Where billable hours go missing

Nobody loses billable hours in one dramatic event. It happens in ten and fifteen minute pieces, spread across a month, each one individually too small to think about. Then the invoice comes out lower than the work felt and there is no single place to point at.

This is an attempt to point at the places. The figures below are computed from assumptions stated next to them, and the assumptions are deliberately modest. If your own numbers are worse, the shape still holds and the total gets bigger.

One: the timer you did not start

You meant to. Then you opened the file, and the thing you were going to do turned out to require another thing first, and forty minutes later you are three levels deep and in flow. The timer is still at zero.

This is the most common leak and the least visible, because the time it removes is time you were never conscious of not recording. You do not feel it go.

Assume ten minutes a day. Not the whole session, just the unrecorded run-up before you remember. Across twenty-one working days that is three and a half hours a month.

Ten minutes is a conservative figure on purpose. The honest range for most people is wider and starts higher, particularly if your work has a research phase before it has an output phase.

Two: the hour you rounded down because you were not sure

Was that three hours or five? You were interrupted twice, you took a call in the middle, and by the time you sit down to write the invoice the block has blurred into “an afternoon”.

So you bill the number you can defend. That is always the smaller one. There is a real asymmetry here: overbilling by an hour is a conversation with a client, and underbilling by an hour is nothing at all, so every uncertain block resolves downward. The incentive works entirely in one direction and it is not yours.

Assume fifteen minutes, twice a week. Across a month that is a little over two hours.

This is the leak that is hardest to argue with yourself about, because the caution is correct. You should not bill time you cannot defend. The problem is not the caution, it is being in a position where the defensible number and the true number are different.

Three: Friday, reconstructed from memory

Calendar, sent mail, a scroll back through the commit log, and a guess. It mostly works. What it reliably misses is the work that left no artefact: the Tuesday afternoon spent reading someone else’s code to understand a system before changing it, the call that got moved and never made it back onto the calendar, the twenty minutes of thinking that made the next two hours possible.

Reconstruction recovers the work that produced a thing. It does not recover the work that produced an understanding.

Assume twenty minutes a week. Across a month, roughly an hour and a half.

Adding it up

On those three assumptions, ten minutes a day, fifteen minutes twice a week and twenty minutes a week, the total is about seven hours a month. Close to a full working day, from habits none of which feel like a problem on the day they happen.

Deliberately, there is no dollar figure here. Converting hours to money means assuming your rate, and a large number arrived at from someone else’s assumptions is a sales pitch rather than an argument. You know your rate. The multiplication is not the hard part.

Which one is yours

The three leak in different ways and they need different fixes, so it is worth knowing which one you actually have. A rough diagnostic:

If your tracked hours are consistently lower than the hours you remember working, but the shape of the week looks right, it is leak one. Time is missing from the start of sessions.

If your tracked hours look right but your invoices feel low, it is leak two. The time is recorded and being discounted at the point of billing.

If your Monday and Tuesday entries are detailed and your Thursday and Friday entries are round numbers, it is leak three. You are tracking live early in the week and reconstructing late in it.

Most people have two of the three.

What actually fixes each one

Leak one is a starting problem, and the only reliable fix is removing the start. Either something records without being asked, or you accept that the first ten minutes of every session are unbilled. Discipline is not a fix. You have already tried being more disciplined about it, which is why you are reading this.

Leak two is a confidence problem, and it is fixed by evidence rather than by resolve. If you can see the block, with its start, its end and what was in focus during it, the defensible number and the true number converge and the rounding stops happening.

Leak three is a latency problem. It is fixed by shortening the distance between doing the work and recording it, which in practice means either recording continuously or reviewing daily rather than weekly. A five minute review at the end of each day removes almost all of it, because you are reconstructing eight hours instead of forty and you have not slept since.

The uncomfortable part

All three of these are worse for the people who are best at the work. Deep focus is exactly the state in which you do not notice that the timer is off. Being conscientious about what you can defend is exactly what makes you round down. Working on hard problems that require understanding before output is exactly what leaves no artefact for a reconstruction to find.

The billing method quietly penalises the working style you would otherwise want to have. That is a good reason to fix the method rather than the working style.

If you want to see where your own time goes without maintaining anything, Stunda records the timeline in the background and turns the blocks into invoice lines. If you would rather do it by hand, the daily five minute review is genuinely most of the benefit, and it costs nothing.