What does it mean that record your calibration over time?

A log of predictions and outcomes shows you exactly where your perception is off.

Example: A product manager keeps a decision journal for six months. Every time she makes a prediction — "this feature will increase retention by 15%," "the candidate will pass the technical screen," "the client will renew" — she writes it down with a confidence level and her reasoning at that moment. Six months later, she reviews. She discovers three patterns that no amount of real-time reflection would have revealed. First, she is chronically overconfident on timeline predictions: events she said were "90% likely within two weeks" actually happened within two weeks only 55% of the time. Second, she is well-calibrated on people assessments — her 70% predictions about candidate performance resolve correctly about 68% of the time. Third, she systematically underestimates technical complexity for projects outside her domain. Without the written record, she would have remembered herself as roughly accurate across all three categories. Memory would have quietly edited her predictions to match what actually happened, and the specific, actionable pattern — overconfident on timelines, calibrated on people, underconfident on unfamiliar technical scope — would have remained invisible. The journal did not make her smarter. It made her perception of her own judgment accurate.

Try this: Start a calibration log today. Choose the simplest format you will actually maintain — a spreadsheet, a notes app, a dedicated notebook. For the next thirty days, record at least one prediction per day using this structure: (1) Date. (2) The prediction, stated as specifically as possible — "Client X will respond to the proposal by Friday" rather than "I think the client will respond soon." (3) Your confidence level as a percentage — 50%, 70%, 90%. (4) Your reasoning — two to three sentences on why you believe this. (5) Your current mental and emotional state. After the outcome is known, return to the entry and record: (6) What actually happened. (7) The date it resolved. (8) One sentence on what the gap between prediction and outcome reveals about your perception. At the end of thirty days, sort your predictions by confidence level. For your 90% predictions, how many resolved correctly? For your 70% predictions? For your 50%? A perfectly calibrated person would see those percentages match their confidence levels. The gaps between your stated confidence and your actual accuracy are the precise coordinates of your miscalibration.