A decision journal in Google Sheets

A spreadsheet is a worse place to write than Notion and a much better place to count, which decides what it's for. The scoring step — sorting your resolved decisions into confidence bands and working out how often each band was right — is arithmetic, and in a spreadsheet it's four formulas that run themselves from then on.

This takes about fifteen minutes. It assumes you've settled on the fields; if you haven't, the template lists them and the method essay explains why each one is there.

One sheet, one row per decision

Twelve columns you fill in by hand, then three the sheet fills in for you.

A Decided	date
B Decision	one sentence, the call itself
C Domain	career, money, health, personal
D Prediction	specific enough that a stranger could mark it right or wrong
E Confidence	a plain number, 0 to 100
F Review date	when you'll know enough to judge it
G Resolved	empty until it resolves; this doubles as the open/closed flag
H Verdict	yes, partly or no, typed exactly
I Satisfaction	1 to 5, and a separate question from the verdict
J Same call again	filled at the second look, not before
K Check-ins	dated lines in one cell, newest at the bottom
L What I missed	that was knowable at the time

The reasoning, what you're most uncertain about and what would change your mind are the ones a spreadsheet handles worst, and they're covered further down.

Log confidence in steps of 10, and don't log 0 or 100. You are never that sure, and a 100 that turns out wrong scores the worst it's possible to score.

Three formulas

Put these in row 2 and fill them down the column.

M — Confidence band. Rounds each entry down to its ten-point band, which everything below groups on.

=IF($E2="", "", FLOOR($E2, 10))

N — Second look due. Stays empty until you record a resolution date, then shows the day the delayed question comes up. Change 45 to whatever gap you want.

=IF($G2="", "", $G2+45)

O — Brier score. The squared distance between what you claimed and what happened.

=IF($H2="", "", ($E2/100 - IF($H2="yes", 1, 0))^2)

Lower is better, 0 is a perfect call, and 0.25 is what you'd score by writing 50 on every row forever. Anything above 0.25 across a lot of entries means your confidence numbers are costing you information rather than adding any. It's a single figure for the whole journal, so put =IFERROR(AVERAGE($O:$O), "") somewhere and leave it alone for a few months.

The table that does the actual work

Off to the right, put the bands 50, 60, 70, 80, 90 down column R starting at row 2, then three formulas alongside, filled down:

S2:  =COUNTIFS($M:$M, $R2, $H:$H, "<>")
T2:  =COUNTIFS($M:$M, $R2, $H:$H, "yes")
U2:  =IFERROR(T2/S2, "")

S counts the resolved decisions in that band, T counts the ones you called cleanly right, and U is your hit rate — format it as a percentage. If the 80 row comes back at 55%, you've found the gap this whole exercise exists to measure, and you found it with four formulas.

Partly counts as a miss here, which is strict. Real decisions land on partly often enough that it matters, and I'd rather the count erred harsh; column H keeps the honest answer either way.

Two conditional formats instead of remembering

Select A2:O, then Format, Conditional formatting, and add a rule of type "custom formula is" for each:

=AND($G2="", $F2<>"", $F2<=TODAY())
=AND($N2<>"", $N2<=TODAY(), $J2="")

The first colors decisions that are open and past their review date. The second colors the ones that resolved a while ago and are still waiting on the question people skip. Give them different colors and the sheet tells you what it wants on the Sunday you open it.

What doesn't fit in a cell

Your reasoning, your uncertainties and what would change your mind are two to five sentences each, and a spreadsheet makes writing them unpleasant enough that you'll start compressing them into three words. That's the wrong direction: the reasoning is the field your future self can't reconstruct.

Two options that work. Keep a Google Doc per decision and put its link in column B, which costs a click but keeps the writing pleasant. Or write the reasoning as a note attached to the cell in column D, which keeps everything in one file at the cost of never really re-reading it. Notion has the opposite balance here: paragraphs are comfortable there, and the calibration counting is what strains.

Where this stops working

Nothing pushes anything at you. The conditional formatting is a real improvement on paper, and it still only helps on the days you open the file. Forty-five days after a resolution, nothing is prompting you to open it.

The confidence trail stays text. Column K holds the line — 70, then 55, then 40 — as words, and charting it means splitting check-ins into a second sheet, one row each. That's the tidier data model and it costs several clicks per check-in, at which point the check-ins quietly stop.

Per-domain calibration is copy-paste. Knowing whether you're sharper about money than about career decisions means duplicating the whole S/T/U block with $C:$C added to each COUNTIFS, once per domain, and again whenever you add one.

It lives in Drive. A decision journal accumulates salary numbers, health choices and candid sentences about people you work with, and a spreadsheet is one accidental share away from being readable, in a place designed to make sharing quick. Worth a minute's thought about which account it sits in.

It's still the right first choice for a lot of people. If the part you want most is the measurement, Sheets gets you there faster than any other free option, and a sheet you actually keep beats a better system you don't.

More guides in Making better decisions.

---

Full disclosure: I make Reckon, an iOS decision journal where the calibration table splits by domain without duplicating anything, the confidence trail is drawn rather than typed, and the second look arrives on its own instead of waiting for you to open a file. One-time, no subscription. The sheet above is a real setup and I'd rather you ran it than nothing; the app is for when the counting and the remembering have become the reason you stopped.