Product recalls: running the retrieval while the factory keeps making parts

Bounding the population before you announce anything

The scope decision drives every cost that follows, and it is made with incomplete information under time pressure. Work outwards from the failure: which lots share the material, the tooling, the setup, the process deviation or the supplier delivery implicated. Then test the boundary honestly by asking what evidence excludes the adjacent lots — if the answer is that nothing suggests they are affected, that is not exclusion. Recalling too widely wastes money and credibility; recalling too narrowly forces a second announcement, which costs far more because it tells everyone your first boundary was guesswork.

Finding out who actually holds the product

Your dispatch records show the first destination, not the current one. Between you and the affected units there may be a distributor, a converter who built your part into an assembly, a retailer, a service network, or an installer. Each layer needs a separate contact, a list of what they received, and a means of confirming what they still hold versus what has moved on. Expect the trail to break somewhere, and plan the wider notification for that segment rather than pretending the records are complete. Consumer-held product is the hardest tier and the slowest to respond.

Controlling returns so the recall does not create its own defect

Returned units arrive unlabelled, mixed with unaffected stock, sometimes damaged in transit and occasionally with the wrong product entirely. They need a dedicated receiving point, physical segregation from anything sellable, clear status marking on arrival, and a booking record that ties each unit to the notification it answers. The failure that turns a recall into an incident is affected material being restocked and shipped again by someone trying to be helpful. Access to that quarantine area should be restricted and disposition decided by a named authority rather than by whoever needs the space.

Disposition, and why every returned unit is evidence

Before anything is scrapped or reworked, decide what will be examined and record what is found, because returned units are the largest sample of the failure you will ever have. A proportion should be held intact for analysis, particularly units reported as failing, and the examination should test whether the assumed cause explains what actually came back. Recall populations frequently reveal a second problem, or show that most returned items were unaffected, which is important information for the boundary. Only then process the remainder through rework, scrap or controlled destruction with a record of quantities.

Effectiveness, reconciliation and running the plant at the same time

The number that matters at the end is what proportion of the affected quantity was accounted for, and what the plan is for the balance — recovered, confirmed already consumed, or unreachable. That requires reconciliation against the original shipped quantities rather than against how many came back. Meanwhile the factory has to keep producing, usually the same product with the corrective change in place, and the same engineers are needed in both places. Deciding early who is on the recall and who is protecting production prevents the common outcome where both jobs are done badly.

Frequently asked questions

What is the difference between a recall and a stock recovery or field correction?

A stock recovery deals with product still under your control that has not been distributed, so it is an internal quarantine exercise. A field correction repairs, retrofits or re-labels product where it sits rather than bringing it back. A recall retrieves product already in other hands. The operational demands differ sharply: corrections need field capacity and parts kits, recalls need reverse logistics and quarantine space, and the choice is driven by whether the defect can be safely remedied in place.

How do we rehearse a recall before we ever need one?

Run a mock exercise on a real shipped lot. Nominate the lot, then require the team to identify every customer that received it, the quantity each took, and how they would be contacted, all within a stated time. Measure how long it took and what could not be determined. Most plants discover the same gaps: consolidated shipments with no lot record, distributor records that stop at the first destination, and out-of-date contacts.

Who should make the decision to recall?

A defined group with the authority to commit the money and the standing to overrule commercial objections, typically including the quality director, the operations lead and an executive. What matters as much as who is present is that the criteria are agreed beforehand, in calm conditions, so the discussion under pressure is about whether the evidence meets the threshold rather than about where the threshold should sit. Record the reasoning, whichever way it goes.