The overrun happens after the contract is signed, in two specific places, and nothing you do at pricing time can reach it.
Written by JD Davenport, who still sells for a living. Updated 4 November 2026.
A job is priced once, in an office, from photographs and a measurement. It is then built in the world, by people who were not in that office, on a day nobody could see from it. The gap between those two jobs is where the margin goes, and it is almost never one large event. It is a second trip, a part nobody ordered, an hour of standing around, and a quantity the estimate was optimistic about.
This is worth saying first because it is where most people look and it is where the answer usually is not. I have audited jobs that sold at or above their margin floor and still lost money. They were not rare and they were not exotic. They were priced correctly and executed expensively.
A margin floor protects the price. It does not protect the outcome. Everything that happens after the signature is outside its reach, and that is where the money leaves.
Before anything is bought. One review, after the job is sold and before material is ordered or a crew is scheduled. It asks whether the job is actually ready: is the scope clear, does the measurement match what was sold, is there anything on this job that is not on the estimate. A job that reaches this point not ready does not get material ordered. That refusal is the whole mechanism. A review that runs after the truck is loaded is a report, not a gate.
On the day the work finishes. One record per job, filled in on site, by whoever was there. What was bought that nobody planned for. What quantities were not what the estimate said. How many visits it took. It takes four minutes and it is the only moment when somebody who saw the job is still standing on it.
Two checks on their own are paperwork. What turns them into a system is that the second one feeds the first.
Every item the close-out records as unplanned becomes a line on the pre-install checklist for next quarter. The thing your crew forgot in March is a box somebody ticks in June. Which means the checklist stops being a generic list of things contractors forget and becomes a specific list of things your jobs forget, sorted by how often it actually bites you.
A parts checklist copied from another company tells your crew what somebody else forgets. That is why generic ones get ignored, and the ignoring is rational.
Because the first time it happens it is not one. A missing piece of trim is bad luck. The second time is bad luck. By the tenth time it is a structure, and the tell is that the same item keeps appearing.
That is also the honest test of whether a fix worked. Not whether somebody was spoken to. Whether the item stopped recurring after the date the fix went in. If it kept happening, the fix was a note.
Two numbers, and neither is a margin.
How often does a job reach the gate not ready? That rate tells you whether the problem is upstream in how jobs are sold and handed over, rather than downstream in how they are built.
How often does the job finish in one visit? Second trips are the cheapest large cost in most contracting businesses, because nobody invoices for them and nobody codes them, so they are invisible in exactly the place you would look.
Both are counts, not percentages of revenue, and both can be kept on one row per job.
Job costing first, then production. Not because job costing is more important, but because the gates are justified by what the jobs cost when the gates were not there. Install them in the other order and you are asking people to fill in two more pages on faith.
The two gates, the trackers behind them, the builder that makes the checklists out of your own jobs, and a board for who is doing what are The Production Standard. Closing job costs at line level first is The Job Cost Standard. More answers on the questions page.