Everyone in manufacturing knows that the fastest quote has an advantage. The problem is that being fast usually means being less thorough. Imagine … sales manager gets an RFQ from a new prospect, and she knows if she return with a number by the end of day, two other shops will. So she pulls up the quote spreadsheet - the one still based on manual time studies from 2015 - and prices the job at what looks like a 35% margin. The quote is accepted, the shop gets the job, and it ships on time.
Then the month-end close comes around, and the plant manager is looking at the P&L saying the job actually ran closer to 12%. The press brake had unplanned downtime. A scheduling change forced an extra set-up on a different workcell. A maintenance call pulled the operator off the flat laser. The plant manager saw all of it happen. The sales manager never heard about any of it.
Now they’re in a meeting arguing about margin, and they’re both right. But the floor data never made it back to the quoting process, so the next RFQ will get priced the same way. Same argument next month.
Where the quote breaks down
Most shops quote prospective jobs by drawing from tribal knowledge, old spreadsheets, and a rough hourly rate per machine. That worked when the product mix was stable. Now, jobs are shorter and the mix is higher. The old estimates drift further from reality every time something changes on the floor.
Meanwhile, your machines could be generating actual-time data on every job they run. Your PLCs already have it: Cycle starts, cycle stops, idle periods, fault events, sitting in registers waiting to be pulled. Your operators could be adding the context that the machine cannot: Why the job stopped, what the set-up involved, and whether the material ran differently than expected.
A 2021 WEF survey of 1,300+ manufacturers determined just 39% of manufacturing executives report that they have successfully scaled data-driven use cases beyond the production process of a single product. D&B found in 2025 that only 36% of manufacturing firms feel they can make informed decisions with their current data. A 2022 HBR report found that only 25% of companies consider themselves truly data driven. Shops are collecting more data than ever, but they don’t trust it or are failing to leverage it… leading them to continue operating off old spreadsheets.
Some shops have quoting software that estimates machine time from CAM simulations. That’s a step forward. But machine time is only part of what a job costs. Load and unload, first article inspection, deburr, material handling, program prove-out, and waiting on the crane. Often that non-value-add time around the machine adds up to more than the cutting time itself.
What recorded time actually tells you
Set-up is the big one. In high-mix shops, a press brake might show 80% utilization on the dashboard. Looks great. But if the operator made six changeovers during that shift, a big chunk of that “utilization” was set-up, not production. The monitoring system did not distinguish between bending parts and adjusting tooling. Tag those intervals against the actual job record, and you will find the set-up-to-run ratio on short jobs is eating margin you didn’t know you were losing.
Maintenance ripple effects are another factor. When a hydraulic pump fails on a forming cell, the repair cost shows up as a work order. The harder number is the time cost: jobs delayed, overtime to catch up, and the risk of scrap from rushing the next run. If you’re capturing actual time-to-complete on every job, you can see that jobs running the week of a major repair took 15% longer.
Now you have a number you can build into your cost model.
From recorded times to predicted times
Once you have a few months of recorded job times tied to machine, operator, material, and job type, you have enough data to start quoting from actuals rather than estimates. If the last 40 runs of a particular part averaged about two hours total, you can quote with a confidence interval instead of a gut feeling. Flag the outliers.
If three of those runs took closer to three hours, look at what was different. With more history, even a basic regression that accounts for material type and part complexity can produce a predicted job time closer to reality than whatever is in the spreadsheet. It updates every time a job runs, because every completed job is new training data.
Even on a brand new part you've never run before, you're not starting from zero. If you know how similar materials and parts of similar complexity have run on that machine, you have a baseline. It won't be as tight as a prediction on a part you've run 40 times, but it beats pulling a number out of the air or distracting the team on the floor asking around.
This is the difference between a spreadsheet based on general standards from years ago and an informed estimate built from actual production data. The number might be higher than what you would have quoted before, or it might be lower. Either way, it's closer to what the plant manager will experience when the job hits the floor.
Additionally, knowing your true cost sometimes means realizing you shouldn’t take the job. Just because you can make the part doesn’t mean you can profitably make it at a price the market will bear. Walking away from a low-margin job with confidence beats finding out at month-end that you worked hard to lose 3%.
So what do you actually do with the data?
Whether you have production monitoring or you’re just starting to think about it, the question worth asking is: What do you want the data to do? Knowing that a machine ran for six hours yesterday is useful. Knowing what those six hours cost you, on that job, for that customer, is what changes your margins. The data is sitting in PLC registers, work orders, scheduling databases, and your operators’ heads.
Every completed job is more data, and shops that compete with data are going to win more of the right work.
