How to keep a decision journal

Cerno is a private AI boardroom that saves every decision with its brief, its assumptions and its vote, and asks you later what happened. This guide is about the practice itself — the format, what to record, and what it shows you — whether you keep it in Cerno, a notebook or a spreadsheet.

Why keep one

You make dozens of consequential decisions a year and remember the reasoning for almost none of them. What you remember is the outcome, and outcomes are a terrible teacher: a good result from a bad decision feels like vindication, and a bad result from a good decision feels like a lesson. Without a record of what you thought at the time, you learn the wrong things from both.

A journal separates the decision from the outcome. It lets you ask, a year on, not "did it work?" but "given what I knew, was that a good call?" — and, across many entries, "what do I keep getting wrong?"

That second question is the one with money in it. Most people have two or three systematic errors — a particular optimism, a particular avoidance, a particular way of caving — that cost them repeatedly. They can't see them because each instance looks like a one-off. The journal is where the one-offs turn into a pattern.

The format

Five things, in this order, at the moment of deciding. Not after.

1. ›The question. As a question. "Should I take the client at the lower day rate?" — not "client pricing".
2. ›The options. All the ones you seriously considered, including the one you didn't take. If there was only one, write down why.
3. ›What you chose, and why. Two or three sentences. The "why" is the part you'll be most surprised by later.
4. ›What you expect to happen. Concrete and dated where possible. "They'll become a regular client at the lower rate and I'll regret it by spring." Vague expectations can't be wrong, which is why people write them.
5. ›What you were feeling. One line. Tired, excited, cornered, relieved. This is the entry that explains the others when you read them back.

Then, at a review date you set when you write the entry:

1. ›What happened. Honestly, briefly.
2. ›Was it the decision or the dice? A good outcome from a lucky break is not a good decision. A bad outcome from a decision that was right on the information is not a bad one. Separate them, or the journal teaches you to be lucky.

What to log

Not everything. The filter is: would you want to be able to explain this in a year? Practically, that means decisions that cost money you'd notice, commit you for more than a month, or land on someone else — a hire, a price, a client, a lease, a supplier, a product.

Two or three entries a month is normal for a small business. If it's more than that, you're logging things that don't need it and you'll stop. If it's fewer, you're probably not counting decisions you made by default — the ones where you did the usual thing without deciding — and those are worth an entry precisely because they didn't feel like decisions.

Reading it back

The value is entirely in the reading, and almost nobody does it. Set two habits.

The review date. Every entry gets one when it's written — a month, a quarter, a year, depending on the decision. When it arrives, fill in what happened. This is the part that feels pointless and isn't; an entry without an outcome is a prediction you never scored.

The annual read. Once a year, read the whole thing in one sitting. Not entry by entry — the whole thing, looking for what repeats. This is where the journal pays for itself, and it takes an hour.

What people typically find:

• ›A stable optimism. Timelines underestimated by about the same fraction every time. Once you know the fraction, you know your base rate for your own estimates, and you can correct for it in advance.
• ›A recurring avoidance. The same kind of decision, always deferred. Usually a conversation nobody wanted to have — the price rise, the underperforming supplier, the client who should have been let go.
• ›A tell. A feeling that predicts a bad call. For some people it's "excited"; for others it's "cornered". The feeling column is where this shows up, and once you know your tell you can treat it as a warning rather than a mood.
• ›A door you keep misjudging. Decisions treated as reversible that weren't, or the reverse — weeks spent on something you could have undone in a day.

The mistakes that stop it working

Writing the entry after the outcome. The entry has to be written before you know. Reconstructed reasoning is fiction with a happy ending. If you missed the moment, log the decision with a note that the expectations are being written after the fact — and weight it accordingly.

Vague expectations. "I think this will go well" can't be scored. "I expect this client to reorder within two months" can. The discomfort of a specific prediction is the point.

Never scoring it. An entry with no outcome recorded has taught you nothing. The review date is not optional; it's the mechanism.

Grading outcomes instead of decisions. A journal that just records wins and losses teaches you to be lucky. The seventh question — decision or dice? — is the one that makes it a learning tool rather than a scoreboard.

Keeping it somewhere you won't look. A notebook in a drawer, a file in a folder. The journal has to be where you'll trip over it, or the annual read never happens.

Where Cerno's journal fits

Cerno's Decision Journal is this practice with the writing done for you: every decision is saved with its brief, its recorded vote, its assumptions and its dissent — the "why" captured at the moment of deciding, in more detail than you'd write by hand — and the journal asks you later what you actually did and what happened. After a year, the pattern reads are there to be run. It's the version for someone who knows they should keep the journal and knows they won't fill in the sixth column by hand.