Skip to main content
IE
Innovate / Engineer
Start Free Sprint
WIKI · STAGE 09 · MANUFACTURE

· Quality Control Docs

ACTIVITY 09.10.04 · 6 MIN READ

Quality control docs, measured.

Also called:  QC plan · Inspection protocol · Acceptance criteria · Quality control documentation

The document that defines what gets inspected, the measurable pass or fail thresholds, the sample sizes, and what counts as a reject.

— TL;DR

QC docs turn “looks fine” into numbers a factory can act on. Each check gets a measurable threshold, a sample size, and a clear reject rule. If a criterion can’t be measured, it isn’t a criterion. There’s a worked plan below, and a checklist you can run.

• • •

What quality control docs are

Quality control docs are the written instructions that tell the factory exactly how to decide whether a unit passes or fails. They name each thing to inspect, the measurable threshold it must meet, how many units to check, and what a defect looks like. Without them, “quality” is whatever the person on the line that morning happens to think it means.

The discipline is forcing every check into a number. “The glaze should look even” is an opinion; “no glaze pooling deeper than 0.3mm, no bare patches over 2mm across” is a criterion two different inspectors will read the same way. A QC doc that leans on adjectives will pass batches you would reject and reject batches you would have shipped. Every line in it should survive the question: how would someone with a gauge confirm this?

The three things every check needs

  • A measurable threshold. A dimension with a tolerance, a temperature with a band, a torque figure. Not “tight enough”, not “smooth”.
  • A sample size. You rarely test every unit. State how many out of the batch get checked, and on what plan.
  • A reject rule. What happens when a unit fails, and how many failures condemn the whole batch rather than one part.

The QC plan, worked

The clearest QC doc reads as a table: one row per inspection point, each with its own threshold, sample and reject rule. Here is the plan we wrote for the proofing box, so you can see the shape of a good answer rather than a generic template.

QC plan · the proofing box
Ceramic checksNo cracks or chips visible under task light. Glaze even, no bare patches over 2mm. Rim diameter 198mm ±0.8mm so the lid seats flush with no rock.
Electronics checksPowered functional test: PCB holds 26°C ±0.5°C across a 20-minute soak. Encoder and OLED respond. Heater current within 5% of nominal.
Final-unit checksSafety: earth continuity and insulation per BS EN 61010, no exposed live parts. Finish: wood cladding seated, no visible adhesive, no scratches over 5mm.
Sampling / AQL100% on the safety test (non-negotiable). Cosmetic and dimensional checks on an AQL 2.5 sample plan; for a 200-unit batch that is 32 units inspected.
Reject criteriaAny safety failure scraps the unit. On the AQL 2.5 sample, 3 or more cosmetic or dimensional defects holds the whole batch for sorting.

Five rows, every one with a number behind it. Notice the safety test is 100% and the cosmetic checks are sampled. That split is deliberate: you never gamble on a safety failure reaching a customer, but you can accept a known, small risk on a scratch.

The same plan, written two ways. The weak version reads like a wish; the strong version reads like something a factory can run without phoning you.

✕  “Looks fine” inspection
  • “Check the ceramic looks good.”
  • “Make sure the heater works.”
  • “Reject anything obviously faulty.”
  • “Inspect a few from each batch.”
✓  Measurable criteria + sampling
  • Ceramic: rim 198mm ±0.8mm, no glaze patch over 2mm.
  • Heater: holds 26°C ±0.5°C over a 20-minute soak.
  • Reject: any safety fail scraps it; 3+ defects holds the batch.
  • Sample: AQL 2.5, so 32 units on a 200-unit batch.

Every weak line depends on the inspector’s judgement that day, so two people produce two verdicts. Every strong line is a number a gauge or a tester can settle, which is the only way quality survives the handoff to a factory you don’t stand next to.

How it fits the bigger picture

Quality control docs are activity 09.10.04 in the framework, inside Stage 09 Manufacture. They sit on the specification and the test results behind them, and they feed straight into the assembly plan (09.10.05), which builds the units these criteria then judge. QC defines the bar; assembly is what gets measured against it.

01 02 03 04 05 06 07 08 09 10 Idea Discover Innovate Evaluate Define Design Engineer Develop Manufacture Deliver YOU ARE HERE

What it can do

It gives a factory you may never visit a single, unambiguous standard to build to, and gives you the right to reject a batch on evidence rather than argument. When a supplier disputes a reject, the doc is the referee: the unit either met the threshold or it didn’t.

What it can’t do

It can’t catch a defect it never thought to inspect. A QC doc only tests what you wrote into it, so a failure mode you missed in FMEA sails straight through. It also can’t enforce itself; someone has to actually run the sample and read the gauge.

See the full 10-stage process →

Try it yourself

Take your product and list every point a customer would notice if it were wrong: dimensions, function, safety, finish. For each one, write the measurable threshold, then ask: “could a stranger with the right tool confirm this without calling me?” If not, the line is still an opinion. Decide which checks run on 100% of units and which can be sampled, then write the reject rule that holds a batch.

Want to get the upstream thinking right first? Start the Free Sprint → and the GPT will help you pin down what actually matters to your buyer.

Your QC-docs checklist

Project notes: writing the reject rule

  From the notebook · optional reading

Two factories, two readings of the same word, and the day “even glaze” turned into a number that settled the argument for good.

3 min read · click to open

The ceramic shell and the PCB were made in two different places: the shell in Stoke-on-Trent, the electronics near Manchester. Our first QC draft said the glaze should be “even and free of obvious defects”. I thought that was fine. It was not.

Where the word broke

The first batch back from Stoke-on-Trent had a run of shells with faint glaze pooling near the rim. The pottery thought they were fine, well within trade-normal. We thought roughly a fifth of them looked tired next to a £149 price tag. Neither side was wrong, because “even” meant two different things to two sets of eyes. We had written an opinion and called it a criterion.

So we rewrote it. “No glaze pooling deeper than 0.3mm; no bare patch wider than 2mm; rim 198mm plus or minus 0.8mm so the lid seats with no rock.” I asked the pottery to sign it off as something they could actually hold to before we ran the next batch. They could, and the dispute never happened again, because there was nothing left to interpret.

The sampling decision

The other thing we got right, more by nerves than wisdom, was splitting the plan. The earth-continuity and insulation test ran on 100% of units, no exceptions, because a safety failure reaching a kitchen is the one outcome we could not accept. Everything cosmetic ran on an AQL 2.5 sample, 32 units on a batch of 200. When a sampled batch threw 3 defects, the whole lot got held and hand-sorted. Boring, written down, and it meant a quiet launch instead of a recall.

— Manufacture stage, project notes, 2026

— Next in Manufacture → Assembly plan