Decision Log Builder

28 views · 0 copies · 0 likes

Almost nobody records why they decided something, and the cost of that shows up later in two specific ways. First, decisions get relitigated - a question settled in March returns in July because nobody remembers the reasoning, and the same three people have the same argument. Second, and more damaging, you never learn whether your judgement is any good, because you cannot compare an outcome against a prediction you never wrote down.

The second problem is the serious one. Without a written prediction, hindsight quietly rewrites what you expected. Outcomes feel more foreseeable than they were, and a decision that went badly gets remembered as obviously risky while a lucky one gets remembered as skilful. You accumulate experience without accumulating judgement, which is why years in a role and quality of decisions correlate far less than people assume.

This prompt captures a decision in a structured form designed for later review. The critical field is the falsifiable prediction with a date. Not this should improve retention, but retention for this cohort will be above twenty percent at ninety days, checked on the fifteenth. A prediction you cannot be wrong about teaches you nothing.

The reversibility assessment does real work too. Decisions that can be undone cheaply deserve speed, not deliberation, and treating them as weighty is a common and expensive habit. Genuinely irreversible decisions deserve the opposite. Recording which kind you thought this was is often more revealing in review than the decision itself.

The field people skip and should not is what would have to be true for this to be wrong. It forces you to name your assumptions while you can still see them, and it converts a vague unease into a specific thing to watch. Very often the honest answer surfaces a check you can run this week for almost nothing.

The rejected alternatives matter for a reason people underestimate. When a decision is revisited later, the question is almost always why did we not just do X - and X was usually considered and rejected for a reason that no longer exists or still holds. Either answer resolves the discussion in a minute.

How to use it. One entry takes about ten minutes and is worth it only for decisions that are expensive to reverse or that you expect to be questioned. Logging everything guarantees the habit collapses. Set a calendar reminder for the review date at the moment you write the entry, because a log nobody revisits is just a diary.

You capture a decision in a structured, reviewable form. Ask for anything missing before producing the entry. Do not fill gaps with assumptions.

INPUTS
The decision: [what is being decided]
Context: [why this is being decided now]
Options considered: [list them]
Who decides: [name or role]

Produce a DECISION LOG ENTRY with exactly these fields.

DATE - today.
DECISION - one sentence stating what was chosen. Written so a reader in a year understands it without context.
STATUS - PROPOSED, DECIDED, or SUPERSEDED.
DRIVER - what forced a decision now rather than later. If nothing did, say so, because that is worth noticing.
OPTIONS CONSIDERED - each option with the reason it was rejected. Be specific. Rejected as too expensive is not useful in a year - state the figure or the constraint.
REVERSIBILITY - classify as EASILY REVERSED, COSTLY TO REVERSE, or EFFECTIVELY PERMANENT, and state the actual cost and time to undo it.
ASSUMPTIONS - what must be true for this to be correct. For each, mark whether it is VERIFIED, ASSUMED, or UNKNOWN.
FALSIFIABLE PREDICTION - a specific measurable outcome with a date. It must be possible to be wrong. Reject any prediction that cannot fail and rewrite it.
WHAT WOULD MAKE THIS WRONG - the observable signals that would indicate this was a mistake, and roughly when they would appear.
CHEAP CHECKS AVAILABLE NOW - anything that could test a key assumption this week at low cost.
REVIEW DATE - when to compare outcome against prediction.

After the entry, output REVIEW QUESTIONS - the three questions to ask on the review date. These must be answerable with evidence, not opinion.

RULES
If the prediction cannot be falsified, say so and rewrite it until it can.
If no option was seriously considered besides the chosen one, flag that explicitly as a risk.
Do not evaluate whether the decision is good. Your job is to record it so it can be evaluated later.
Do not soften an assumption marked UNKNOWN into one that sounds verified.