26 October 2026 · 3 min read

Nobody writes down why they decided not to recall

Short answer: Most food businesses document recalls well and non-recalls badly. When you decide a deviation is contained, a complaint is isolated or a batch is safe to release, that decision is usually made in a call, agreed in a chat thread and never written up, because nothing happened. Then a second complaint arrives, or a retailer asks why that batch shipped, and you are reconstructing your own reasoning from memory and timestamps. The record you need is short, what you knew, what you assumed, what evidence you relied on, who decided, and when. Write it at the time, because written later it reads like a defence rather than a decision.

This piece covers what belongs in that record, who owns it, and the parts people leave out.

The decisions that go unrecorded

The recall itself generates paperwork by force. Everything upstream of it does not.

Hold and release. A batch is held on suspicion, tested, released. The result is filed. The reasoning that made the result sufficient usually is not.
No further action on a complaint. One consumer, no pattern, no retained sample. A reasonable call. Also the one you will be asked about if a second arrives.
Scope decisions. You recalled three batches, not nine. Why those three, and what ruled the others out.
Downgrades. Something opened as a potential recall and was reclassified. The reclassification is the decision, not the opening.

What a usable record contains

Not a report. Half a page, written while the detail is fresh.

What you knew at the time, the facts, separated from what you assumed or were waiting on.
The evidence relied on test results, traceability output, retained sample checks, supplier confirmation. Name them specifically.
Who decided, by name and role, not 'the team agreed'.
What would change the decision, which is the line that protects you most. If a second complaint on the same batch reopens it, say so.

Why reconstruction fails

You can rebuild a timeline from emails and messages. You cannot rebuild judgement, the thread shows that someone said 'I think we're fine', not why. Under audit, a retailer reads that as absence of process, and it is hard to argue otherwise. Capturing decisions as they happen, in one place rather than across inboxes, is the only version that holds up.

Who owns it

One named person per incident, usually whoever is running it. Not everyone, because then it is nobody. Agency and temporary cover over a Christmas run makes this sharper, the person who made Tuesday's call may not be in the building on Friday.

Frequently asked questions

Do we need to log decisions where we took no action?

Those are the ones most worth logging. Action leaves its own trace. Inaction leaves nothing.

How detailed does it need to be?

Enough that a competent colleague who was not there could follow the reasoning. Usually a few paragraphs.

Will a regulator or retailer ask for this?

They generally ask how you reached a decision, not just what you decided. Specific expectations vary by jurisdiction and by customer, so check your own.

Should the decision record sit with the quality file or the incident file?

The incident file, with a clear reference from the batch records. Splitting it means neither is complete.

Friday4:30 is the coordination layer for the food and drink business, from product incidents to business and industry issues, first alert to closed-out review.

This article is general guidance, not legal or regulatory advice. Always confirm your specific obligations with your regulatory authorities and legal advisors.