All modules

Module 06 — Decision Log

Six months from now,
will you remember why?

Decision amnesia is real. Without a record of your reasoning, you can't tell good process from bad luck — or learn anything from either.

The problem

You forget what you decided. More importantly, you forget why.

Six months after a decision, you can't recall the options you considered, the information you had at the time, or what made you choose what you chose.

So when things go wrong, you're left with two bad options: relitigate the decision unfairly — judging past-you with information they didn't have — or repeat the same mistake because you can't trace the pattern.

The decision log exists so future-you has access to past-you's reasoning. Not to second-guess it. To learn from it.

Without a decision log
Options you considered Gone
Information you had Gone
Rationale for choosing Gone
What you expected Gone
Ability to learn from it Gone

The framework

A good decision with a bad outcome is still a good decision.

Annie Duke's core insight in Thinking in Bets: you can't judge the quality of a decision by its outcome. Sometimes good processes produce bad outcomes. Sometimes bad processes get lucky.

Conflating the two is how intelligent people stop improving. They avoid decisions that produced bad outcomes — even when the decision itself was sound. They repeat decisions that produced good outcomes — even when they were pure luck.

Separating process quality from outcome quality is the only way to actually improve. The decision log makes this distinction concrete — because you can only evaluate process quality if you recorded the process.

Process vs. Outcome — Duke's matrix
Good process + good outcome
Deserved win
The best case. Reinforce the process.
Good process + bad outcome
Still a good decision
Bad luck. Don't change the process.
Bad process + good outcome
Got lucky
Fix the process before luck runs out.
Bad process + bad outcome
Expected result
Fix the process. The outcome was inevitable.
Based on Annie Duke · Thinking in Bets (2018)

How you use it

Log it in plain language. Query it like memory.

Log a new decision

"Log a decision: we're going with annual pricing over monthly. Options were monthly, annual, lifetime. Rationale: lower churn risk, better cash flow. Revisit in 90 days."
→ Decision logged. Revisit date set: Apr 15 2026.
→ 3 options recorded (monthly, annual, lifetime).
→ Tagged: pricing, strategy.

Query your decision history

"What decisions have I made about pricing in the last year?"
"What decisions are due for revisit this month?"
"Show me decisions involving Rebecca"
"What was the rationale for choosing Stripe over Paddle?"
1
Log the decision
Plain language. No forms. Claude extracts the structured fields automatically.
2
Set a revisit date
When the outcome becomes knowable. 30 days for tactical decisions, 90 days for strategic ones.
3
Get surfaced when it's time
SoY prompts you with the original reasoning and expected outcome. No digging required.
4
Record the actual outcome
What happened vs. what you expected. This is where learning gets compounded over time.

What a record contains

Seven fields. Everything you need to evaluate the decision later.

Not a journal entry. Not a post-mortem. A structured record of process — captured at the moment of decision, before outcome bias sets in.

1
The decision made
A clear, unambiguous statement of what was chosen.
2
Options considered
What you could have done instead. These are the counterfactuals future-you needs.
3
Rationale
Why this option. What information or reasoning led here.
4
People involved
Who was consulted or had input. Auto-linked to your contacts.
5
Expected outcome
What you thought would happen. This is the prediction you'll evaluate against.
6
Revisit date
When the outcome becomes knowable. SoY surfaces the decision automatically on this date.
7
Actual outcome
Recorded on revisit. What actually happened vs. what you expected. The learning lives here.

Revisit prompts

When the date arrives, SoY brings the decision back.

Most decision logs fail because the revisit never happens. You forget to check, or by the time you remember the outcome you've lost the thread of the original reasoning.

SoY surfaces the decision automatically — with the original record fully loaded. Your only job is to record what happened.

Over time, this builds a real picture of your decision quality. Not your feelings about your decisions — your actual track record of prediction accuracy, separated from outcome luck.

Revisit prompt — Apr 15 2026
"90 days ago you decided to go with annual pricing over monthly. You expected lower churn risk and stronger cash flow. What actually happened?"
→ Type your outcome here...

Claude will record it, compare it to your prediction, and update your decision quality score.

Example record

What a logged decision looks like.

This is a real record structure — clean, queryable, and retrievable months later with a single question to Claude.

Annual pricing model over monthly
Awaiting revisit
Date logged
Jan 15, 2026
Revisit date
Apr 15, 2026 — 90 days
Options considered
Monthly billing Annual billing Lifetime license
Rationale
Lower churn risk on annual vs. monthly. Stronger cash flow signal for planning. Simpler operations — one billing event per customer per year. Lifetime felt too early given unknown retention rates.
People involved
Kerry Morrison Marcus (advisor)
Expected outcome
Lower monthly churn rate. Upfront revenue concentration in Jan/Feb. Some conversion friction vs. monthly option.
Actual outcome
Not yet recorded — revisit Apr 15, 2026

✦ · A real example

What you logged.
What it surfaces 90 days later.

A decision record, split into what you entered and what the system tracks and resurfaces automatically.

What you logged
Decision
Go with annual pricing over monthly
Options considered
✓ Annual — lower churn risk, better cash flow predictability
— Monthly — easier conversion, higher flexibility for customers
— Hybrid — both offered, discount for annual
Rationale
Monthly churn in the first cohort was the primary risk. Predictable ARR lets us hire 6 months ahead instead of reacting.
Revisit date
90 days — after Q1 conversion data is in
What the system tracks
Status
Logged Feb 3 — revisit window: May 3
↻ revisit in 69 days
Linked contacts
Rebecca Chen Marcus Webb
Both were present when decision was made — linked from transcript
Process / outcome rating
—
Process: strong
Outcome: not yet recordable
Check back May 3
Related decisions
2 other pricing decisions in the log — one from Oct (monthly trialed, abandoned) surfaces as direct evidence for this call.
The point
You logged the reasoning once. The system links it to the people in the room, surfaces the revisit automatically, and connects it to prior decisions you'd already forgotten about. That's decision memory.

Software of You

Build a real picture of how you decide.

One-time payment. No subscription. Works with your existing Claude subscription.

Get SoY — Free Next: Module 07 — Journal