Every few months I look back. What did I say I'd do? What did I actually do? What went wrong? What will I do differently? Four questions, written down.
A quarterly note is a short review of the last three months. It compares what I planned with what happened, lists anything that broke, says what I learned, and records which plans I'm changing. Keeping the same headings each time makes it easy to compare quarters.
Fixed-structure quarterly retrospective: plan versus delivery, incidents and fixes, learning, decision changes with links to the original records, and the next quarter's commitments with confidence levels. Scoring calibration is checked against the previous quarter's estimates.
One review, the same shape every time
A quarterly note is a review of the plan. Its value comes from comparison: this quarter against the last one, and what I expected against what happened. That only works if every note has the same shape, so I use a fixed template.
The template
Each note has six parts, always in this order.
- What I planned comes first, as a short list copied from the previous note's commitments, so I can't quietly rewrite them.
- What shipped follows, with facts only, taken from the change log, with links.
- What broke covers anything that was wrong or broken, how I found out, and what I changed so it doesn't happen again.
- What I learned is one or two points, in plain words. If I can't explain it simply, I haven't really learned it yet.
- Decisions I'm changing lists any planning decision I now think was wrong, with a link to its original record and the reason.
- Next quarter sets three to five commitments, each with a confidence level.
Why confidence levels
Writing a confidence level next to each commitment lets me check how good my estimates are. If I say I'm 80% sure of five things and only two happen, my estimates are too optimistic. Over several quarters, that tells me whether to plan less or plan differently.
A worked example
This is an illustrative next-quarter section and the review that might follow it. None of it is a real commitment.
| Commitment | Confidence | Outcome | Hit? |
|---|---|---|---|
| Every collection has two articles | 90% | Done | Yes |
| Publish the threshold calculator | 70% | Done | Yes |
| Write the Decision Diagnostic in full | 60% | Half drafted | No |
| Add audio to all new articles | 50% | Not started | No |
Two out of four were done. The average stated confidence was 67.5%, against a 50% hit rate. On this small sample, that suggests I'm a little overconfident. Four items is far too few to be sure, which is why the comparison is worth running across several quarters.
What doesn't go in
No traffic figures and no sales. The site carries no offers until 2027, and I don't want visitor numbers to steer what I write. The note is about whether I did what I said, and whether what I did was right.
I also leave out anything I can't support from the change log or a decision record. If I remember something differently from the log, the log wins, because memory tends to flatter.
Where I got stuck
The "decisions I'm changing" section is the hardest to write. It's easy to list what shipped and harder to say "I chose badly here". I've added a rule that each note must name at least one decision I'd make differently, even if it's small.
Sources
This article describes my own template. It uses no external facts or figures; the example is illustrative.