Estimating From Your Own History

Estimating by imagining the work is unreliable in a known direction. Estimating by looking up what similar work actually took is not, and it requires only that you kept the record.

A concrete way to test the method is outlined in the detailed reference. Use a short pilot and keep only the fields that still make sense when comparing the record months later.

Why it works

Past actuals already contain everything an estimate omits: the interruptions, the waiting, the rework, the part nobody thought about.

A related source worth keeping beside this note is ICO data-protection guidance, particularly for decisions involving data protection.

No correction factor is needed, because the correction is embedded.

And it improves with the record rather than with your judgement, which is the useful property — judgement does not improve on this task, and the evidence for that is strong.

Building the class

For each completed piece of work: what it was, what it actually took, and two or three words about what made it unusual.

Three lines. Kept separately from the daily record, or derived from it.

After twenty items you can estimate. After fifty you can estimate well, and a working practice reaches fifty in two or three years.

Using it

Find three to five past items most like the new one.

Take the range, not the average. "Between four and seven days, most likely five" is both more honest and more useful than a single number.

Adjust for the stated differences, which is where the two or three words earn their place.

And quote the range. A single figure invites being held to its lower end.

What makes a good comparison

Same kind of work rather than same client.

Similar scope, similar unfamiliarity, similar dependency on other people.

That last one matters most and is most often ignored: a job waiting on three approvals is a different job from the same work with none.

When there is nothing comparable

Decompose until the parts are comparable. You may not have built this before, but you have written a specification, done a review, and fixed the thing that did not work.

Where genuinely nothing compares, say so and price the uncertainty rather than producing a confident figure.

And treat the first instance as an investment in the class, which makes the second estimate real.

Recording the outcome

Whether the estimate held, as a yes or no.

What the overrun consisted of, which after a while shows one or two recurring categories rather than general slippage.

And elapsed time separately from effort, because for anything with dependencies those are different numbers and clients usually care about the first.

The window

Use the last two or three years, or the last twenty comparable items.

Older data describes different tools, different skill and a different you, which the relevance note covers.

Estimating for others

Your comparison set is yours. Somebody else's work will differ, usually by more than people expect.

In the absence of their history, assume worse than your own by a meaningful margin, which is uncomfortable and consistently borne out.

What to check

Do you have a list of completed work with actual durations?

How many comparable items could you find for your next estimate?

Do you quote a range or a number?

And do you record what the overrun consisted of when one happens?

The point

Take the range, not the average, and quote the range.

A single figure invites being held to its lower end, and the range is what your history actually supports.

Worth adding

Record what the overrun consisted of, not just that it happened.

After a while it shows one or two recurring categories rather than general slippage.

Additionally

Decompose until the parts are comparable when nothing matches.

You may not have built this before, but you have written a specification and fixed the thing that did not work.

Finally

Use the last two or three years, or the last twenty comparable items.

Older data describes different tools, different skill and a different you.

In summary

Past actuals already contain the interruptions, the waiting and the rework that an estimate omits by construction.

No correction factor is needed.

Last word

The practice this points at is unglamorous: a small daily act, a short weekly correction, an annual read, and a written page of conventions beside the data. None of it is difficult and all of it is easy to skip, which is why most records do not reach the point where they start paying.