How to track time on jobs accurately

Last updated: September 2026

Inaccurate in a predictable direction

Ask anyone to remember a week of work and the errors are not random. Reconstructed hours round to comfortable numbers, drop the twenty-minute return visit, forget the half hour spent finding parking, and quietly shorten the job that went badly.

Which means the hours you most needed, the ones on the jobs that overran, are exactly the ones that go missing. Quote from that data and you will keep underpricing the same work, for years, with numbers that look like evidence.

Accuracy here is not about honesty. Everyone is honest. It is about the gap between the work and the writing down.

Record against the job, on the day

One change does most of the work: log time against a job at the point you are doing it, rather than against a week at the end of it.

It is faster, because you are answering a question you already know the answer to. It is more accurate, because you are not remembering. And it produces something a weekly grid never can: a per-job figure you can compare with what you charged.

A weekly total tells you what to pay somebody. Per-job time tells you which work is worth taking, and those are completely different questions.

Agree the rules before you switch anything on

Time recording goes wrong when people are unsure what counts. Settle these four first, because consistency matters more than the answers:

1. Does travel count? Either answer works. An unspoken answer does not.
2. What about collecting materials? Usually yes, and usually against the job it was for.
3. How is a return visit handled? Logging it against the original job keeps the true cost of that work together.
4. What happens to time on a job you quoted and lost? It still happened, and knowing what unsuccessful quotes cost is useful in itself.

Be straight about why you are collecting it, too. People assume time recording is surveillance unless told otherwise, and it lands very differently when the stated purpose is pricing work properly.

Where Dispatch fits

Time logging in Dispatch is switched on in Settings and off until a business turns it on, then opt-in again for each engineer. It is deliberately not something that appears on people’s phones one morning without a conversation.

Once on, hours are logged as Labour or Travel against the job being worked on. A timer on the job screen prefills the form rather than running the record, which suits a trade where you put the phone down to work. Each person sees their own week on an Hours screen, so they can check what they recorded rather than hope it saved.

A week can be exported as CSV or PDF, or emailed to yourself, which is the format a bookkeeper or an accountant wants. There is no approval workflow: nobody signs your timesheet off, it is a record rather than a submission.

Because the hours sit on the job, they line up with the quote, the photographs and the invoice for the same work, which is the comparison that makes the whole exercise worth doing.

Turning hours into better quotes

After a couple of months of honest per-job time you can answer questions that were guesswork: whether a particular job type is consistently underpriced, whether one customer absorbs more time than they should, whether travel wipes out the margin on work the other side of the county.

Most trades businesses find at least one job type they had been quoting at a loss. Fixing that one price is usually worth more per year than everything else on this page put together.

If you want the mechanics of the timesheet itself rather than the accuracy problem, timesheets and logging time covers the weekly output, and improving job profitability covers what to do with the numbers once you have them.

What to be realistic about

Nobody logs time perfectly. Aim for good enough to price with, not for a legal record. A week where four of five jobs have honest hours on them is already transformative compared with none.

It needs a connection, so time on a job in a basement or a plant room is entered when you are back in signal, and that is worth agreeing as a habit rather than treating as a failure.

And do not use it as a stick. The fastest way to get inaccurate data back is to question every entry, because the entries will simply start looking like whatever avoids the conversation.