What Didthe JobActually Cost?
June 5, 2026 · 6 min read
The machine ran for six hours. The quote said it would run for four. Nobody caught it until the P&L came out.
Walk a shop floor and ask the operators where the bottleneck is. You'll get one set of answers. Ask the same question in the front office, and you'll get a different set. Both groups are looking at real problems. Neither group has the whole picture of any given job.
Margin leaks out of that gap.
Most shops are quoting blind
Most discrete manufacturers don't have this data, at least not in any form a quoting team could use. The equipment is producing it. Modern controls track cycle starts and stops, fault codes, alarm states, and program changes inside the machine. Almost none of that gets out of the PLC and into anywhere it can be analyzed. The gap between machine monitoring and operational intelligence is exactly this: one tells you the machine was running, the other tells you what the run actually cost.
The maintenance team logs work orders in one system. Schedulers track jobs in another. Operators carry the rest in their heads and on handwritten tags.
A few shops have a monitoring dashboard bolted on to one or two cells. Even those usually stop at uptime and downtime by machine. The job-level tie-back, what this part costs on this run with this operator on this material, isn't there.
When a sales engineer prices a new RFQ, the inputs are usually a CAD file, a material spec, an hourly rate sheet that hasn't been updated in two years, and whatever the estimator remembers about the last time a similar part ran. The actual production history of similar work doesn't make it into the calculation. The estimating team would happily use that data if it showed up in a usable form. Nobody has built the path from the machine to the quote.
So shops quote off what they can see, which is the rate sheet, and the rate sheet drifts further from reality every quarter.
The rate sheet drifts further from reality every quarter.
Where the standard cost goes wrong
A standard hourly rate is a simplification. It bundles together a machine, an operator, overhead allocation, an assumed setup time, and an assumed run time. On a stable mix, that simplification is close enough. In a high-mix shop running short orders, the assumptions inside the standard start to crack.
- Setup variance Setup is treated as a fixed number. In reality, setup varies by tooling, by operator experience on that specific job, and by what the cell was running before. A blanking line changeover with similar material might be 15 minutes. Going from carbon to aluminum it's 60 or more, since the machine needs to be thoroughly cleaned. The rate sheet has one number.
- Indirect labor A job hits the laser for forty minutes. Before and after that, an operator spent fifteen minutes pulling material from the rack, ten minutes at the deburr bench, and twenty minutes walking parts to QA and waiting on the inspection. Forty-five minutes of paid labor, none of it on a machine, none of it on the production dashboard. It gets absorbed into a generic shop overhead rate or lost entirely, and the part comes out looking more profitable than it ran.
- Scrap and rework Scrap and rework get tracked, but not against the original quote. A part that gets remade on the second shift shows up in the scrap log and in the rerun work order. It doesn't show up as a hit against the margin on that customer's PO. By the time anyone reconciles the two, the quote on the next RFQ from the same customer has already gone out.
The result is a P&L surprise at month-end. The plant manager has seen every minute of it happen and doesn't have a clean way to explain it. The quote came out of a system that didn't know any of it.
What a closed loop actually looks like
Another dashboard won't close the gap. What will is ensuring every completed job leaves behind a record that next quarter's quote can actually use.
That record needs a few things. Machine time, broken down between setup, run, and downtime, tagged to the job. Operator context for anything the machine couldn't see on its own, captured at the moment, not reconstructed from memory a week later. Material consumption and scrap, tied to the same job ID. Indirect labor that touched the job, even if it never touched the machine.
None of those individually is hard. The hard part is the tie-back, getting all of it onto the same job record so it can be summed up and compared to what was quoted.
Once that exists, the quoting question shifts. Instead of starting from a standard rate, you start from history. The last forty runs of this part averaged 2.3 hours total, with setup running between 35 and 50 minutes depending on the operator. On a new part, you don't have history for that exact part, but you have history for parts of similar complexity in similar materials on the same machine, and that's a better baseline than a 2015 time study. That history is what cost-to-serve analysis is built on.
This is also where the conversation about a job stops being adversarial. The sales side and the plant side are working off the same data. Either the historical actuals support the price, or they don't. If they don't, the disagreement happens before the quote goes out, not at the month-end review.
A 90-minute job with a 40-minute setup is a different economic event than a 90-minute job with a 10-minute setup. Pricing them the same is a real way to lose money on small orders while convincing yourself you're winning them.
Setup is the biggest line item nobody quotes
If a shop is going to instrument one thing first, instrument setup.
Setup is the most under-measured and most under-priced category of time in high-mix metal fabrication. A press brake at 80% utilization sounds healthy. If five of those hours were six changeovers, the press brake produced parts for three hours and adjusted tooling for the other five. The dashboard rolls all of that into "utilization." So does the standard hourly rate and the quote.
For shops running shorter and shorter orders, the setup-to-run ratio matters more than the machine rate. Separating setup time from run time on the floor, even imperfectly, changes the quoting math. It also gives the engineering team something concrete to attack. If a particular family of parts has a consistent 35-minute setup, you can put a dollar value on reducing it to 20 minutes. That's a number the CI team can actually work against.
Knowing your costs lets you say no
Better quoting is the obvious benefit. The underrated benefit is permission to walk away from work you shouldn't be doing in the first place.
Most shops have a few customers, or a few part families, that the operations team knows are unprofitable. Walking away from them feels risky because nobody can prove the alternative is better. With actual cost data tied to the work, the conversation gets concrete. This customer's parts run 22% over the quoted hours on average. The margin for this part family is negative once setup and rework are accounted for. You can either reprice or stop quoting.
Reprice or stop quoting. Either one is a defensible call once the numbers are on the table. The only bad option is to continue guessing.
Shops that quote from cost data win more profitable work and stop subsidizing the rest. Some of the data is already in the building. The rest is a question of what to start capturing, and what to wire together with what. How floor data connects to business outcomes is the longer version of that answer.
The data is already on your floor.
Wire it to your quotes.
Your quoting team could start from history instead of a rate sheet that drifted two years ago. The machine time, setup, scrap, and indirect labor are already in the building. MACH puts them on one job record.
Book a Diagnostic