"I knew it." Your memory says you did. A 1975 study says it probably edited the tape.
You say it when the new hire quits in month two, or when the project you dropped takes off for someone else. On other days you say the opposite: "How could I have known?" Both lines feel like memory, and both are often your mind editing the past.
Then, a month later, you make the same kind of call again.
In 1975, the psychologists Baruch Fischhoff and Ruth Beyth published a study called "I knew it would happen". Before Richard Nixon’s 1972 trips to China and the Soviet Union, they asked people to rate how likely different outcomes were. After the trips, they asked people to recall their own ratings.
People remembered giving higher odds to whatever had happened.
A decision journal is a short record you write before a choice: what you’re deciding, what you expect to happen, how sure you are, and why. You check it once the result is in. Because the entry came first, hindsight can’t edit it, so you can tell a bad decision from bad luck.
Hindsight and resulting: the two habits that grade you on luck
Hindsight is the first habit. Once you know the ending, your mind rewrites what you used to think, and you lose the record of your own reasoning. Here’s how hindsight bias quietly rewrites your past calls, if you want the long version.
The second habit has a name from poker. Annie Duke, a former professional player, calls it "resulting" in her book Thinking in Bets. Resulting means you judge a decision by how it turned out.
Duke opens the book with a famous case. In the last seconds of the 2015 Super Bowl, Seattle coach Pete Carroll called a pass near the goal line. It was intercepted, and fans and commentators tore the call apart.
Duke walks through why the call had good reasons behind it, even though the ending was bad. A run would have kept the clock moving, while an incomplete pass stops it and leaves time for another try.
Together, the two habits hand you the wrong lessons. You drop sound calls that got unlucky, and you repeat sloppy calls that got lucky. Trying to remember better won’t fix it, because memory is the part doing the editing.
What a decision journal borrows from science
Researchers had your problem first.
A scientist runs a study, sees the data, and then honestly "remembers" that this was the result they set out to test. Nobody has to lie for it to happen.
Science answered with preregistration. The Center for Open Science describes it as writing your research plan down in advance and filing it with a registry. Once the plan exists on paper, with a date, the data can’t quietly move the goalposts.
You can run the same protocol on your own life. File the prediction before the world answers, and then the world grades the prediction instead of your memory.
That’s all a decision journal is: a registry with one researcher in it.
What goes in a decision journal template?
Five lines. Farnam Street’s well-known decision journal template has 8 parts. This one keeps 5, including the one that’s easiest to skip: how you feel when you decide.
DECIDE: Today's date, and what I'm choosing, in one sentence.
EXPECT: One outcome I can check, the date I'll check it, and my confidence: __%
BECAUSE: The 1 or 2 reasons that tipped it.
DROPPED: The best option I said no to, and why.
STATE: Time of day, and how I feel (tired, rushed, flattered, calm).
Copy it into a notes app, a notebook, or a spreadsheet. Write it before you act, in whatever tool you’ll open.
Here are two sample entries. They’re written to show the form, so swap in your own calls.
DECIDE: Mon 3rd. Pitch the new client report format to my manager on Thursday.
EXPECT: She agrees to try it on one client by Fri 14th. Confidence: 55%
BECAUSE: She complained about the old format twice. The new one is ready.
DROPPED: Waiting for the quarterly review. Lost because the problem is live now.
STATE: Monday, 8am, calm, a little nervous about looking pushy.
DECIDE: Tue 4th. Buy the annual gym plan instead of paying monthly.
EXPECT: I go at least 8 times by Tue 18th. Confidence: 50%
BECAUSE: The annual plan is cheaper per month. Paying up front might commit me.
DROPPED: Monthly plan. Lost because I wanted the push of having paid.
STATE: Tuesday, 9pm, tired, just finished a fitness video.
Each line does one job.
DECIDE forces one sentence. If you can’t write it in one, you haven’t decided yet. If you’ve been stuck on that sentence for days, a simple rule for overthinkers will move you faster than another entry.
EXPECT is the heart of it, and it holds one outcome only. "She agrees and the project goes well" can’t be scored when half of it happens. "She agrees by the 14th, 55%" can be checked on the 14th. Duke’s whole book argues for thinking in bets, and a percentage is a bet.
Before a raise or price talk, put your walk-away point in EXPECT. A repeatable prep routine for negotiations shows how to set it.
BECAUSE keeps you honest later. At review time you check whether the reasons held, along with whether the result was nice.
DROPPED stops you from grading yourself against a fantasy. You compare the result with the road you really turned down.
STATE often explains the pattern later. The gym entry says "9pm, tired, just finished a fitness video." That’s a different decision from the same purchase made on a calm Saturday, and only the entry remembers which one you made.
How do you review a decision journal?
Once a week, open every entry whose check date has passed. Under each one, write a single line: what happened. Then sort it with one question: with only what I knew on that date, would I make the same call again?
Your answer picks the row. The result picks the column.
| Good result | Bad result | |
|---|---|---|
| Good decision | Earned it. Keep the process. | Bad luck. Keep the process anyway. |
| Bad decision | Got lucky. Fix the process. | Earned it. Fix the process. |
The two boxes that matter are the off-diagonal ones, where the decision and the result disagree. In those boxes the entry sets the verdict. That undoes resulting.
Look across the boxes for anything that repeats. It might be a word in the STATE line, a type of call, or a time of day. When the same thing keeps landing in a "fix" box, you’ve found the mistake you keep making, in your own handwriting.
That’s the moment to pair it with the 3 mental models worth wiring in first, so the fix fires before the next call instead of after it.
Your first week: 3 fresh calls, 3 old ones, 1 AI check
Fresh entries need time to resolve, so a journal started on Monday has nothing to review on Sunday. Fix that by working both ends of the calendar.
- Pick 3 calls that settle in 7 to 14 days. A reply to a pitch, a purchase you’re unsure about, a meeting you expect to go one way. Log each one before you act.
- Back-fill 3 calls from the last month. Write each one in the 5 lines from memory, and mark it LATE. These already have results, so your first Sunday has something to sort.
- Book 10 minutes on Sunday. Run the 3 LATE entries through the table. Next week, the first fresh entries come due.
- Bring in a second opinion before your next big call. Ask AI to argue the other side, then put what it finds in BECAUSE or DROPPED.
By Sunday you’ve recorded 6 calls and reviewed 3. The Sunday after, the fresh ones join them.
Reality check The first review runs on back-filled entries, and hindsight wrote part of them. You’ll remember your old reasons as better or worse than they were. Use that Sunday to practice the table, and save your conclusions for the entries you wrote before the result.
FAQ
Which decisions belong in a decision journal?
Log a call if you’ll still care about it in a month and you can name one outcome to check. Skip daily small choices. Three careful entries a week are enough to start.
Should I keep a decision journal on paper or in an app?
Use whatever you’ll open in under 10 seconds. Paper works if it lives on your desk. An app works if it sits on your phone’s home screen, because the entry has to happen before the choice.
How do I know if my confidence numbers are any good?
Group your entries by the percentage you gave them. Of all the calls you rated around 60%, roughly 6 in 10 should come true. If far fewer do, your confidence is running ahead of the evidence, and you need many entries before the pattern means much.
Write the bet down before the world grades it.