Industrial track & trace: from machine to product

A serial number only tells you something if you can find the components, operations and checks linked to it. That is what industrial track & trace is about: knowing where a product is and being able to reconstruct how it was made. For factories and machine builders, it starts not with a dashboard, but with reliable records on the shop floor.

Tracking and tracing: the difference

Tracking answers the question: where is this product and what is its current status? Tracing looks back: which material batch did it come from, which machine processed it and what were the measurement results? How current that information is depends on when it was recorded. A scan on departure does not automatically tell you exactly where a product is afterwards.

The level of detail you need varies by process. Sometimes batch-level records are enough; sometimes every component needs its own serial number. What matters is keeping the links intact: which raw materials went into which batch, which components were assembled together and which products were delivered?

From identification to a useful product record

Barcodes and QR codes are simple ways to identify products, containers or packaging. They need a readable code and a clear line of sight. RFID can work without a direct line of sight, but needs testing with the actual materials and reading environment. Choose the technology to suit the process, not the other way round.

At each relevant process step, record at least the product or batch ID, the time, the machine or workstation and the result. Where needed, link material consumption, process settings and quality measurements. Record scrap, rework and replacement components too. Otherwise, the record only shows the planned route, not what actually happened.

Connecting machines with OPC UA and MQTT

Manual scanning remains useful, but machine data can be collected automatically where possible. OPC UA can provide structured access to machine data. MQTT can publish events and measurements to other systems. The two can complement each other; neither makes data meaningful on its own.

That takes agreed definitions and conventions. What does ‘operation complete’ mean? Which serial number does the measurement belong to? Do the machines use the same time reference? Without that context, you have isolated signals, not reliable traceability. Also allow for buffering when connections drop and processing that detects duplicate messages. That way, a network outage does not quietly create gaps or duplicate records.

A Unified Namespace (UNS) can make this information available through a shared structure, for example by factory, line and machine. This helps systems use the same names and context. A UNS does not replace a historical product record, though: you still need storage, retention periods and traceable corrections. Connections to ERP, MES and WMS then link production data to orders, materials and deliveries.

What it lets you demonstrate

When a material batch is suspect, you need to identify which products were made with it and where they ended up. Conversely, you need to trace a defective product back to its components, process steps and checks. This supports quality investigations and helps define the scope of a potential recall, provided the records are complete and reliable.

Traceability also helps with audits and sector-specific requirements. Requirements vary across food, pharmaceuticals and other manufacturing environments. So agree upfront which data you need as evidence, who can access it and how long you must keep it. Collecting more data is not automatically better.

Start small and test the exceptions

We would start with a clearly defined product flow and a specific question, such as: can we find every material batch used from the serial number of a delivered product? Then test missing scans, batch splits, rework and network outages too. Involve operators and quality staff: they know where records are likely to drift from reality.

AI can later help search records or flag possible links between deviations. We keep a clear boundary here: AI only reads, a human decides. The original records remain the source; an AI summary is not evidence and must not release a product or control a machine.

Want to explore where traceability starts in your production process, or where it is missing? Book a 15-minute conversation with Vlegel Technology in Genk, the software engineering team for machines, data and AI. We will discuss your product flow, existing machine connections and a technically sound first step.