Ask a shop owner whether they compare estimated hours to actual hours and the answer is almost always yes. Ask them which line of last month's worst job missed, and the answer gets slower.
That second question is the real one. A total tells you that you lost. A line tells you why, and only the line changes the next quote.
Four ways the comparison breaks
It arrives too late to matter. The comparison lands when the job closes. By then the parts have shipped, the invoice is out, and — this is the part that costs real money — the same part has often already been requoted at the same wrong number. If the loop is slower than the sales cycle, it is a history lesson, not a control.
It compares two different things. The estimate is usually hours at a blended rate. The actual is usually clocked labor from the floor. Those are not the same quantity. Setup gets clocked but was quoted inside the run. An operator runs two machines and clocks to one. Rework gets clocked to the original operation and disappears. The two numbers can differ by twenty percent with nobody having made a mistake, which teaches a shop that the comparison is unreliable, which means it stops getting looked at.
It is a total, so it cannot be acted on. "Quoted 16 hours, ran 19" is interesting once. It is useless the second time, because there is no decision in it. Material, setup, run, outside service, and inspection fail for entirely different reasons and call for entirely different fixes. A single number averages all five into something no estimator can use.
It never reaches the person who priced it. In a lot of shops the comparison is a monthly accounting report. The estimator quoted the job in March, finds out in June, has quoted two hundred jobs in between, and is told a total. Nothing transfers.
What the comparison has to do to be worth anything
Four properties. Miss any one and it degrades into reporting.
Same units on both sides. If the quote says 0.32 hours per piece of spindle time, the actual has to be spindle time per piece — not clocked labor, not door-to-door. Comparable or it is noise. That depends on what the floor actually captures.
Per line, not per job. Material against material. Setup against setup. Run against run. Outside service against outside service. The shape of the miss is the information.
Live, not at close. Hour forty of a hundred-hour job is when the number is still worth something — there is still a job to fix, a scrap call to make, a customer to warn. At close there is only a lesson.
Routed back into the estimate. The next quote for a similar part should start from what actually happened on the last one, not from the same standard that already missed. If a human has to remember to go look, it will not happen at volume.
A worked example
Here is a real-shaped one. A housing, 303 stainless, fifty pieces, passivate outside.
Quoted at 0.32 hours per piece. Sell price $3,062.50 for the lot. Estimated cost of goods $2,147.70, so a quoted margin of 37.9 percent. Nothing wrong with any of that; it is a normal, defensible quote.
The job runs at 0.39 hours per piece. Margin lands at 29.9 percent.
Eight points gone. Now: the useless version of this is "we missed by eight points, quote it higher next time." Higher by how much, and on which line?
The useful version puts the lines side by side. Material came in where it was quoted — the bar stock was bought right. Outside service came in where it was quoted — the passivate house billed what it said it would. Setup ran long against the estimate. Run time per piece ran long too, but less than setup did, proportionally.
Now there is a decision, and it is specific. Setup on this family of parts is underquoted, which is a fixable standard. Run time is slightly underquoted, which is a different and smaller correction. Material and outside service are accurate and should not be touched — and that last part matters, because the instinct after a bad job is to pad everything, which makes the shop uncompetitive on the lines that were never the problem.
Same eight points. One version produces a shrug and a pad. The other produces two specific corrections to two specific standards.
Why this is hard to do with the stack most shops run
The reason this comparison is rare is not that nobody wants it. It is architectural.
The quote lives in a quoting tool. The hours live on the floor — a terminal, a traveler, a spreadsheet. The costs live in accounting. To compare line to line you need all three in the same place, keyed the same way, at the same granularity, at the same moment.
So the shops that get this working usually get it working through a pair of systems and an integration between them, and it does work. It also means the comparison depends on a field mapping staying correct across two vendors' release schedules, and on somebody owning that mapping. When the mapping drifts, the comparison does not break loudly. It goes subtly wrong, and subtly wrong is worse than broken, because broken gets fixed.
The alternative is that the quote, the clocked hour and the posted cost are the same record rather than three records that get matched. Then there is no mapping to drift, because there was never anything to map — which is how Si-Q prices the next one, and why the two-vendor version of this loop costs more to keep than it looks.
Five questions worth asking of whatever you run
Ask these about your current setup before you ask them of any vendor.
- If I open a job that is running right now, can I see estimated against actual by line — material, setup, run, outside service?
- Is the actual hour on that screen the same record the operator clocked, or a copy that something moved?
- How long after a job closes does the estimator find out it missed, and does anyone tell them which line?
- When I requote that part next quarter, does the system start from what happened, or from the standard that already missed?
- Who owns the connection between the quoting side and the costing side, and what happens when one of them ships an update?
If question one takes more than one screen, the comparison is not running. It is available on request, which is not the same thing.
The honest summary
Estimated versus actual is not a report. It is a control loop, and a control loop that reports a total, at close, to the wrong person, is not doing anything.
The shops that get margin back are not the ones that look at the comparison more often. They are the ones where the comparison is granular enough to name a cause and fast enough to act on, and where the answer walks back into the next estimate without anyone remembering to carry it.
KG Ai ERP keeps the quote, the clocked hour and the posted cost on one record, so estimated against actual is live and per line rather than a month-end reconciliation. Built inside a working precision machining shop. Request a demo.