Most traceability exercises are designed to succeed.
The batch chosen is recent. It used a single lot of each ingredient. It is still in the warehouse. The exercise is run by the quality manager, who knows where every record lives, on a quiet afternoon, with no time pressure. Twenty minutes later the result is 100%, the form is signed, and the file grows by one page.
Nothing has been tested. What has been demonstrated is that a well-informed person working under ideal conditions can retrieve records they already knew the location of. A real recall will not resemble that in any respect.
What the exercise is actually for
Traceability is a baseline legal requirement in most jurisdictions — one step forward and one step back — and certification schemes typically require documented traceability and mock recall exercises at a defined minimum frequency, often with a specified completion time.
But the compliance requirement is the floor. The purpose of the exercise is to find out where your traceability breaks before a real incident finds out for you, at 11pm on a Friday, with a customer on the phone and the person who normally does this on annual leave.
An exercise that cannot fail tells you nothing about that scenario.
Choose the batch adversarially
The single highest-leverage change is how the test subject is selected. Pick for difficulty, not convenience:
- A batch containing rework. Rework is the most common point at which lot identity is lost. If your system can trace a batch that absorbed rework from two earlier runs, it can probably trace anything.
- A batch that consumed two or more lots of the same ingredient mid-run. Systems that assume one lot per batch break here, and many do.
- A batch split across a shift change, where the handover of record-keeping is tested alongside the data.
- An ingredient common to many finished products, tested forward. This is the mass balance scenario and the one that most closely mirrors a supplier-driven recall.
- Something six to nine months old, close to the edges of your record retention and archiving arrangements, where records may have moved to storage or to a superseded system.
- A product that went to multiple customers through multiple distribution routes.
Run it without the usual person
If the exercise is always run by the person who designed the system, it tests that person, not the system.
Run at least one exercise per year with that person deliberately unavailable. Have someone competent but not expert follow the documented procedure. If they cannot complete it using the written instruction alone, the finding is that your traceability depends on an individual — which is precisely the dependency a real incident is most likely to expose.
Run it unannounced, and ideally at an awkward time. A traceability system that works only during office hours is not a recall capability.
Mass balance is where exercises actually fail
Most sites can produce a chain of records linking an ingredient lot to finished product. Far fewer can reconcile the quantities.
Mass balance asks: take the full quantity of ingredient lot X received. Now account for all of it. How much went into each production run, how much remains in stock, how much was discarded as waste, how much is in rework, how much left as sample, and how much is unaccounted for?
The target is normally 100% ± a small tolerance defined by your own process. Common reasons sites fall short:
- Waste and giveaway not recorded against a lot
- Rework quantities tracked loosely or not at all
- Partial bags or containers returned to stock without lot identity
- Samples taken for QA or customers, never recorded
- Line flush and start-up product not accounted for
Trace without balance is a partial result. If you can identify which products contain the ingredient but cannot say how much of the lot went where, a recall's scope cannot be bounded — which in practice means recalling more than necessary.
Test forward and backward, and test all the way out
Backward from finished product: which lots of which ingredients, which production records, which equipment, which staff, which packaging and label versions.
Forward from an ingredient lot: every finished batch it entered, and every place those batches went.
The forward direction is where exercises are most often truncated. "Dispatched to Customer A" is not a completed trace. The question a recall asks is who physically received it — which delivery, which date, which quantity, to which site or distribution centre. If your trace ends at the dispatch door, your recall will stall there too.
Where your product goes on to further distribution, test whether you can obtain onward distribution data from your customer, and how long that takes. That dependency is part of your recall timeline whether you have tested it or not.
Test against the clock, and test the contact chain
Customer specifications frequently require traceability within a defined window — commonly two to four hours. Run the exercise against a timer and record the actual elapsed time, including the time taken to retrieve archived or off-site records.
Then test the part almost universally skipped: the contact chain. Can you actually reach the people you would need to?
- Are the customer emergency contacts current, or three staff changes out of date?
- Do you have after-hours contacts, or only switchboard numbers?
- Who contacts the regulator, and do they know they hold that responsibility?
- Is the recall team contactable outside working hours, and is there a defined deputy for each role?
A perfect data trace and an out-of-date contact list still produces a failed recall. Verifying two or three contacts at each exercise keeps the list alive at negligible cost.
An exercise with no findings is a design problem
If your traceability exercises consistently return 100% with no observations, the most likely explanation is not that your system is flawless. It is that the exercises are designed not to find anything.
Record what was difficult, even where the result was ultimately achieved: records that took longer than expected to locate, a step that required a phone call to one specific person, an ambiguity in the procedure, a quantity that needed reconciling by hand. Those observations are the output. The percentage is just the headline.
Raise the genuine shortfalls as nonconformances and run them through your corrective action process. A traceability exercise that identifies a weakness and closes it is worth more to an auditor — and to you — than five consecutive clean results.
Practical takeaways
- Select batches for difficulty: rework, multiple ingredient lots, shift splits, multi-customer distribution, older stock.
- Run one exercise a year without the person who normally runs it, using only the written procedure.
- Always include mass balance. Trace without quantity reconciliation cannot bound a recall.
- Trace forward to actual receipt, not to the dispatch door.
- Time it, including archive retrieval, and compare against your customer commitments.
- Verify emergency contacts at every exercise.
- Record what was hard, not just the score, and raise real gaps as nonconformances.
The question a traceability exercise should answer is not whether you can trace a batch. It is whether you could do it on your worst day, without your best person, under a deadline you did not choose.