Workmanship criteria in machined parts inspection: part 2
Severity classes, boundary samples, and counting rules are what turn a workmanship sheet from decoration into an enforceable standard. This second installment explains how buyers assign defect classes without endless debate, why physical samples beat photographs for finish disputes, how to count visual defects fairly, and how sampling mathematics applies to traits the eye judges rather than the caliper measures. It closes with the feedback loop that keeps workmanship stable across repeat orders instead of resetting the argument every shipment.
Key takeaways
- Fix one severity class per defect trait before the visit, because the class decides whether a finding fails the lot or merely notes it.
- Keep physical boundary samples for every contested trait, since photographs alone cannot settle texture and reflectivity arguments.
- Count defects per piece and per sample under written rules, so neither side can inflate or dilute the numbers after the fact.
- Apply the same acceptance sampling logic used for dimensions to visual traits, as the underlying statistical reference describes. [1]
- Push repeat findings into the supplier's corrective-action system, following the quality management principles that turn one inspection into lasting change. [2]
Part 1 of this series built the workmanship sheet: named defect families, illustrated limits, and a sampling rule both sides accept. Part 2 goes deeper into the machinery that makes the sheet work in practice. Severity classes decide how a single finding moves the verdict. Boundary samples settle the arguments photographs cannot. Counting rules stop inspectors and factories from gaming the numbers.
These details matter because workmanship inspection is inherently judgmental. A diameter is either inside tolerance or outside it; a brushed finish is acceptable until someone decides it is not. The structure in this article narrows that judgment zone until different inspectors, on different days, reach the same verdict on the same lot.
We also cover the longer game: feeding findings back into the supplier's process so the third shipment needs less arguing than the first. Inspection catches defects; only process change prevents them.
Assign severity classes before the visit
Every workmanship trait needs one severity class agreed in advance: critical, major, or minor. Critical means the defect creates danger or destroys function, and a single critical finding fails the lot. Major means the defect impairs use or presentation beyond a stated allowance. Minor means a small imperfection inside an agreed limit that only matters when too many accumulate.
The test for the right class is the worst credible consequence. A sharp burr on a handle the installer grips bare-handed is critical even if it looks small, because the injury risk decides the class, not the size. The same burr profile on a hidden mounting ear is minor. UTS Inspection UTS Inspection applies these pre-agreed classes on site so the verdict follows the sheet instead of the inspector's mood, and factories can predict the outcome of their own process.
Document upgrades explicitly. Safety fasteners, fluid-path parts, and food-contact surfaces promote several defect families one class higher than the default table. Write these promotions into the sheet so the factory prices the order against the real standard instead of discovering it on inspection day.
Boundary samples settle what photos cannot
Counting rules both sides can audit
| Counting question | Recommended rule | Why this rule |
|---|---|---|
| Several scratches on one piece | Count per defect occurrence against the per-piece limit, then the worst piece decides the piece verdict | Stops one bad piece hiding behind sample averages |
| Same defect on many pieces | Each affected piece counts once toward the sample defect total | Reflects how customers experience the lot |
| Cluster in one carton | Counts normally, but note the cluster and pull extra cartons nearby | Clusters signal process pockets worth investigating |
| Reworked pieces | Pieces the factory re-polishes during the visit are flagged and re-checked separately | Prevents showcase polishing from diluting the sample |
| Touch-up paint or filler | Treat concealed rework as its own defect class in the sheet | Hidden corrections fail differently in the field |
Who owns each workmanship decision
Confusion about authority ruins more visits than defects do. The buyer owns the criteria, meaning only the buyer can change a limit or grant a concession during the visit. The inspector owns the verdict against the written criteria and must not invent new rules on the floor. The factory owns the process and the corrective action that follows a fail.
This separation is why concessions need a paper trail. If the buyer accepts a finish variation on this lot, that concession covers this lot number only, in writing, with photographs attached. Verbal dispensations have a way of becoming permanent standards by the next shipment, and the rework loop starts over.
Repeat findings should travel upstream into the supplier's system. The international vocabulary of quality management describes corrective action as eliminating causes rather than sorting symptoms, and buyers can reasonably ask which procedure changed after a repeated defect. [2] Suppliers that answer with a changed deburring station, a new handling rack, or a revised finishing schedule are investing; suppliers that answer with fresh promises are spending your patience.
A short script for the closing meeting
End every visit the same way: read the counts trait by trait, state the verdict against the pre-agreed limits, and agree the next step before anyone leaves the room. Fails need a rework scope, a deadline, and a re-inspection date. Passes with observations need a written note of which traits sit near the limit so the next order starts with attention there.
Sampling gives this meeting its authority. Because the sample was drawn at random under a published acceptance logic, the verdict describes the lot rather than the inspector's luck. [1] Both sides may dislike a fail, but neither side can fairly call it arbitrary.
Keep the standard alive between orders
Workmanship sheets decay. Tooling wears, operators rotate, and the golden sample slowly stops resembling current output. Review the sheet every few shipments: retire limits nobody enforces, tighten traits that keep generating customer complaints, and photograph the new boundary pieces when finishes legitimately evolve.
Track workmanship findings across orders in a simple log, trait by trait. A defect family that appears in three consecutive reports is not bad luck but an uncontrolled process, and the conversation with the factory should shift from sorting to root cause. That shift, from verdicts to prevention, is the point the whole series has been building toward.
Sources
- Wikipedia: Acceptance sampling —— Overview of acceptance sampling including AQL concepts used in product inspection.
- Wikipedia: ISO 9000 —— Overview of the ISO 9000 family of quality management standards.
Frequently asked questions
How do you assign a severity class without endless debate?
Give each defect trait one class that matches its worst credible consequence. A burr on a sealing face is critical because it can cause leaks; the same burr on a hidden bracket edge is minor. If the team argues for ten minutes about a class, the trait needs a boundary sample more than it needs a debate.
Can AQL sampling really apply to visual workmanship traits?
Yes, and recommend it. Basic sampling mathematics, as explained in the standard reference on acceptance sampling, works identically whether the trait is a diameter or a scratch. Fix the sample size and acceptance numbers before the visit, select cartons at random, and judge workmanship findings against the same plan.
How do buyers keep workmanship stable across repeat orders?
Stable workmanship across batches comes from process control, not from sharper eyes on inspection day. The vocabulary and principles of quality management systems describe how documented procedures, corrective action, and management review turn one good shipment into a repeatable result.
What is the correct way to handle rework after a workmanship fail?
Segregate the lot, agree whether rework or 100 percent sorting applies, record the corrective action in writing, and re-inspect the corrected quantity before release. Do not let reworked cartons rejoin the shippable stock without a second check; that is how rejected defects quietly sail.