Inside your pay cycle if you pay anybody on profit, inside the month if you want to price off it, and you should know the number rather than assume it.
Written by JD Davenport, who still sells for a living. Updated 4 November 2026.
Almost nobody measures this, which is strange, because it silently decides more than any figure that gets reported in a meeting. The lag between a job finishing and you being able to state what it actually cost determines which decisions you are allowed to make and which pay plans you are allowed to run.
Take twenty jobs that closed at least a quarter ago. For each one, find the date the work finished and the date of the last cost that landed against it. Subtract.
That distribution is your answer. Not the average, the distribution, because the tail is what hurts. A business whose costs typically close in eleven days but occasionally in ninety does not have an eleven-day lag. It has a ninety-day exposure that usually does not fire.
Most people are surprised twice. Once by how long the tail is, and once by how much of it is one or two predictable sources: a supplier who invoices monthly, a subcontractor who bills when he remembers, a credit for returned material that nobody chases.
Inside the pay cycle. You can pay commission on profit. This is the only condition under which a profit-based plan is honest, because every job on a statement is a job whose costs are final. Miss it and you are paying people an estimate and reconciling later, which damages trust faster than the margin is worth.
Inside the month. You can price off your own history. Your rate card and your estimating assumptions are built from delivered cost rather than from what you hoped it would be, and they are current enough to be worth using.
A quarter or more. You can still audit, and you should, but you cannot steer. By the time a job's real cost is known, you have sold a season of work on the old assumption. Everything you learn is history.
Never, cleanly. Then the honest position is that you do not know your job costs, and any per-job figure in circulation is an estimate wearing a decimal point. Say so out loud. It is a better place to start from than a precise number nobody can defend.
Rarely the accounting. Usually four ordinary things.
A supplier who invoices on a cycle rather than on delivery. A subcontractor with no agreed rate, so nothing can be checked until somebody asks him. A credit for returned material that goes to the account rather than to the job, so the job stays overstated forever and nobody notices because the account is right. And a cost with no job number on it, sitting in a general bucket, waiting for somebody to remember which job it belonged to.
Every one of those is a structure, not an effort problem. Every one has a fix that does not require anybody to be more diligent: an agreed rate card, a required job reference on every payment, and a monthly check that credits landed where the charges did.
You do not need the lag to be zero. You need it to be shorter than the decision it feeds.
The cheapest improvement is almost always the rate card, because it converts a question that requires a conversation into a comparison a spreadsheet can make. The second cheapest is refusing to let a payment be recorded without a job reference, because a cost with no job is a cost that will be allocated by memory weeks later, and memory allocates it to whichever job is being discussed.
Measure it again after a quarter. If the tail shortened, the change worked. If only the average moved, you fixed the easy half.
Closing job costs at line level, with the rate card and the register, is The Job Cost Standard. Whether your lag allows a profit-based pay plan is the question behind revenue versus profit. More answers on the questions page.