πŸƒ DryFood KB knowledge base
/

Compliance

Recall and incident management

Nobody plans a recall, which is exactly why you must. The businesses that survive one are the ones that could trace the batch in an hour, had the contacts on a card, and wrote the notice from a template instead of from panic. This page is the plan β€” write yours, then prove it with the mock-recall exercise at the end.

The recall decision

The decision is a risk classification, not a feeling. Most jurisdictions use a two- or three-tier system; the questions below mirror them.

LevelSituationTypical action
Class I / highestReasonable probability of serious health consequences or death β€” undeclared major allergen, pathogen (Salmonella, Listeria, botulinum toxin), glass/metal in packImmediate public recall; notify authority; media statement
Class II / mediumTemporary or medically minor harm possible β€” undeclared sulfite at moderate level, mold-adjacent off-odour, mislabelled varietyRecall or market withdrawal per authority guidance
Class III / lowestUnlikely to cause harm β€” labelling error without safety impact, weight shortfallsMarket withdrawal or correction
Not a recallInternal quality failure caught before shippingRe-grade or destroy; document as internal nonconformity
β›”
When in doubt, act up a class

No business has ever been ruined by recalling product that turned out fine. Businesses and people have been ruined β€” and hurt β€” by the opposite. The cost asymmetry is total.

Roles and contacts β€” write this down now

RoleOwnsFill in for your operation
Incident managerDecision authority; single spokesperson
Traceability leadPulls batch records; quantity, customers, destinations
CommunicationsCustomer notice, public statement, social channels
OperationsHalts production, quarantines stock, arranges returns
Authority liaisonNotifies regulator; manages inspection follow-up

In a one-person business you hold all five roles β€” write the card anyway, and add the external numbers: local food authority emergency line, lab, lawyer, insurer, largest customer's quality contact.

The first 24 hours

  1. Hour 0–1: stop and confirm Halt shipments of the suspect batch. Pull a retain sample and the batch record. Decide the class (above).
  2. Hour 1–2: quantify Traceability lead answers: how much product, which lots, where is it now β€” warehouse, customers, sold on. One step back (supplier), one step forward (customers).
  3. Hour 2–4: notify Authority notification per your jurisdiction's rules (in the EU via the local authority into RASFF; in the US, FDA recall notification; UK, FSA incident team). Draft the customer notice from the template.
  4. Hour 4–24: contact every affected customer Directly, with: product, lot, dates, hazard, exact instruction (do not eat; return/dispose; refund route), and a contact for questions. Log each contact.
  5. Day 1: public notice if needed Class I almost always needs a public statement β€” plain language, hazard, action, apology, contact details.
Every notice must stateWhy
Product name, size, lot/date codes affectedCustomers must identify their packs precisely
The specific hazardEnables informed risk decisions (allergen sufferers especially)
The exact action: do not eat; return or dispose; howAmbiguity is what spreads harm
Refund/replacement route and timescaleReduces friction, accelerates returns
A reachable human contactTrust is the recovery currency

Root-cause analysis after the fire is out

  1. Preserve evidence Retain samples, batch records, logs, cleaning records, calibration records β€” freeze the paper trail literally and figuratively.
  2. Ask "why" five times e.g. Mould arrived in retail packs β†’ jar moisture high β†’ conditioning skipped β†’ batch rushed for market deadline β†’ no CCP check on conditioning completion β†’ corrective action: add a signed conditioning-release step to the batch record.
  3. Fix the system, not the symptom If the fix is "be more careful", it is not a fix.
  4. Verify the fix β€” next batches measured against the failure mode; record the verification.
  5. Update the SOP and retrain anyone it affects. An incident that changes nothing repeats.

The mock recall exercise

A mock recall is a drill: pick a real past batch and time how fast you can answer the five questions below, using only your records. Do it twice a year. Most first attempts fail on questions 4 and 5 β€” that is the point of practising.

  1. Pick a lot code at random from a real shipment (e.g. L2026-07-11-B, apple rings, 22 packs).
  2. One step back β€” which raw-material lots went into it? Supplier, dates, quantities. Time target: 15 minutes.
  3. One step forward β€” who received the finished packs? Markets, direct customers, online orders. Time target: 30 minutes.
  4. Mass balance β€” kg in, kg packed, kg sold, kg still held: do the numbers reconcile to within normal drying yield? A mismatch is an unaccounted lot β€” find it.
  5. Simulate the notice β€” draft the customer/public notice from the template using only record data. If you cannot write it in 20 minutes, your records are thin.
  6. Debrief honestly β€” log the gaps found, assign fixes with dates, and re-drill after fixes. Score: questions answered within target = pass.
🧾
The traceability dividend

Operations with tight batch records can recall one lot precisely. Operations with loose records must recall everything β€” the whole product line, months of sales. Records are not bureaucracy; they are the size of the blast radius.

Recovering after a recall

  • Close the loop with the authority β€” submit the effectiveness check: what was recovered, disposition of returns, verification that the fix works.
  • Handle returns β€” segregate, document, destroy or denature; returned product never re-enters sale.
  • Tell the truth once, completely β€” customers forgive incidents; they do not forgive discovering you under-communicated.
  • Consider a relaunch gesture β€” improved packaging, a cleaner label, third-party testing published. Turning the fix into a visible feature is the strongest recovery there is.