JPH Cycle Time Estimation: How AI Quotes Production Cells
A Ford F-250 cargo door cell needs to hit 32.4 seconds per cycle to make 55 jobs per hour. A GM Silverado underbody cell might run 38 seconds for 47 JPH. A Stellantis Ram 1500 trim line could be anywhere from 41 to 55 seconds depending on the configuration.
Every Tier-1 automotive tooling RFQ starts with this number. And every quote depends on whether your senior engineer can correctly remember which past job, among the last 200 you've built, is closest to the new geometry, alloy mix, robot count, and conveyor topology.
That memory game is where the millions live. And it is exactly the part of the quoting workflow that AI is now eliminating.
What JPH Cycle Time Estimation Actually Encodes
JPH, Jobs Per Hour is the OEM's hard constraint on the production line you're being asked to build. It is not a target you negotiate. It is the rate the OEM's downstream plant must hit to feed their assembly line at the contracted run-rate.
If Ford says 55 JPH for a cargo door cell, the math is unforgiving:
- 3,600 seconds per hour ÷ 55 JPH = 65.4 seconds available per job
- Of those 65.4 seconds: ~45% is value-add work (welding, fastening, joining)
- The remaining ~55% is robot travel, part transfer, part presence checks, and idle waits
That means the actual robot weld time, gluing dispense time, and fastener-driving time has to fit inside roughly 29-32 seconds. Miss that by 4 seconds and your cell underdelivers. Miss it by 8 seconds and the OEM rejects the line at acceptance.
The estimator who quoted at 32 seconds and built at 36 seconds doesn't just lose margin. The builder may eat the entire program. That's why JPH cycle time estimation is important.
Why JPH Estimation is the Hardest Part of a Tooling Quote
A new RFQ arrives with three things from the OEM:
- A plant layout PDF (24" × 36" or 48" × 36") showing the cell footprint, available column-grid, conveyor entry/exit, and approximate robot positions
- A 50-page requirements document specifying production rate, OEE target, weld counts, fastener counts, ergonomic clearances, and dozens of plant-specific rules
- 2D and sometimes 3D models of the product being built (the actual cargo door, underbody, body-in-white, trim assembly)
What the OEM does not give you:
- Cycle time for any individual operation
- Whether 14 robots or 18 robots are the right count
- Whether to use a single-station serial line or a parallel twin-line
- How much time to budget for inter-station transfer
- What gripper change-over allowances to assume
Every one of those is the supplier's responsibility to figure out. And every one of them is conditioned on what worked on the last similar job and what didn't.
That is why a senior estimator at a 70-year-old tooling builder doesn't quote from scratch. They look at the last F-150 door cell, the last F-250 cargo door cell, the last RAM Heavy Duty door cell. They identify which one has the closest part geometry, weld count, alloy mix, and plant layout constraints. They Frankenstein a baseline. And then they adjust.
The "Frankenstein" Workflow That Currently Eats 5-6 Weeks
In every tooling builder I've talked to, the same workflow shows up:
Week 1: A senior engineer reads the 50-page requirements doc and the layout PDF. They identify which past jobs are "similar enough" to study. They request the historical files from the document control system. They spend 2-3 days reviewing the cells they built 2, 4, or 7 years ago.
Week 2: A junior engineer cross-references the OEM specs (Ford BMS, GM GMW, Stellantis MS standards) cited in the new RFQ against what was used in the historical jobs. They flag deltas: new alloy, new tolerance band, new safety standard.
Week 3: The estimating team Frankensteins. They take 3-4 past cells, identify the closest match for the door panel section, the closest match for the hem flange section, the closest match for the inner panel. They mash them into a baseline cell concept.
Week 4-5: Layout team adjusts the baseline for the new plant's column grid and conveyor topology. They simulate cycle time. They iterate.
Week 6: Pricing pulls labor cost, material cost, and risk reserve. Sales reviews. Quote goes out.
Total: 5-6 weeks per quote. And the senior engineer's time the most expensive engineer in the building is concentrated in Weeks 1-3, where the entire downstream chain depends on whether they correctly identified the right past jobs to study.
Where the Million-Dollar Mistakes Happen
Two failure modes show up consistently in this workflow.
Failure mode 1: Missed cell. The estimating team Frankensteins from three past jobs and forgets that on the F-150 door cell, there was a fourth cell, a hem cell, that the new RFQ also requires but wasn't visible in the simplified summary they were working from. The quote goes out without the hem cell. The customer signs. The supplier has to eat the cost of building the hem cell. Real cost: $1.3 million on one program.
Failure mode 2: Cycle time underestimation. The senior engineer remembered that the F-250 cargo door ran at 32.4 seconds, but forgot that the F-250 used a unique adhesive that cured 2.8 seconds faster than the standard. The new program uses standard adhesive. The cell quotes at 32.4 but actually runs at 35.2. Either the supplier eats the speed-up cost (extra robots, faster conveyors) or the OEM rejects acceptance. Real cost: $4-8 million depending on program size.
Both failures trace to the same root cause: the senior engineer's memory is the system of record for "what was similar to this new RFQ." If they remember wrong, miss a detail, or never saw the original cell, the quote is wrong.
What AI Actually Changes
The shift is not "AI replaces the senior engineer." It is "AI gives the senior engineer a 90% complete first pass to verify, instead of a blank page to build from."
Specifically, AI handles three parts of the JPH workflow that currently eat senior engineer time:
1. Past-cell similarity matching. Given the new RFQ's layout PDF, requirements doc, and 2D/3D models, the AI identifies the closest 3-5 past cells from the builder's historical job database, ranked by geometry similarity, weld count overlap, alloy match, JPH target proximity, and plant layout congruence. The senior engineer reviews the matches and confirms or overrides in 30 minutes instead of 3 days.
2. Spec callout extraction. The AI extracts every cited OEM specification (BMS, GMW, MS, AIPS, AIMS) from the 50-page requirements document, identifies which are present in the historical jobs, and flags the deltas. The junior engineer reviews the delta list they no longer have to scan 50 pages line-by-line.
3. First-pass cycle time estimate. Based on the matched past cells, the AI proposes a cycle time decomposition: this many robots, this many weld seconds per station, this much transfer time, this much idle budget. The senior engineer adjusts the proposal rather than building it from scratch.
The compression is dramatic. Weeks 1-3 of senior engineer time collapses from ~15 days to ~3 days of senior engineer time, with AI doing the document reading, similarity ranking, and first-pass synthesis. The Frankenstein step happens in hours, not weeks.
The Accuracy Standard That Matters
A 50% accurate AI doesn't help. The senior engineer would have to redo the work anyway.
An 80% accurate AI is the threshold where it becomes genuinely useful. The senior engineer verifies the 80% that's right and fixes the 20% that's wrong, saving net effort. This is exactly the target an estimator at a 70-year-old Tier-1 tooling builder named when we ran the numbers with him: "if it can give me 80% of what our outcome is, and our professionals tweak it to get it to where it needs to be, I can potentially get to the same result in a week or days versus a month and a half."
At 80% accuracy on past-cell similarity matching, the quote cycle compresses from 5-6 weeks to 1-2 weeks. The senior engineer's time stays on the high-judgment 20%, the alloy-specific cure time exceptions, the plant-specific column-grid constraints, the customer-specific acceptance criteria. The mechanical 80% (which past jobs are similar, what specs are cited, what the baseline cycle decomposition looks like) gets handled before the senior engineer opens the file.
What This Looks Like in Practice
A new Stellantis RAM 1500 trim cell RFQ lands. Within an hour:
- AI extracts the JPH target (47 JPH from the requirements doc, page 12)
- AI surfaces the 4 closest past cells from the builder's history (2 RAM trim cells, 1 Jeep Wrangler trim cell, 1 RAM HD trim cell)
- AI extracts and compares 14 cited MS standards (12 are matches with past jobs, 2 are deltas, one new welding standard, one updated safety standard)
- AI proposes a baseline cycle decomposition (16 robots, 4 stations, 71.5-second cycle time at 47 JPH target)
The senior engineer opens the file the next morning, reviews the proposal in 90 minutes, confirms 80% of it, overrides the welding standard delta and the inter-station transfer time, and hands the file to the layout team. By end of week 1, the quote is at the pricing review stage that used to take until week 6.
When AI Is Worth Building For
This workflow is worth automating for any tooling builder where:
- Jobs are $5M+ in value (so a single error costs more than a year of automation cost)
- The estimating cycle is 4+ weeks (so compression has compounding effect)
- There are 50+ past jobs to learn from (so similarity matching has signal)
- Senior engineer time is the bottleneck (so freeing it up has direct revenue impact)
That describes essentially every Tier-1 automotive tooling supplier. It also describes specialty steel multi-site groups, aerospace tier-1 contractors, custom machinery builders, and most complex industrial automation shops.
The senior engineer's memory was the system of record for 70 years of accumulated tooling expertise. It is also a single point of failure that costs $1-8M when it slips. The shift to AI-assisted similarity matching is not about replacing that expertise , it is about giving it a verification surface instead of a blank page.
See it on your own past cells. Book a 30-minute demo with us, we'll walk through how AI matches past jobs to a new Ford, GM, or Stellantis RFQ and where it compresses the cycle.