The typical situation
Every quote depended on one senior estimator reading the drawing and guessing the cycle time. Quotes took one to two days, and the factory lost orders to faster competitors. When the guess was low, the job ran at a loss; when it was high, the order went elsewhere.
Factories like this already have the answer in their records: every past job has a drawing, a material, an operation list and the actual machine time from the production log.
What the data looked like
Four years of job cards and production logs with actual machine hours, the quotation register, material grades and operation lists. Drawings are PDFs; the useful features (dimensions, tolerances, number of setups) are read from them once, with the estimator checking a sample.
What we built
- A cycle-time model trained on past jobs, giving an estimate with a range and the three most similar past jobs beside it.
- A quote sheet the estimator adjusts and signs, so the person stays accountable for the price.
- A monthly comparison of estimated against actual machine time, which decides when the model is retrained.
- A shadow period of six weeks in which the model estimated every quote but the estimator's figure was used.
Typical outcome
| Measure | Before | After | Period |
|---|---|---|---|
| Time to quote a new drawing | 1 to 2 days | about 1 hour | typical outcome |
| Estimate within ±15% of actual machine time | about 55% of jobs | about 80% of jobs | typical outcome |
What usually trips these projects up
The model is only as good as the production log. Where operators book time to the wrong job card, the first month goes on fixing the booking habit, not on the model.