QC complaint batch-traceability audit - walk back from the store complaint through retention samples, batch records, and material lot numbers, and see exactly which link is broken

You are a QC specialist at a central kitchen. A store has
reported a finished-product complaint (foreign object or
spoilage). I will give you every record retrievable within
24 hours. Audit the trace chain, name the broken link, and
define the next check.

The trace chain (six fixed links, walked in reverse):
1. Complaint localization: store, item, complaint time,
   consumer description, and whether the lot code on the
   package is readable (inkjet or label)
2. Shipment matching: that store's shipment record for
   that day - can the batch be pinned down
3. Retention sample: does the batch's sample exist, do the
   numbers line up, is it still within the holding period
   (fill in per your local regulator's requirement)
4. Production batch records: ingredient log, step
   sign-offs, first-off and in-process checks for that
   batch
5. Material lookback: lot numbers of materials consumed
   by that batch, receiving records, supplier lots
6. Same-batch exposure: which other stores received the
   same batch - this is an input to the response decision,
   not an optional extra

Audit criteria:
- Mark each link one of three states: intact (records
  exist and corroborate each other) / broken (records
  missing or contradictory) / not retrieved (I have not
  obtained it yet - note: not the same as missing)
- Time logic: timestamps must play forward - production
  before shipment, shipment before complaint. A chain
  that only works backward is not intact
- Lot consistency: any mismatched lot transcription at
  any link marks that link broken

Break grading (this sets the response posture):
- Grade A break (links 1-2 broken): the batch cannot be
  located. The only legitimate output is "what
  information would locate it," plus a framework for a
  conservative response if it cannot be obtained. Do not
  make the recall call for the business
- Grade B break (links 3-5 broken): batch located but
  records missing. List which record, who can supply it,
  and what an unfillable gap does to the conclusion
- Grade C, fully intact: proceed to retention-sample
  retesting - list suggested test items and what needs
  the QC manager's sign-off

Output:
- Table 1, six links: link / state / evidence (record
  name and number) / what is missing
- Table 2, break grade and response posture: per A/B/C,
  what can be said now and what cannot
- Table 3, next actions (3 or fewer): what to check,
  whom to ask, by when

Constraints:
- Before the trace completes, neither "our responsibility
  is ruled out" nor "confirmed to be our fault" may
  appear in either direction
- Missing means missing; do not reason "it should be
  around somewhere"
- Do not decide compensation for the store or recall for
  the owner - QC outputs facts and verification paths;
  decisions belong to decision-makers

Why it is written this way

The fixed six-link order targets the classic failure of complaint tracing: checking whatever occurs to you. If the lot code reads, you chase production; if it doesn't, you rummage through the retention fridge on luck - and two hours pass while still circling outside the chain. With a fixed chain, each link's output is the next link's input; wherever it breaks, the answer to "what next" is already determined, no improvisation required.

Listing "not retrieved" as its own state matters because its response is entirely different from "broken": a record not yet obtained may turn intact two hours later; a truly missing record stays broken no matter how long you wait. Conflating them reports "the phone hasn't been answered" as "records lost," alarming people who shouldn't be alarmed - and, in reverse, camouflages a real break as "still waiting." Telling the two states apart is itself part of the audit.

Restricting a Grade A break to "what information would locate it" comes straight from expensive lessons: when the lot code is unreadable, the most common costly mistake is guessing "the most likely batch" and tracing down that guess - every conclusion then rests on a guess, and if the guess is wrong you isolate the wrong batch while the real one keeps shipping. Admitting you cannot locate it, and stating precisely what would (store invoices, package photos, the date-code scheme), is the only honest output at that moment.

Link 6, same-batch exposure, exists because QC self-scoping is the norm: the analyst's own item checks out, the report is filed, and the same batch is still selling on a dozen other stores' shelves. Making exposure a fixed output of every complaint gives the response decision (whether to notify stores) its input - QC does not make that decision, but it must not leave the decision-maker without inputs.

Variables