A decision journal template

This is the template I use for consequential decisions. It works on paper, in any notes app, or in a spreadsheet; the fields matter, the medium doesn't. If you want the reasoning behind each field, the companion essay walks through it.

The one rule that makes the whole thing work: fill in the first section before you know how it turns out. A decision journal written after the fact is a diary of justifications.

Part 1 — When you decide

Date:
The decision (one sentence):

What I predict will happen:
  (Specific enough that a stranger could later say "right" or "wrong.")

How confident I am: ____%
  (A number, not a word. "Probably" can mean almost anything.)

Why — the actual reasoning:
  (2–5 sentences. What you know, what you're assuming, what you weighed.)

What I'm most uncertain about:

What would change my mind:

State of mind right now:
  (Tired? Rushed? Excited? It matters more than we like to admit.)

Review date:
  (When will you know enough to judge this? Put it in your calendar.)

Part 2 — While it's open (repeat as needed)

Decisions don't sit still between the call and the outcome. When something new arrives, log it:

Date:
What's new:
Does it point for or against my prediction:
My confidence now: ____%

Two or three of these per decision is plenty. The point is a paper trail of when your thinking changed and what changed it.

Part 3 — When it resolves

Date:
What actually happened:

Was my prediction correct?  yes / partly / no

How satisfied am I with the outcome?  1–5
  (Separate question from correctness. You can be right and unhappy,
   or wrong and relieved.)

What did I miss at decision time that was knowable?

Part 4 — The second look (30–60 days later)

One question, and it's the one people skip:

Knowing only what I knew then — would I make the same call again?  yes / no
Why:

The gap between this answer and the resolution entry is where most of the learning sits. Outcomes fade with time; the reasoning you wrote down doesn't.

Scoring yourself

After twenty or so resolved entries, bucket them by confidence: everything you logged at 60–70%, at 70–80%, at 80–90%. For each bucket, count how often you were actually right. If your 80% bucket hits around 80%, you're calibrated in that range. If it hits 55% — a common first result — you've just measured a gap you couldn't see before. (Want a baseline while your first entries accumulate? Take the ten-question test.)

A worked example

Date: 2026-03-04
Decision: Take the six-month consulting contract instead of
  going full-time on the app.

Prediction: The contract runs its full term and I still ship
  the app's next major version by the end of September.
Confidence: 70%

Reasoning: The client's scope is well-defined and I've worked
  with them before. App work needs ~10 focused hours/week, which
  survived the last contract. Money removes the runway anxiety
  that killed progress last year.

Most uncertain about: whether "well-defined scope" survives
  their Q2 reorg.
Would change my mind: scope creep past ~30 hrs/week for more
  than two consecutive weeks.
State of mind: calm, slightly flattered by the offer. Noting it.

Review date: 2026-10-01

Check-in — 2026-05-11: Reorg happened. New PM wants weekly
  on-sites. Points against. Confidence now: 55%.

Resolved — 2026-10-22: Contract completed. Major version shipped
  Oct 21, three weeks late.
Correct? partly (contract: yes; September ship: no)
Satisfaction: 4/5
Missed at decision time: I'd priced in scope creep but not
  travel time. It was knowable; the reorg rumors were public.

Second look — 2026-12-01: Same call again? Yes. The three-week
  slip cost little; the runway mattered more than the date.

---

Full disclosure: I make Reckon, an iOS app that is this template with the math done for you: the confidence buckets, the per-domain calibration curves, the scheduled second look. One-time, no subscription. The paper version above works fine; the app exists because I wanted the scoring to be automatic and honest.