Quality Holds & Traceability: Recall One Batch, Not a Warehouse
When the lab result fails or KEBS calls, the question is "which batches, where, how fast?" — quality holds that freeze suspect stock and traceability that recalls one batch instead of a warehouse.
Every processor eventually gets the call. A customer found something in a packet. A lab result came back wrong. A KEBS sampling visit wants records. What happens next depends entirely on infrastructure built long before the call: if stock moves through the system in batches with holds and a traceable chain, the incident is an afternoon of precise action. If it doesn't, the incident is a total recall, priced in brand.
Quality holds: undecided stock does not move
A quality hold is a status that makes stock untouchable — not issuable to production, not sellable, not transferable — until someone with authority releases it. The four places it belongs:
- Incoming materials pending inspection — the maize at the weighbridge, the packaging delivery, the chemical drum: received into hold, released after checks, rejected back to the supplier with the reason on record.
- Production output awaiting results — batches that need lab confirmation (micro, moisture, aflatoxin) sit in hold; sales cannot promise what QC has not released.
- Customer returns — returned goods enter hold for assessment: restock, rework, or destroy — decided and recorded, never quietly reshelved.
- Suspect stock during an investigation — the moment a complaint implicates a batch, the hold freezes whatever remains, everywhere, in one action.
The hold must be systemic, not procedural
A red sticker and a "do not use" shelf work until the night shift runs short of materials. A hold that the system enforces — the issue screen simply will not release held stock — works at 2am too. Culture backs the control; the control cannot be only culture.
Traceability: the chain in both directions
| Direction | The question | When you need it |
|---|---|---|
| Backward (finished → source) | This packet came from which run, which materials, which suppliers? | Customer complaint, failed retained sample |
| Forward (source → finished) | This supplier's delivery went into which batches, sold to whom? | Supplier notifies a problem; a material batch fails late |
Both directions ride on the same records: batch-tracked receiving, production runs that consume material batches and produce output batches, and dispatches that record which batches went to which customer. None of it is extra work at incident time — it is the ordinary batch loop, captured while it happens. The test of a traceability system is speed: batch-to-customer in minutes is a controlled incident; in days, it is a news story.
The recall drill
Run this as an exercise before it runs you
- Pick a finished batch at random. Trace its material batches and suppliers — time it.
- Trace forward: every customer who received any of that batch, with quantities.
- Place a hold on the remaining stock of that batch across all locations — one action, verified.
- Produce the paper: receiving records, QC releases, dispatch documents for the batch.
- Debrief: anything that took longer than 30 minutes is the gap to close.
What regulators and buyers actually ask for
- KEBS and county public-health inspections want batch records, QC results tied to batches, and disposal evidence for rejects — a system query, or a very long day of files.
- Supermarket chains and institutional buyers increasingly audit traceability before listing suppliers — a working recall drill is a sales asset, not just insurance.
- Export markets make it existential: no chain, no certificate, no shipment.
Be ready for the call
Holds that freeze stock everywhere in one action, and batch-to-customer tracing in minutes — built on the production records you already keep.
See holds & traceability in AWRAFrequently asked questions
Who should have authority to release a hold?
Someone whose role is quality, not output — the QC lead or a designated manager, never the production supervisor whose numbers the hold is hurting. Releases are logged with name and reason; that log is exactly what an auditor reads first.
How granular should batches be?
Granular enough that recalling one batch is survivable: a day's run for most SME processors, a shift or a silo-lot where risk is higher. The finer the batch, the smaller the recall — balanced against the recording effort your line can sustain honestly.
What do we do with stock that fails and cannot be reworked?
Destroy it documented: quantity, batch, reason, method, witness — with the write-off in the system and disposal evidence kept. Failed stock that "disappears" instead of being destroyed has a way of reappearing in the market with your name on it.
Does traceability require barcodes and scanners?
No — it requires batch identity captured at receiving, production, and dispatch. Scanners make capture faster and less error-prone at volume, but a disciplined SME line runs full traceability with batch numbers on paperwork keyed into the system. Start with the discipline; add hardware when volume justifies it.
