The shelf was supposed to take an hour. You had the brackets, the drill, the level; you had done this before. Three hours in, you were driving to the hardware store for the second time, and the hour had quietly become an afternoon.

Everyone has a version of this story. The report that needed "one more evening" for a week. The kitchen renovation that outlived a season. The quick email that swallowed the morning. In 1979, Daniel Kahneman and Amos Tversky gave the pattern a name — the planning fallacy — and made a claim that still stings: we underestimate how long our own tasks will take even when we know perfectly well that similar tasks have always run long.

That last clause is the strange part. This is not a bias born of ignorance. Most of us can recite our own history of overruns on demand. We just don't believe it applies to the task in front of us.

A Bias That Survives the Evidence

The cleanest demonstration comes from a 1994 study by Roger Buehler, Dale Griffin, and Michael Ross. They asked psychology students finishing their honors theses to predict when they would submit. The students guessed about 34 days. The actual average was nearly 56, and fewer than a third finished by the date they had named.

Here is the detail that made the study famous. The researchers also asked for a worst-case estimate — how long the thesis would take if everything went as poorly as it possibly could. The students' catastrophic scenario averaged about 49 days. Reality overran their worst case.

And these were not naive forecasters. Asked about previous projects, the same students readily admitted they usually finished later than planned. The knowledge sat right there in memory, fully acknowledged, and changed nothing. Whatever the planning fallacy is, it is not a shortage of information.

The Inside View: Plans Are Best-Case Stories

Kahneman and the economist Dan Lovallo later described the mechanism as a choice between two ways of looking at a task. The inside view builds a forecast from the case in front of you: you imagine the steps, walk through them mentally, and add up the time. The outside view ignores the details and asks a colder question — how long have things like this taken before?

The inside view feels rigorous, but it has a structural flaw: it is a simulation of the plan, and plans are stories about things going right. When you mentally rehearse hanging the shelf, you rehearse drilling, leveling, mounting. You do not rehearse the drill battery being dead, because you have no reason to predict that specific failure — or the stripped screw, or the stud that isn't where the finder says it is.

That is the trap in one sentence: any particular obstacle is unlikely, so you predict none of them — but some obstacle is close to certain. The smooth path requires many things to each go right, and that combined probability is almost always smaller than it feels. An estimate built from an obstacle-free story is not an average expectation. It is a best case wearing an average's clothes.

Your Memory Is a Sympathetic Biographer

There is a second mechanism, and it explains why experience doesn't rescue us. In a 2005 review, Michael Roy, Nicholas Christenfeld, and Craig McKenzie argued that part of the problem sits upstream of prediction: we misremember how long past tasks took, recalling them as shorter and smoother than they were. When you consult your history, you are reading an archive that has already been edited in your favor.

Even when overruns are remembered accurately, we file them under circumstances. The thesis ran late because the printer broke and a roommate had a crisis — one-off bad luck, external and unrepeatable, therefore irrelevant to the next project. Every delay gets a private explanation, so the lesson never generalizes. We treat our overruns as anomalies and our plans as the norm, when the record says it is precisely the other way around.

One more finding from Buehler's group sharpens the picture: when observers predicted how long someone else's task would take, they were more pessimistic — and more accurate. Other people's projects we see from the outside, as members of a category. Only our own get the flattering inside view.

The Outside View, Scaled Down to a Tuesday

The institutional fix is called reference class forecasting, developed by the planning researcher Bent Flyvbjerg: forecast a project not from its plan but from the recorded outcomes of similar finished projects, then adjust. It has been written into official guidance for large public works, where the planning fallacy costs billions rather than afternoons.

You can run the same logic on a Tuesday. The personal version needs no statistics — just a record your memory can't edit.

Keep an estimation log. Two columns. Before starting anything nontrivial, write down the task and your honest guess. When it's done, write what it actually took. That is the entire practice.

The first thing the log gives you is a number: your personal ratio. After a few weeks, most people discover something like a consistent multiplier — estimates times 1.5, or times 2, depending on the kind of work. That ratio is your outside view, distilled. You don't have to feel pessimistic to use it; you just multiply. The forecast stops being a mood and becomes arithmetic.

The second thing it gives you is subtler. Buehler and colleagues found that merely recalling past experiences didn't fix predictions — people had to explicitly connect those experiences to the current task before their optimism budged. A log forces that connection at exactly the right moment. You cannot write "shelf: 1 hour" in the same column where "shelf: 3 hours" is already staring back at you. The record argues, in your own handwriting, at the only moment arguing helps.

Two rules keep the log honest. Log before you start, not after — a retroactive estimate is a memory, and we have established what memory does. And log the messy tasks, not just the tidy ones; a reference class made of your greatest hits is just the inside view with extra steps.

What the Ratio Buys You

It is worth saying what this practice is for, because it is not self-punishment. The planning fallacy is not stupidity; it is optimism doing its ordinary job, and optimism is what gets projects started at all. The goal is not to become the person who expects everything to go wrong. It is to plan with your real numbers while still hoping for the fast ones.

Concretely, the ratio changes three decisions. You promise dates you can actually hit, which quietly transforms how colleagues and family experience you. You schedule buffers as policy rather than as guilt. And you say no better — because when a "two-hour favor" reads as six in your log, you can decline with evidence instead of vague dread.

None of this requires software. A notebook page ruled into two columns has been fixing estimates for longer than psychology has had a name for the problem.

A Place Where the Ledger Lives

The practice only works if the record is effortless to keep — an estimate takes ten seconds to write down, and any tool that takes longer than that to open will lose to the optimistic voice saying you'll remember. That is the quiet reason we built Pagebox the way we did: it opens in under a second, and a lightweight database — task, estimated, actual — takes a minute to set up and a glance to consult. Your estimates live next to your notes and daily journal, so the evidence is already in your hand at the moment you're about to promise someone "an hour." If you'd like a memory that doesn't flatter you, start your two-column log at pagebox.lumenlabs.works.