Tangle Research
On-Time Delivery in Metal Fabrication: Closing the Gap to Top Quartile
FMA reports an industry-average on-time delivery rate of 84% for metal fabricators. Top-quartile shops exceed 90%. An 8-point OTD uplift on a $5M shop reduces late-cost exposure by $20,000/year and removes (uncounted) customer-retention risk. Tangle's deployments deliver 27% more on-time deliveries.
1. Why on-time delivery slides
OTD does not collapse for the reasons shop owners usually name. The press brake breaking down, the operator calling in sick, the steel arriving late — these happen, but they are not the structural cause of low OTD.
The structural cause is that the schedule gets built once, on a wall-board or in Excel, and then immediately starts diverging from reality. A job runs longer than estimated. A rush job lands midday. A part fails inspection. An operator's day-shift is reassigned. None of these are unusual. All of them are invisible to the schedule until a supervisor notices, manually re-plans, and tells someone.
By Tuesday afternoon the plan from Monday is fiction, and the team is operating on memory and judgment. That works for a few jobs but breaks down across a full schedule. Late jobs accumulate not because any single thing went catastrophically wrong but because the schedule never caught up.
2. What AI-assisted scheduling does differently
AI-assisted scheduling does three things wall-boards and spreadsheets cannot.
It absorbs real-time signals. When a machine reports it is running slow, when a job finishes early, when a part fails inspection, the schedule updates automatically rather than waiting for a supervisor to redo the plan.
It surfaces conflicts before they happen. A scheduler human cannot hold a hundred jobs in their head and see that two of them will need the same machine on Thursday afternoon. The schedule can.
It optimizes against capacity, not against the wall-board. Sequencing decisions made by an algorithm that knows actual machine cycle times and changeover costs tend to fit more work through the same capacity than sequencing decisions made by judgment.
3. The math
Annual cost of late = annual revenue × (1 − OTD rate) × late-cost factor. The late-cost factor captures expedite freight, contractual penalties, and the labor cost of re-planning and chasing. This model uses 5%, at the low end of published estimates of 5 to 12%.
| Scenario | OTD uplift | OTD rate | Annual cost of late | Annual saving |
|---|---|---|---|---|
| Baseline | — | 75% | $62,500 | — |
| Conservative | +5 pts | 80% | $50,000 | $12,500 |
| Realistic | +8 pts | 83% | $42,500 | $20,000 |
4. The customer-retention effect this model does not count
Late delivery is consistently cited as the top reason customers leave job shops in industry trade-publication surveys. A single lost long-term customer can dwarf the entire late-cost calculation above. This article deliberately does not include that value in the headline number, because the dollar value of a retained customer cannot be defensibly extracted from the inputs to the model.
It should still be part of the conversation. The operations leader who recovers OTD also recovers customer trust. The numerical case for AI-assisted scheduling is real. The strategic case is bigger.
The schedule is fiction by Tuesday. Not because anything dramatic broke. Because nothing was ever fast enough to keep up.
Run the model with your own numbers
Three to five minutes. Five inputs. Same framework, applied to your shop.
Open the ROI calculator· Frequently asked questions
What is the average on-time delivery rate for metal fabricators?
The FMA Financial Ratios & Operational Benchmarking Survey reports an industry-average OTD rate of 84% for US metal fabricators. Top-quartile shops exceed 90%.
How should on-time delivery be measured in a custom job shop?
Against the original promise date — not the most recently rescheduled date. Shops that measure against the latest reschedule produce inflated OTD numbers that hide the real customer-experience problem. Honest measurement is against the original promise.
What is a reasonable late-cost factor to use in a financial model?
Published estimates range 5 to 12% of late job value, capturing expedite freight, contractual penalties, premium labor, and the administrative cost of replanning. This model uses 5%, at the conservative end.
Does AI scheduling work in a true mixed-mode job shop?
Yes. AI-assisted scheduling matters more in mixed-mode shops than in repetitive ones, precisely because the variation is what overwhelms wall-board and spreadsheet planning. The harder the planning problem, the larger the lift from automating it.
· Sources
- FMA Financial Ratios & Operational Benchmarking Survey — Industry-average OTD benchmark of 84%.
- The Fabricator — 12 metrics for metal fabrication success — OTD and customer-retention framing.
- Eziil — Metal fabrication performance KPIs — Improvement targets and measurement practice.
- Tangle deployment data — 27% more on-time deliveries across customer deployments.