Lot and batch traceability: defining the lot you would have to recall

A lot needs a stated rule for where it begins and ends

Write the rule per process rather than per plant: a lot ends at a material change, a tooling change, a setup, a shift change where that matters, a cleaning cycle, or a defined quantity, whichever comes first. Ambiguity here surfaces during containment, when nobody can say whether the material loaded mid-run belongs to the same lot as the pieces before it. The rule should be enforceable by the person running the job with a container label and a production record, because a definition that requires the office to reconstruct it afterwards is a definition nobody follows.

Lot size is the main lever on recall exposure

Large lots reduce administration, changeover marking and paperwork; they also mean that any doubt about one piece contaminates a large quantity. Small lots multiply handling and record-keeping, and they narrow the population you would have to reach. The way to size them is to work backwards from the consequence: how much would it cost to recover, replace and account for one lot of this product if it had to be recalled entirely. Where that figure is severe, smaller lots are cheap insurance; where the product is low-consequence, the paperwork is the larger cost.

Splitting and merging is where genealogy is won or lost

Real production constantly divides and combines. A lot is split across two machines, part of it is held for rework, remainders from several lots are combined to complete an order, a hopper is topped up mid-run, and packing consolidates output from a week into one pallet. Each event has to be recorded as a relationship rather than a replacement, so the resulting quantity carries the identity of everything that fed it. Where combining is unavoidable, the resulting mixed unit inherits the exposure of every parent lot, which is an argument for not doing it casually to save a container.

Reconciliation is the check that catches the missing pieces

At the end of a lot, the quantity issued should equal the quantity good, plus scrap, plus samples taken, plus material still held. When it does not, something left the process unrecorded: pieces taken for a trial, parts put aside by an operator, an unlogged rework, or a miscount at packing. Discrepancies look like an accounting nuisance until a containment, when the unaccounted quantity becomes the population you cannot rule out. Making reconciliation a closing step of every lot, with a threshold that triggers investigation, keeps the numbers meaningful. It also exposes yield problems that scrap reporting alone would never surface, because material lost without a ticket looks like efficiency.

Rework, returns and re-lotting

Material that leaves the flow and comes back is the most common source of a broken chain. A batch reworked weeks later, a customer return repaired and resold, or held stock released after a concession all need a decision: does it keep its original identity, or does it become a new lot with the original recorded as a parent. Either is defensible, and doing it inconsistently is not. The one rule worth making absolute is that reworked material never rejoins production unmarked, because it carries a different process history and a different risk profile than the pieces beside it.

Frequently asked questions

Can a date code serve as a lot number?

It can, provided the production rule ties it to conditions rather than to the calendar alone. A code covering a whole day of output across two shifts, several material deliveries and a tooling change describes a period, not a batch made the same way. Date coding works well where the process runs continuously with one material feed and a clean break each day, and poorly where setups and materials change several times within the coded period.

How do we handle bulk materials where individual lots merge in a tank or hopper?

Treat the vessel as the point where a new lot is created, and record which input lots contributed to each fill. For continuous processes, define the output lot by a time window and hold the input record for that window, accepting that the boundary is approximate and that adjacent output may share inputs. Document the assumption. It is far better to state a known imprecision than to imply a granularity the process cannot support.

What identification should appear on the container itself?

Part number and revision, lot identifier, quantity, the date the container was filled, and its status — released, on hold, or awaiting inspection. Status marking matters more than people expect, because most mixing incidents involve stock that was physically fine but not yet cleared, or material rejected and never removed. A container that has to be interpreted by consulting a system elsewhere will eventually be moved by somebody who did not consult it.