Rework management: deciding what gets fixed, what gets scrapped, and what the fixing costs

The disposition decision and who is allowed to make it

Four outcomes exist for nonconforming product: rework to specification, repair to a usable but different condition, accept as-is under a documented concession, or reject. The choice depends on whether the defect is correctable without damaging other features, whether the customer or the regulator permits a concession, and whether the arithmetic favours fixing over remaking. Authority must sit with quality rather than production, because the person under schedule pressure is the wrong person to decide that a marginal part is acceptable. Where a customer specification governs, the concession route needs their written agreement before the part ships, not after.

Rework needs a route and a record, not a corner of the floor

Reworked parts pass through real operations consuming real capacity, and if those operations are not on an order, the labour and material vanish into overhead and the parts lose their history. Raise rework as its own order or as recorded additional operations against the original, define the sequence of steps, and specify the re-inspection required. This is also what stops parts circulating: without a route, a batch gets partially fixed, set down, picked up by a different shift and worked on again. The physical control matters as much as the paperwork — segregated, labelled, and not stored where good stock is.

Spare rework capacity conceals the process that created it

A dedicated rework cell staffed by experienced people quietly makes the defect rate tolerable. The line ships on time, the customer sees acceptable quality, and the process problem never becomes urgent. This is why rework hours should be reported as a loss alongside scrap rather than absorbed as normal labour. A useful test is to ask what would happen if the rework bench were removed for a fortnight: if the answer is that shipments would stop, the plant is buying its quality with hidden labour and has been doing so long enough that nobody remembers the underlying cause.

Proving the fix actually worked

Reworked product needs verification at least as rigorous as first-time product, and sometimes more, because rework can damage features that were previously fine. Straightening a component may relieve stress somewhere unintended; re-soldering may affect a neighbouring joint; recoating may alter dimensions. Define what gets re-inspected — not merely the corrected characteristic — and record the result against the part or lot. In regulated production the reworked item's record has to show what was done, by whom and to what instruction, because a subsequent field failure will be investigated by reading exactly that record.

Costing rework so the decision is not sentimental

The instinct is always to save the part, especially when material is expensive or the customer is waiting. An honest comparison includes the labour, the re-inspection, the capacity used at operations that could have been making new product, the delivery risk from parts sitting in a rework queue, and the residual risk that a repaired item behaves differently in service. Against that sits the cost of remaking and the lead time to do so. Plants that record rework hours properly often find that some routine repairs cost more than the replacement part, and had been continuing for years on the assumption that fixing must be cheaper.

Frequently asked questions

Can rework be done by the same operator who made the defect?

Often yes, and it can be the fastest way to close the feedback loop, since the person sees the consequence directly. Two conditions apply: the rework must follow a defined instruction rather than improvisation, and the verification must be done by someone else. Where the defect indicates a training or method gap, having the operator correct it under supervision is genuinely useful. Where it indicates a machine or material problem, sending it back to the same person changes nothing.

How should reworked units be counted in production figures?

Report them separately from first-time-good output. Counting them as normal production makes a shift look productive while the plant repeats work it has already paid for, and it erases the gap between first-pass and final yield that shows how much quality is being bought with labour. Show good output, rework hours and scrap side by side. Managers then see that a shift which hit its number using a large block of rework did not perform as well as one that did not.

When is a concession the right answer rather than rework?

When the deviation genuinely does not affect fitness for the purpose, the customer or regulator agrees in writing, and the concession is limited to a defined quantity or lot rather than left open. It is the right call for cosmetic deviations on hidden features or for dimensions outside a drawing tolerance that is tighter than function requires. It becomes a problem when concessions become routine for the same characteristic, which is a signal that the specification should be formally changed rather than repeatedly waived.