A precision parts maker: 2,100 paper QC sheets replaced with digital pass/fail
A precision parts manufacturer in South Yorkshire was verifying tolerances on paper sheets filed in a Portakabin. The quality manager could not answer basic questions about the firm's own output.
The reading
Forty-one employees, nine CNC centres, aerospace and defence customers who required full traceability. Every part was inspected against a tolerance specification; the inspection was recorded on a paper sheet; the sheet was filed in a Portakabin behind the main building.
The quality manager, formally responsible for these records, could not answer, within a working day, the question "how many parts in batch 2024-0417 failed first-pass inspection?" He could get there in about three days, by sending a junior inspector into the Portakabin with a torch.
The audit was due in November. The managing director had, in September, instructed us to "make this not be a problem".
Paper in a Portakabin
The inspection itself was not the problem. The inspectors were skilled, the measuring equipment calibrated, the tolerances correctly applied. The problem was that the output of all this competence was a sheet of paper that went into a filing cabinet and effectively vanished.
Audit response, trend analysis, operator feedback, customer queries: all of these were theoretically possible and practically impossible. The firm was running a serious manufacturing operation and a Victorian records office at the same time.
- 2,100 paper QC sheets generated per month
- Portakabin storage with no index beyond date
- Average 3-day retrieval for any specific record
- No trend analysis possible on inspection failures
- Audit preparation estimated at 6 weeks of inspector time
What we built
We gave each inspection station a ruggedised tablet with a digital version of the QC sheet. Measurements were entered against the tolerance specification, which the system had already loaded for that part number. Pass/fail was computed and the sheet closed with the inspector's PIN. The paper sheets stopped being generated at the end of week two.
The technical detail worth noting: we built the tolerance engine as a declarative rule set, so quality engineering could add or amend a specification without a developer. A new part number was onboarded in under an hour. Previously, this had been a two-day exercise involving the quality manager, a draughtsman, and a photocopier.
The stack ran on-premise because the aerospace customer required it: a small Postgres cluster in a locked cabinet, synced nightly to an off-site backup.
The numbers at ninety days
| What | Before | After |
|---|---|---|
| QC sheet retrieval time | 3 days | 4 sec |
| First-pass failure rate (visible) | Unknown | 3.2% baseline |
| New part onboarding | 2 days | 50 min |
| Audit preparation estimate | 6 weeks | 3 days |
| Portakabin filing hours | 18/week | 0/week |
The audit as revealer
The November audit passed without incident, which was the stated objective. The quieter outcome was that the firm, for the first time, had a numerical view of its own quality performance. The 3.2% first-pass failure rate was not a problem to be hidden; it was a baseline to be improved against.
The one-week training test here was passed on day three. A new inspector, hired in month two, needed no induction on the paper system because there was no longer a paper system. That is almost always the goal.