Traceability has two granularities. Lot-level answers which batch a problem touched, right for anything made in bulk. Unit-level, serialization, answers what happened to this exact one, and costs more to build: every unit marked, every step captured against the mark. The decision is economic, and the boundary is clearer than vendors make it sound.

When the serial pays

When lot-level is the right answer

Bulk, blended, and consumed products: food, beverages, resins, board, chemicals. The lot is the natural liability unit, and lot genealogy answers every question regulators and customers ask (see how lot genealogy builds itself). Serializing each snack bag adds marking cost and data volume without a better answer to any real question.

How serial genealogy gets built

Same principle as lot trace, finer grain: mark early (laser, label, RFID at first operation), then capture against the mark at the steps that matter, component lots joined at assembly, test results at the stand, torque values at critical joints, rework in and out. RFID earns its cost where units go dark between operations, the part pulled for rework that vanishes from tracking is the classic case. Every serial then answers 'how was I made,' and the aggregate becomes a quality goldmine: ML clusters field failures by mold, shift, component lot, and cell, which lot-level data cannot resolve. The buying framework lives in our traceability software guide.