Why founders should keep a decision log (and how to keep one you'll use)

Here's a pattern every operator recognises: a question gets debated, resolved, and quietly reopened four months later — because the reasoning left with the conversation. You end up paying for the same decision twice, sometimes reversing it without ever learning whether the first answer was right.

A decision log fixes that for roughly two minutes of writing per decision. It's the cheapest institutional memory available, and it works whether you're solo or running a team of thirty.

The five fields that matter

Longer templates get abandoned. These five are the load-bearing ones.

• Decision — one sentence, in the past tense, stating what was chosen.
• Date — when it was made, not when it was written up.
• Reasoning — the two or three things that actually tipped it, including the constraint you were under.
• Alternatives rejected — what you didn't do, so future-you doesn't rediscover it as a bright idea.
• Revisit by — a date, or 'when X changes'. Most decisions deserve an expiry.

Skip owner fields, confidence scores, and RACI grids until you've kept the log alive for three months. Complexity is how logs die.

What's worth logging

Not everything. Log a decision when reversing it would be expensive, when you'd struggle to explain it in six months, or when someone else will need to act on it. Pricing changes, positioning, hiring, tooling commitments, saying no to a market segment — those belong in the log. Which font to use does not.

If reversing it costs more than an afternoon, write down why you chose it.

The review cadence

A log with no review is an archive. Two habits keep it live:

1. 1Weekly, log anything you decided in the last seven days — usually zero to three entries.
2. 2Monthly, open the decisions whose revisit-by date has passed. Confirm, adjust, or reverse, then log that as a new entry rather than editing history.

That second habit is where the compounding happens. Reviewing your own past reasoning against what actually occurred is the fastest available feedback loop on your judgement — far faster than waiting for outcomes to be undeniable.

When it starts paying off

Within a quarter you'll notice fewer circular conversations. Within two you'll have calibration: a real sense of which kinds of calls you tend to get right and which you consistently overestimate. And the first time you onboard someone, the log does an hour of explaining for you.

EmpireOS ships a decision log as a core record type — with reasoning, alternatives, review dates, and reminders when a decision comes due — so it sits beside your offers, pipeline, and action queue instead of in a document you forget.