A 1:1 notes template | Roland Leth

This is the template I use for notes on the people I manage. It works on paper, in Apple Notes, in Obsidian, or in one doc per person; the fields matter, the medium doesn't. The reasoning behind the fields is in the companion how-to.

The rule that makes it work: notes on the person, not just the meeting. A typical 1:1 template is a meeting template — agenda, updates, action items — and six months later it can tell you what was discussed but not how the person changed. Parts 1 and 2 below take two minutes after each 1:1; Part 3 is the page you'll actually be glad exists.

Part 1 — The meeting note

Date:
What we covered:
Decisions / follow-ups:

Whatever you already write here is fine; this template doesn't change it. Two or three lines. The next two parts are the ones meeting notes usually skip.

Part 2 — What you noticed

Observed:
  (One line of behavior per entry. Something you saw or heard,
   not your interpretation of it.)
Direction, against their usual:  up / steady / down

The one discipline: behavior, not conclusions. "Quieter in group settings, third week running" is an observation. "Disengaging" is a theory, and theories go in Part 3, where they carry a date and a sureness word. The test: someone else in the room would agree it happened, whatever they made of it.

Not everything arrives in a 1:1. The tone of a PR review, the hallway answer, the incident at 2am: date those too, wherever you keep the person's page.

Part 3 — Your read of the person

One page per person, revised over time. This is the part most note systems never give a home.

I currently believe:
  (A plain sentence. "Does his best work with a deadline and
   no check-ins in between.")
How sure:  uncertain / likely / plausible / certain
Latest evidence, with date:
Last confirmed:

And the revision rule, the one that makes Part 3 worth keeping: when your read changes, don't delete the old line. Strike it through or move it below, date the new version, and note what moved you. Keep the old reads. They're the record of how the current one formed, and of how long your first impressions tend to last.

Part 4 — Before a review

Re-read the person's page, oldest entries first.
For each belief: check "last confirmed."
  Older than a quarter → it's a question for the next 1:1,
  not a line in the review.
Write the review from the page, not from memory.

This is the sweep, and it's most of the payoff. Fifteen minutes, once per person per cycle. The recency research is the reason it works: without the re-read, a review weights the recent weeks no matter how good your intentions are.

A worked example

Three months of one (invented) person, condensed:

1:1 — 2026-03-06
Covered: Q2 scope, conference talk draft.
Follow-ups: intro to the design lead.

Observed:
- Asked to own the payments migration end to end.  (ownership up)
- Estimated it at four weeks.  (estimates steady: still short)

Read — Andrei, March
- Ready for more scope than he has.        likely
  Evidence: asked for the migration (Mar 6).  Confirmed: Mar 6.
- Estimates run short when he's excited.   certain
  Evidence: three of the last four projects.  Confirmed: Mar 6.

1:1 — 2026-05-15
Observed:
- Migration a week behind; he flagged it himself,
  two weeks before I'd have seen it.  (clarity up)

Read — Andrei, May
- Estimates run short when he's excited.   certain
  Revised: still true, but he now catches it early himself.
  Different problem, smaller. (May 15)

By the June review, the growth case wrote itself from the March line, and the estimation note went in with its revision attached, which is a fairer sentence than the one memory would have offered.

How long this takes

Parts 1 and 2: two or three minutes, right after the 1:1, while the observed column is still cheap. Part 3: touched only when a read forms or moves, which some weeks is not at all. Part 4: fifteen minutes per person per cycle. If it regularly takes more than that, you're writing a dossier, not keeping notes; trim.

If you want to know how far behind your current picture is before any of this accumulates, ten questions will tell you.

Full disclosure: I make Continuum, a Mac app shaped like Part 3 with the housekeeping automated: you write the entry and tag the signals you noticed, tagged signals nudge the confidence of the beliefs they touch and leave the revision note, and "last confirmed" is tracked without you carrying it. Stale beliefs get flagged before a review instead of discovered during one. It runs entirely on your Mac, no account, free for up to three people; Pro is a one-time purchase if you'd rather not subscribe. The paper version above is complete; the app exists because Part 4 is the step that gets skipped.