Track decisions. Learn from them.
Decision Journal
Log the reasoning behind significant decisions, then review them. Eliminates hindsight bias and builds genuine calibration.
After a decision plays out, humans automatically rewrite the history of why they made it. We incorporate the outcome into our memory of the reasoning. If it worked, we remember being more confident than we were. If it failed, we remember doubting it from the start. This is hindsight bias — and it makes it nearly impossible to learn from experience without a written record.
What to log
For each significant decision: (1) The decision and context — what were you choosing between, and why now? (2) Your reasoning — what mattered, what you were trading off. (3) Your confidence level — 0-100, how sure are you this is right? (4) What you expect to happen — specific, falsifiable predictions. (5) Date of the expected outcome — when will you know if this was right?
The review
Quarterly, revisit past entries whose outcome date has passed. Compare your predicted outcome to what actually happened. Compare your stated reasoning to your actual post-hoc reasoning. The gap is where you learn. You'll find patterns: where you're systematically overconfident, where you consistently underweight certain factors, where your stated reasoning wasn't your real reasoning.
Calibration vs outcomes
A good decision made from good reasoning can have a bad outcome — probability doesn't guarantee results. A bad decision made from bad reasoning can have a good outcome — for a while. The journal helps you separate decision quality from outcome quality. You're not trying to have more good outcomes; you're trying to make better decisions. Better decisions produce better outcomes over time.
In engineering leadership
Log: hiring decisions, architectural choices, prioritisation calls, team structural changes. The typical pattern on review: you underweighted culture fit vs skill in hiring, you overestimated your team's capacity to absorb change, you were too anchored on the loudest stakeholder's view. These patterns don't become visible from memory alone.
In practice
You decide to promote an engineer to a lead role. Log: confidence 70%, reasoning — strong technical judgment, showed ownership on the last project, team respects them, uncertain about their conflict resolution style. Expected outcome in 6 months: team velocity maintained or improved, they own at least one cross-team initiative. Review at 6 months — did velocity hold? Did they own anything? What was the actual delta vs your prediction?
TL;DR
Write down decisions + reasoning + confidence before outcomes are known. Review quarterly. The gap between prediction and reality is your actual learning.