Why a written sorting instruction
When a complaint comes in, the first sort usually starts with a phone call: “check everything for scratches.” Two days later nobody can say which lots were checked, by which method, or why the rejects in the next shipment were missed. A written instruction makes the sort repeatable and gives D3 the evidence a reviewer asks for: what was suspect, where it was, how it was checked and what was found.
This template is generic. It doesn’t replace your procedure or the customer’s rules, such as a named sorting company, controlled-shipping level or label format. Where the customer requires it, agree the instruction with them before sorting starts.
What the form covers
| Section | What goes in it |
|---|---|
| 1. Identification | Instruction number and revision, complaint / 8D number, part and drawing revision, failure mode, who issued and approved it |
| 2. Suspect range | From and to (date, shift, lot, serial), last known good with its evidence, machine / tool / cavity, material lots, quantity |
| 3. Where to sort | Every location: work in progress, finished goods, warehouse, in transit, customer stock, service parts. “None” instead of a blank |
| 4. Method and reject criteria | Characteristic, limit, method, gauge ID, lighting and conditions, OK / NOK photos or limit samples |
| 5. Who sorts | Names or sorting company, who briefed them and when, signatures |
| 6. Handling of results | Marking of good parts, segregation and disposition of rejects, escalation when a reject is found, clean-point marking |
| 7. Results log | Per shift and location: lot, checked, OK, NOK, sorter, remarks |
| 8. Exit criteria | When sorting stops (agreed with the customer) and the D6 reference that justifies stopping |
How to fill it in
- Range first: define first known bad and last known good from evidence (a tool change, a material lot, a measured retained sample). If you can’t prove the boundary, widen it.
- Every location: list WIP, finished goods, warehouse, transit, the customer’s stock and service parts. Write “none” rather than leaving a line empty, so the reader knows it was considered.
- Yes/no criteria: a sorter should not have to judge. Give the limit, the method, the gauge, the lighting and OK / NOK photos or limit samples.
- Check the method: for visual defects, mix a few known-bad parts into the lot and see whether the sorters find them. Write the result down.
- Log per shift: checked, OK and NOK per lot and location. Escalate the first NOK straight away; it may move the range.
- Exit criteria: agree when sorting stops, usually once the permanent action is in place and verified, and reference the D6 evidence.
Writing the 8D around this sort? Get a free Accept / Conditional / Reject advisory on the text, with missing D3 evidence named, before it goes to the customer.
8D Gate is a free pre-send readiness advisory. It doesn't replace the customer's decision. We don't store your 8D body in our database; you can delete anytime. Your text is sent to Anthropic (Claude API) only to run the review; Anthropic doesn't use API data to train its models and by default deletes it within 30 days. Content flagged under its Usage Policy may be kept up to 2 years.
Weak vs stronger entries
Illustrative, fictional examples from a moulded part with a surface defect:
| Weak | Stronger |
|---|---|
| “Check for scratches.” | Reject any scratch on surface A longer than the limit sample L-2, under 1000 lux at 40 cm, 5 seconds per part. |
| “Sort all suspect stock.” | Lots 2231–2240 at the plant (WIP and FG), 2 pallets in transit, customer warehouse lots 2236–2240. |
| “Range: last week.” | From tool change on 12 May, shift 2 (first known bad) to the tool repair on 15 May, shift 1; last known good lot 2230 measured OK. |
| “Sorted, all OK.” | 4,800 checked, 4,761 OK, 39 NOK, 3 sorters on shifts 1–3; method checked with 5 seeded NOK parts, all found. |
The same structure works outside automotive. A food supplier sorts by batch code and best-before date and records metal-detector checks; an electronics supplier sorts by date code and reel and retests on the flying probe. The questions stay the same: what is suspect, where is it, how do you know a part is bad, and who checked.
Sorting is containment, not the fix
A clean sort result shows the suspect stock is under control. It doesn’t show why the defect happened or why it escaped, and it isn’t evidence that the corrective action works. Customers often reject an 8D whose D6 is “no complaints since sorting started”. Keep the sort data in D3, and verify the permanent action separately in D6 with data from after the change. See corrective action vs containment, containment action examples and D6 verification of effectiveness.
Common mistakes
- Suspect range based on a guess, with no last-known-good evidence.
- Customer stock and parts in transit left out.
- “Visual check” with no limit sample, lighting or time per part.
- Results reported as “all OK” with no quantities.
- No exit criteria, so sorting either never ends or stops before the fix is verified.
FAQ
- Is sorting a corrective action?
- No. Sorting is containment (D3): it protects the customer while the cause is found. The corrective action removes the cause (D5), and D6 shows with data that it works. A clean sort result doesn’t prove the permanent fix.
- Who decides the sorting method and when it can stop?
- Your procedure and your customer. Many customers specify the sorting company, the labels or a controlled-shipping level, and want to agree the exit criteria. Send the instruction to them before you start where it is required.
- What does 8D Gate check?
- It gives a pre-send readiness check on the 8D text: an Accept / Conditional / Reject advisory with missing evidence and questions. It isn’t a customer decision. The review tool is v0.1: it reviews English 8D reports, and the interface and report are in English. Templates are also available in German, Czech and Polish.
Free pre-send readiness check
Before the 8D goes to the customer, run the text through the free supplier 8D readiness check.
Upload a PDF, Word (DOCX) or Excel 8D. You get an Accept / Conditional / Reject advisory, a score, section ratings, missing evidence, and questions to close gaps, with a PDF download. It is a pre-send readiness check for your team, not a customer decision.
8D Gate is a free pre-send readiness advisory. It doesn't replace the customer's decision. We don't store your 8D body in our database; you can delete anytime. Your text is sent to Anthropic (Claude API) only to run the review; Anthropic doesn't use API data to train its models and by default deletes it within 30 days. Content flagged under its Usage Policy may be kept up to 2 years.
Get the free 8D / D4 checklist
Keep it next to the sorting instruction for the rest of the 8D. It covers D1–D8 and goes deeper on root cause, escape and verification.
Get the free 8D / D4 checklist