A delivery arrives, a few spools are wrong, and the question is whether to reject the batch or pick out the bad ones. Most people pick out the bad ones because rejecting a whole delivery feels disproportionate. Sometimes it is the right call and sometimes it is avoiding a difficult conversation.
When individual retirement is right
- Sporadic damage with no pattern — a few spools got knocked about in transit
- Failures that are visible and unambiguous, so you can find them reliably
- The defect is detectable by the person who will next handle the spool
- No safety implication and no risk to product
This is the right default. It is what most situations warrant.
When the whole batch should go
- A systematic dimensional problem. If the bore is wrong, it is wrong on all of them, and inspecting every one costs more than they are worth.
- A material problem. Wrong grade, wrong colour, contamination. You cannot inspect for it.
- A defect you cannot reliably detect. If you cannot find the bad ones, you do not have a bad-ones policy — you have a hope.
- A safety implication. Anything where failure could injure someone.
- The failure rate is high enough that inspection costs approach replacement costs.
That third point is the useful test. Ask: can we actually identify which ones are affected? If the honest answer is no, batch rejection is the only real option.
The economics
Worth actually doing rather than estimating:
- Cost of inspecting every unit — time, per unit, multiplied
- Cost of a failure that gets through — which is not the cost of a spool, it is the cost of the product and the disruption
- Probability that inspection misses one — rarely zero, and higher than people assume for anything visual
- Cost of replacement including the delay
Once you include the cost of a failure getting through, batch rejection looks proportionate much more often. A spool is cheap; a line stoppage is not.
How to decide quickly
Three questions, in order:
- Is the defect detectable and by whom?
- What happens if one gets through?
- What proportion is affected?
If it is undetectable and the consequence is significant, reject the batch regardless of proportion. If it is detectable and the consequence is trivial, pick them out regardless of proportion.
Making the call with a supplier
Things that make this go better:
- Raise it immediately with photographs and numbers, not as a complaint but as a problem to solve
- State what you found and what you propose rather than asking what they want to do
- Keep the affected units until it is resolved — discarded evidence ends the conversation
- Reference the quality agreement if you have one, which is why the quality agreement article exists
Avoiding the overreaction
Batch rejection is sometimes used to avoid thinking. Before rejecting:
- Check your own measurement. A batch that is out of spec is occasionally a measurement that is out of practice.
- Check whether it actually matters. A dimension outside tolerance that affects nothing is a commercial conversation, not a rejection.
- Check whether you can sort it with a go/no-go gauge in seconds per unit.
A simple gauge often converts a batch rejection into a ten-minute sorting job, which is better for everyone.
Afterwards
Worth doing regardless of outcome: record what the defect was, how many, and what you did. Two reasons — it supports the conversation about preventing recurrence, and it is the input to the failure library if the defect is one you will see again.
Our position
If a delivery of our hollow board spools is wrong, we would rather you told us immediately with photographs and measurements than sorted it quietly and said nothing next time. We will say when we think a batch rejection is proportionate and when we think sorting is sufficient — and we will ask you to check your measurement first, because occasionally that is the answer.