Serial number management: allocating, marking and following individual units through the plant

One source of numbers, and never reuse one

Serial numbers should be issued from a single authority so no two units can share an identity, and a used number is retired permanently even if the unit was scrapped before completion. Reuse seems harmless and creates an unresolvable ambiguity years later when a returned unit matches two records. Decide the format deliberately too: whether it encodes product, plant or period, or is a meaningless sequence. Encoded formats are convenient for humans and become a liability when the encoding runs out or the meaning changes, so many plants keep the number meaningless and hold the attributes in the record instead.

Marking that survives the process and then the field

The mark has to outlast heat treatment, painting, washing, handling and years of service in whatever environment the product lives in. Options — laser etch, dot peen, engraved plate, printed label, moulded-in character — each fail somewhere: labels come off in solvent, etched marks disappear under coating, moulded characters fill with paint. Choose against the actual process route and the service environment, then verify readability after the harshest step rather than at the point of marking. Where a unit is marked before a process that could destroy the mark, define how identity is maintained across that step.

Choosing where in the route the number is captured

Capturing at every operation produces a complete history and a large amount of operator effort. The pragmatic approach captures at the points where the information will be needed: the birth record where the serial is created and associated with its order, any step where a serialised or lot-controlled component is joined to the unit, any process whose parameters must be recoverable per unit, final test where results are attached, and dispatch where the unit is associated with a customer and a shipment. Between those points, lot-level records usually suffice, and adding scans nobody will query is pure cost.

Rework, scrap and the unit that gets taken apart

This is where serialisation quietly breaks. A unit fails test and is stripped for parts, and its components go into other units without record. Another is scrapped but its number stays open forever in the system. A third is rebuilt so thoroughly that it is arguably a different unit. Write the rules: a scrapped serial is closed with a reason and never reissued; components recovered from a serialised unit are either quarantined or re-identified, not returned quietly to a bin; and a unit rebuilt beyond a defined threshold gets a new identity with the old one referenced. Without those rules, the genealogy stops being trustworthy after the first year.

The value only appears after the unit leaves

Serialisation costs money in the plant and pays off outside it. A field failure can be traced to the components, settings and date of build, which turns a blanket recall into a defined population. Warranty claims can be validated against actual build and ship records rather than a customer's word. Aftersales can determine the correct spare for the configuration actually shipped. In medical devices and several other sectors a unique identifier is a regulatory requirement with its own database obligations. Deciding what to capture should therefore start with the questions the business will need to answer once the product is out of reach.

Frequently asked questions

When should we serialise rather than use lot control?

Serialise when a question will one day be asked about an individual unit: high-value products, anything with a service life and spare parts, safety-critical items, products under warranty, and anywhere regulation requires unique identification. Lot control is sufficient where units are interchangeable and any problem would be handled at lot level anyway, which covers most fasteners, chemicals and commodity components. Serialising everything is expensive and generates data nobody queries, so the decision belongs product by product.

At what point in production should the serial number be assigned?

At the first step where the item becomes a distinct unit that will be tracked — often the first major sub-assembly or the point where a housing or frame is committed. Assigning too early means numbers are consumed by material that may never become product; assigning too late means the early operations cannot be attributed to the unit. Where the earliest processes matter for traceability, a temporary carrier identity linked to the final serial at assignment is a workable compromise.

What happens if a serial number becomes unreadable during production?

Have a defined procedure rather than letting an operator improvise. Typically the unit is held, its identity established from surrounding records such as the order, the position in the batch or the test data, then re-marked under authorisation with the event recorded against the unit. What must not happen is a unit being re-marked with a plausible number by whoever noticed, since that creates a duplicate identity that will surface later as two conflicting histories nobody can reconcile.