JH
Jonathan Haber

Log actual vs. estimated time for every task

Build a personal database of your own estimation errors so you can calibrate future predictions.

Key takeaways

  • What it is: Build a personal database of your own estimation errors so you can calibrate future predictions.
  • Why it works: The planning fallacy persists partly because its errors are invisible: you finish late, update the calendar, and forget the original estimate. Tracking creates a feedback loop where the cost of optimism is made explicit and memorable. Calibration research shows that people can become better-calibrated forecasters through repeated feedback on their predictions — but the feedback must be systematic, not occasional. A running log of your own ratio (actual / estimated) is a personalized reference class.
  • Evidence: Plausible mechanism, limited direct outcome data.
  • Avoid: Logging time only for tasks that went over estimate — this gives a biased sample. Logging only surprises skews your correction factor and makes you overcorrect.

Why it works

The planning fallacy persists partly because its errors are invisible: you finish late, update the calendar, and forget the original estimate. Tracking creates a feedback loop where the cost of optimism is made explicit and memorable. Calibration research shows that people can become better-calibrated forecasters through repeated feedback on their predictions — but the feedback must be systematic, not occasional. A running log of your own ratio (actual / estimated) is a personalized reference class.

How to do it

  1. 1Before starting a task, write down your time estimate.
  2. 2After finishing, record the actual time taken.
  3. 3Calculate your personal ratio (actual / estimated) over a month of data.
  4. 4Multiply your estimates by your average ratio going forward.
  5. 5Review monthly to see if the ratio improves.

What the evidence says

Mechanistic

Calibration training research (e.g., Lichtenstein & Fischhoff, 1980) shows that feedback on confidence intervals improves forecasting accuracy over time. Application to personal time estimation is mechanistically plausible but less formally studied.

Honest caveat: Most calibration training research is on probability judgments, not time estimates; transfer to scheduling is reasonable but not directly proved.

References
  • — Lichtenstein & Fischhoff (1980), Training for calibration, Organizational Behavior and Human Performance

Common mistake

Logging time only for tasks that went over estimate — this gives a biased sample. Logging only surprises skews your correction factor and makes you overcorrect.

IX Coach maintains your estimation log automatically, shows you your running accuracy ratio, and pre-adjusts new goal timelines based on your actual history.

Practice this with IX Coach →

Related practices