Production Tracking: Traceability, Yield and Getting Off Paper Travelers

Most small manufacturers can tell you how many units they shipped last month, and almost none can tell you which station cost them the most time. Production tracking closes that gap: it records what was built, by whom, from which lot, how long it took and whether it passed — one unit at a time. In regulated production it is mandatory. Everywhere else it is the cheapest yield improvement available.

Start one level below what looks impressive. A barcode scan at five stations, feeding one database, beats an unfinished MES rollout every time — and it can be running before the next production lot.

For medical devices, records must show that each unit was built to the released device master record, by trained staff, on validated equipment, with unique device identification where required. Electronic records themselves fall under 21 CFR Part 11, meaning audit trails and controlled access, not an editable spreadsheet. Plan tracking as part of medical device manufacturing setup, not as an afterthought once the auditor books a date.

Traceability feels like overhead right up to the first field failure. Then it decides whether you recall two hundred units or twenty thousand. If you can tie a failing unit back to a serial number, a build date, a work order and the component lot fitted that day, containment is a narrow, survivable event. If you cannot, the only defensible action is to recall everything that might contain the suspect lot, and the cost difference between those two outcomes typically dwarfs a decade of tracking effort.

The same records answer questions you will face regularly even without a failure: which units received a firmware revision, whether a warranty claim is inside the covered build window, and whether a supplier's claimed process change actually correlates with a defect rate shift. Each of those is unanswerable from paper travelers filed in a cabinet.

Scope matters. Full component genealogy on every resistor is rarely justified. Trace what is safety-critical, expensive, single-sourced or historically troublesome, and record the rest at lot level. A focused system that people actually use beats a comprehensive one they route around.

The failure mode of shop-floor digitization is buying a system before defining the process. Start with the data you would need in a recall, work backwards to the minimum number of scans that capture it, and only then choose tooling. Most small operations need three events recorded: work order opened against a build, serialized unit associated with its key component lots, and final test result attached to the serial. Everything else is refinement.

Keep data entry inside the operator's existing motion. A scan while the part is already in hand costs seconds; a form filled at a shared terminal at the end of a shift produces fiction. If a step cannot be captured without pulling someone away from the bench, either redesign the capture or accept that the record will be unreliable and plan accordingly.

Run paper and digital in parallel for two or three weeks, then stop paper deliberately on a named date. Indefinite parallel running guarantees the digital record stays incomplete, because everyone knows the paper copy is the real one.

We build these records into production programs through short-run manufacturing and manufacturing consulting, and design them in early during engineering so serialization is part of the product rather than bolted on.

Collected records only pay off if someone looks at them on a schedule. A weekly review of first-pass yield by station tells you where the process is drifting before it produces scrap. A monthly look at defect Pareto by component lot exposes a supplier problem while you still have leverage. A quarterly comparison of field returns against build date validates whether the corrective action you implemented actually worked, which is the question most organizations never answer.

Assign one owner for the review and give them authority to stop a build. Data with no decision rights attached becomes an archive, and archives do not prevent recalls.

A workable serialization and traceability setup for a small assembly operation is far cheaper than most teams assume. Barcode or QR labels, two or three scanners, a tablet per station and a modest cloud database land in the low thousands of dollars for hardware, with software either a small monthly subscription or a few weeks of internal build. The real investment is process design and the discipline to keep it running, which is measured in attention rather than capital.

Full component genealogy with supplier lot integration is a different scale of project, typically requiring a proper MES and months of rollout. Do not start there. Implement the minimum viable record, run it for a quarter, and let the gaps you actually experience justify the next increment. Systems specified in advance of experience are consistently over-built in the areas that do not matter and thin in the ones that do.

One more consideration is auditability. If you sell into medical, aerospace, automotive or industrial channels, a customer or a registrar will eventually ask to see records for a specific unit, and the request usually arrives with a short deadline. Being able to produce a build history, component lots and test results for a serial number within an hour is the difference between a routine audit finding and a corrective action that stalls shipments. Design the retrieval path deliberately, and test it once a quarter by pulling a random serial and timing how long the answer takes.

It is the practice of recording what happened to each unit as it moves through manufacturing: component lots consumed, station and operator, cycle time, test results and any rework. The output is unit-level traceability plus yield and throughput data by station.

Usually not at first. Barcode scanning into a single database, costing a few thousand dollars and a few weeks, delivers most of the traceability and yield benefit. A full manufacturing execution system makes sense with multiple lines, frequent engineering changes or formal regulatory requirements.

Because each finished serial number is linked to the component lots inside it, a defective supplier lot can be traced to exactly the units that contain it. Containment then covers those units and their customers rather than every unit built that quarter.