polished paragraph that never gets written.

What good looks like

Somebody says "haven't we been here before?", finds the record in thirty seconds, reads the trigger, and the group either says "conditions changed, let us reopen it" or "no, still holds" — in four minutes rather than an hour.

A new person reads six months of records in an afternoon and understands the organization better than three weeks of shadowing would have taught them.

Try this week

Take the next decision your group makes and write the five fields. Especially the fourth.

Then find the last thing you re-decided, and write its record retroactively. You will discover whether anyone can still remember the alternatives — and how close you came to losing the reasoning entirely.