“Two hours and I'm done.”
Six hours later it still isn't finished. In between there was a query, a file that wouldn't open, and twenty minutes in which something else was more urgent.
The remarkable part isn't the misjudgement. That happens to everyone.
What's remarkable is that next time the estimate is two hours again.
A mistake that doesn't wear off
This pattern is well studied. In a well-known series of studies, students were asked to estimate when they would finish their final thesis. On average they took considerably longer than they had said.
The genuinely interesting part came next: the same people were also asked to estimate a pessimistic case — how long it would take if everything went wrong.
That pessimistic estimate also came in, on average, below the actual duration.
So this isn't carelessness. Even people deliberately trying to be cautious estimate too tightly.
Why experience doesn't protect you
You might assume this sorts itself out over time. Someone who has written a hundred proposals surely knows how long a proposal takes.
They don't. The misjudgement persists even for tasks someone has done often.
The reason is the heart of the matter: people don't estimate from their experience. They estimate from their mental image of the specific case.
And in that image, it goes smoothly.
The inside view and the outside view
When planning, you almost automatically take the inside view: you walk through the task in your head, step by step, and add up how long the steps take.
That method has a systematic flaw. It contains only what you can picture — that is, the planned route. What it doesn't contain is the things that come up: the query, the error, the phone call, the day after too little sleep.
Not because you consider them unlikely. Because they have no shape. You cannot anticipate a specific interruption without knowing which one it will be.
The outside view works differently. It asks not “how long will this take?” but “how long did comparable things take?”
That second question consistently yields better answers — precisely because the interruptions are already contained in those past cases. You don't have to foresee them individually. They are baked into the figure.
What memory has to do with it
A second mechanism makes this worse.
People remember durations as shorter than they were. What stays in memory is the actual activity — not the interruptions, not the searching, not the pauses in between.
So someone thinking back to how long something took last time doesn't recall the six hours that passed but the two hours of focused work inside them.
The experience you draw on is therefore already flattering. That explains why even an honest look back isn't enough as long as it comes only from memory.
On top sits the cost of every task switch — something estimates practically never include. See the piece on what switching between tasks costs.
Why a buffer rarely covers it
The obvious conclusion is to add something on top.
The problem is the order of magnitude. Most people add ten or twenty per cent, while the actual overrun is frequently fifty per cent or more. The buffer moves in the right direction but not far enough.
A second effect comes with it: generously allocated time tends to get used up entirely, even when it could have gone faster. Give yourself a whole day for something and it usually takes a whole day.
Together, those two mean a buffer alone doesn't solve the problem.
What actually helps
Three things work demonstrably better than estimating more carefully.
- Force the outside view. Don't ask how long it will take; ask how long the last three comparable tasks took. That is a different question, and it produces a different number.
- Break the task apart. Itemising a task into its parts usually produces a higher — and more realistic — total. What gets underestimated as a block becomes visible in pieces.
- Ask in advance what will go wrong. Not “could”, but “will”. The wording makes a difference: it forces you to name concrete interruptions rather than abstract risks.
It also helps to distinguish between how long a task takes and how long until it is finished. Two hours of work can spread across two weeks when you are waiting for answers in between. Most estimates answer the first question and get treated as answers to the second.
When the estimate isn't the problem
Sometimes the cause lies elsewhere.
When a task regularly runs far past its time, it is worth asking whether it was being worked on at all. Preparation that feels like work is one of the most common forms of avoidance — see the piece on when preparation becomes avoidance, and the one on productive procrastination.
And when plans drag on for months without ever being closed, that is rarely a time problem. The psychology of unfinished goals is a subject of its own — as is the question of whether time management solves the right problem at all.
Why underestimating has a use too
An honest addition belongs here.
If people estimated reliably and accurately, a great many things would never get started. Hardly anyone begins a project that realistically takes three times as long as they thought.
So optimism in planning isn't only an error. It lowers the barrier to starting.
That is no argument against estimating better — but it explains why the mistake is so persistent. It has a function.
Why your own figures beat any rule of thumb
All of the above comes down to the same thing: you need past figures that don't come from memory.
This is where Mentavo can come in — not by timing anything, but by keeping plans, appointments and notes visible over a longer period. In hindsight it becomes apparent when something was started and when it was actually finished. That is often a considerably wider gap than the one you remember.
Whether a delay was a one-off or a pattern only shows after several rounds anyway — see the piece on patterns and coincidence.
The short version
The estimate is too tight because it maps the planned route rather than the actual one.
No stronger resolution fixes that; a different question does. Not how long it should take, but how long it took last time.
The answer to that is usually uncomfortable.
And almost always more accurate.




