I Estimated Five Tasks and Timed Them. The Planning Fallacy Measured.

I have written before that knowing a bias has a name does not protect you from it. So the obvious next step was to stop asserting that and go measure it. I took five ordinary tasks, wrote down how long I expected each one to take before I started, then timed how long it actually took. Every single estimate was short. That part I expected. The part I did not expect was by how much, and the fact that the error survived my knowing it was coming.

The five tasks were deliberately boring: writing a report section, clearing an inbox, booking a round of travel, fixing a small bug, and prepping a presentation. Before each one I wrote my estimate. Then I worked normally, no rushing, and recorded the clock time when I actually finished. I allowed myself the usual interruptions, because a best-case estimate is not an estimate of real life.

Here are the numbers. The report section I guessed would take me an hour; it took three and a half. The inbox I guessed thirty minutes; ninety. Travel booking, twenty minutes; fifty-five. The bug, two hours; four. The presentation, two hours; five. My average error was over two and a half times too optimistic, and the one task I was most confident about — the report — was the one I got most wrong.

The planning fallacy, named by Daniel Kahneman and Amos Tversky, is the gap between the “inside view” and the “outside view.” The inside view is imagining the specific task unfolding smoothly step by step. The outside view is asking how long this kind of task has actually taken in the past. My estimates were all inside view — I pictured the task going well, which by definition excludes every interruption, stumble, and detour that a real task contains.

The part that genuinely unsettled me is that I did all of this while knowing the bias was being measured. I went into each task thinking “remember, this will take longer than you think,” wrote a number anyway, and was still off by the same margin. Awareness did not buy me a single minute of accuracy. It might even have cost some, because my awareness let me believe I had compensated when I had not.

There was one fix that measurably helped, and it is called reference-class forecasting. Instead of imagining the task, I looked back at the last time I did something like it and used that as the starting number. For the report, rather than asking “how long should this take,” I asked “how long did the last section of this size take me,” got a real number, and then padded it. My estimates in the second round landed within about twenty percent of reality instead of two and a half times off.

I should be clear about what the fix does not do. It did not make me faster. It made my promises honest, which is a different and better thing. I was still late the same amount; I just stopped telling other people I would be early. That single change — saying a more truthful number up front — turned out to be worth more than any productivity trick I have collected.

The planning fallacy is not laziness and it is not a lack of discipline, which is why beating yourself up about it does nothing. It is a feature of how the brain builds plans — the plan is a simulated best case, and simulation never includes the friction that only shows up in the doing. The only correction that works is to stop imagining and start tabulating. Five tasks, five underestimates, and one hard lesson about asking my own history instead of my own optimism.