Every job is different, which is the problem
Print, signage and fabrication are job-shop businesses. Each order differs in material, size, finish, quantity and complexity, and estimating it depends on whoever is quoting.
That creates inconsistency: the same job quoted differently by two estimators, and systematic under-quoting of certain job types that nobody notices because profitability is measured at month level rather than job level.
Predicting the real cost of a job
- Setup time, which dominates on short runs and is frequently underestimated
- Run time, usually the best-understood component
- Material usage including waste, which varies with nesting and size
- Finishing and hand-work, the most variable and most under-quoted element
- Rework probability for that job type and customer
- Machine choice, where the same job runs differently on different equipment
Finishing is where quotes most often go wrong. It is manual, variable, and hard to estimate, and a model trained on recorded actuals captures patterns that individual experience does not.
Job-level profitability first
Before any prediction, analysing recorded job costs against quoted prices usually reveals clear patterns - particular job types, sizes, customers or finishes consistently losing money.
| Pattern | Typical cause |
|---|---|
| Short runs unprofitable | Setup time under-recovered |
| One customer consistently unprofitable | Scope creep, repeated amends |
| Certain finishes lose money | Hand-work underestimated |
| Rush jobs unprofitable | Disruption cost not charged |
The rush job row is worth testing specifically. Many shops charge a premium that does not cover the disruption to everything else in the schedule, and the true cost only appears when other jobs run late.
Scheduling and material waste
Production scheduling in a job shop is an optimisation problem - sequencing to minimise changeovers while meeting due dates. Prediction contributes the durations it needs.
Material nesting - fitting jobs onto sheets to minimise waste - is also optimisation, and combining jobs across orders where material and finish match can reduce waste substantially. The constraint is usually that jobs must be scheduled together to benefit, which links back to the scheduling decision.
Start by recording actuals properly
All of this depends on knowing what jobs actually consumed. Many shops record quoted values and not actuals, or record them so inconsistently that analysis is impossible.
Improving that capture is the enabling step. Even approximate actual time and material per job, recorded consistently, becomes valuable within months and pays for itself in quoting accuracy alone.
If you only record what you quoted, you will never find out which jobs lose money.