Overrun and Optimism
Tasks take longer than estimated, by a consistent factor, and knowing your own factor is the single most useful number personal tracking produces.
Estimates are wrong in one direction. This is among the best-established findings about how people plan, and it does not improve with experience or effort. For teams applying “Overrun and Optimism” to client or project work, the Basecamp connection provides an operational record to compare with estimates; the useful conversation is about scope and delivery, not activity as a substitute for results.
The shape of it
Underestimation is the norm and overestimation is rare.
The factor is fairly stable per person and per kind of task — commonly between one and a half and three times.
Experience improves accuracy for familiar tasks and not for unfamiliar ones.
And knowing about the bias does not remove it. People who know they underestimate continue to underestimate by about the same amount.
Why it persists
Estimates are made by imagining the work going well, and the imagined version contains no interruptions, no waiting and no rework.
The parts that overrun are the parts not imagined: the thing that did not work, the dependency, the bit nobody thought about.
And each individual estimate feels reasonable, which is why the correction has to be applied mechanically rather than by trying harder.
Finding your factor
Record estimate and actual for twenty tasks. Nothing else — two numbers each.
Divide actual by estimate; take the median rather than the mean, because one disaster distorts an average.
That median is your factor, and it is more useful than any other number in this collection.
Twenty tasks takes a few weeks and the number stays valid for a long time.
Using it
Multiply, do not add. Adding a buffer produces a figure you then work against; multiplying produces a realistic one.
Apply it openly. "That is a two-day job, which for me means four" is a defensible statement backed by your own records.
And apply it by category if the factors differ — most people find familiar work near one and a half and novel work near three.
The reference class alternative
Instead of estimating the task, find three similar past tasks and take their actual durations.
This bypasses the imagination step entirely and is more accurate than any adjusted estimate.
It requires a record of past actuals, which is the strongest practical argument for keeping one at all.
Where the overrun actually happens
Rarely in the main work.
Usually in: setup, the thing that did not work first time, waiting for an answer, review and revision, and the last ten percent.
Which means a better estimate of the main work does not help; the correction has to cover the surrounding structure.
What this does for commitments
It converts a vague sense of always being behind into an arithmetic problem with a solution.
Most chronic overcommitment is not a willpower failure but an estimation error compounded across many tasks.
Correcting the estimate corrects the commitment, and that is the whole mechanism.
What to check
Do you know your factor, or are you guessing at it?
Have you recorded estimate and actual for even ten tasks?
Do you apply the correction, or note it and then commit to the original figure anyway?
And could you find three comparable past tasks if you needed to estimate something today?
Multiply, do not add
A buffer produces a figure you then work against.
Multiplying by your own median factor produces a realistic one, and stating it openly — two days of work, which for me means four — is defensible because it comes from your own records.
Worth keeping in mind
Find three comparable past tasks and take their actual durations instead of estimating. This bypasses the imagination step entirely and is more accurate than any adjusted estimate.
Where the overrun happens
Rarely in the main work.
Usually in setup, the thing that did not work first time, waiting for an answer, review and revision, and the last ten percent. Which means a better estimate of the main work does not help — the correction has to cover the surrounding structure.
In short
Twenty tasks with estimate and actual takes a few weeks and produces a number that stays valid for a long time.
It is the most useful single figure personal tracking produces. For a complementary perspective on work, measurement or planning, consult the Information Commissioner's Office.