x
It's a running note of the decisions you make, why you made them, and what you expected to happen. Five minutes, at the point of deciding. Not a diary — a record of reasoning.
You don't forget what you decided. You forget why. Six months later you're looking at a supplier you don't rate, a pricing tier nobody buys, or a tool you're still paying for, and the honest answer to "why did we do this?" is that you can't remember. So you either leave it alone because changing it feels risky, or you change it and rediscover the original reason the hard way.
That gap — between the decision and the reasoning behind it — is where solo businesses quietly accumulate drag.
Four lines. Resist adding a fifth; the format survives because it's small.
The decision. One sentence, plainly. "Moving the client onboarding call from 60 to 30 minutes."
The reason. Why now, and what you're optimising for. "Onboarding is eating Fridays and the second half is mostly repetition."
What you expected. This is the line that does the work. "Expect no drop in project clarity; expect ~2 hours a week back." A decision without a prediction can't be evaluated later — you'll simply rationalise whatever happened.
What would make you reverse it. "If two clients in a row need a follow-up call to cover what we cut." Naming the reversal condition in advance is what stops a decision becoming permanent by default.
That's the whole format. It takes longer to explain than to use.
Directly, and it's the reason this sits in a week about trust boundaries.
Every automation is a decision about what you're willing to stop watching. Six months on, you'll have a scheduled job running, an autoresponder answering, an agent filing — and no memory of what you decided it should never do. That's the failure mode behind the escalation rule: the boundary was clear when you set it and invisible by the time it matters.
The decision log is where that boundary stays legible. "Automated the FAQ replies. Expected: 3 hours a week back, no drop in enquiry quality. Reverse if: a customer gets a wrong answer about price or timing." Now the automation has a written owner's manual, and the handover pack has something real to collect.
Three things, and only the first is obvious.
You stop relitigating. Decisions you've already made stop consuming energy, because the reasoning is retrievable instead of reconstructable.
You find out whether you're any good at this. Here's the checkable part: a decision without a recorded prediction cannot be scored. Read back a quarter of entries and compare what you expected against what happened, and you get something most solo operators never obtain — a calibration score on your own judgement. Most people discover they're reliably optimistic about time and reliably pessimistic about how much customers notice change. Knowing your specific bias is worth more than any productivity technique.
It makes AI genuinely useful on strategy. Paste a quarter of decisions into a model and ask what pattern it sees. It isn't guessing about your business any more — it has the actual reasoning trail. Ask it to flag which decisions had no reversal condition, or which predictions you never checked. That's a real analysis on real inputs, and it's the difference between advice and a horoscope.
Whenever you'd have to explain the choice to someone. Pricing, tools, scope, hiring, what to automate, what to stop doing. Not every small choice — the test is whether future-you would ask "why did we do that?"
Keep it in one file, newest at the top, wherever your operational documents live. Not an app. The value is in it being trivially open-able, which is also why the twenty-minute Saturday reset is the natural moment to add anything you missed in the week.
Write the entry for the most recent decision you made — from memory, right now. It'll be harder than you expect, and that difficulty is the argument.
For the prompt set that runs the quarterly read-back, the free 5-min read is here: The 10-Hour Audit.
Related Decodes
TL;DR: A decision log records what you decided, why, what you expected, and what would make you reverse it — four lines, five minutes, at the point of deciding. You don't forget decisions; you forget the reasoning, which is why old choices become untouchable. The expectation line is the one that matters: a decision with no recorded prediction can't be scored later, so you never learn whether your judgement is calibrated.
