When a Pack Fails in the Field, Can You Find the Cell Lot That Made It?

Yes, you can find the cell lot behind a field failure, but only if the path from pack serial to work order to cell lot, and back out to every sibling unit built in that run, was written into the purchase terms before the first pack shipped. If that chain has to be negotiated after the failure, you are not running a lot trace anymore. You are running broad containment, and broad containment means pulling units that are almost certainly fine.

The pack I keep coming back to arrived on a pallet in a service bay, still in its shipping cradle, with the top corner of the case scorched and the shrink band bubbled. You could smell it before you got the wrap off. And the first thing you reach for, the white label with the serial and the barcode, was the exact thing the heat had eaten. Half the digits gone. That is the moment the whole question stops being theoretical: the record you need most is the one the failure is most likely to destroy.

So you do what you always do. You stop touching it and you photograph everything. The melted label, the laser-etched number on the case end, the barcode sticker on the BMS board under the lid, the ink-stamped code on the cell cans you can see through the wrap, the residue pattern. Then you write the serial down twice, once from the pack and once from the asset record in the fleet system, and you compare them. If those two numbers disagree, you have a bigger problem than the failure.

The walk backward goes through a work order, not a database

People assume the supplier types your serial into a search box and the answer appears. Sometimes it does. More often the serial resolves to a work order, and the work order is where the real information lives: build date, shift, cell vendor, cell lot or date code, firmware version flashed at end of line, and the count of units built. The serial is just the key that gets you into the room. The lot is what you actually came for.

This is why I ask suppliers, during the audit and not during the crisis, to show me a finished pack and walk me from the label to the lot record. Not describe it. Walk it. I want to see whether the lot is recorded per pack or per batch, whether cells get kitted by lot or pooled across lots on the same day, and whether the person on the line scans the cell carton or just the pack label at the end.

If cells get pooled, your trace comes back as a date range instead of a lot number. That is a materially worse answer. A date range turns containment into "every pack built between these two Mondays," which can be ten times the population of a genuine lot. Worth knowing before you sign, because it is very hard to change afterward.

RMA records and production records live in different buildings

Here is the friction nobody warns you about. The person handling your RMA is in customer service or quality, working in a ticketing system that knows your serial, your complaint, your ship date, and your replacement. The person who can tell you the cell lot is in operations, working in an MES or ERP that has never heard of your ticket.

You are asking one system a question only the other system can answer, and the handoff between them is a human being sending an email. That handoff is where the clock dies.

Which is why the terms should not say "supplier will provide traceability." They should say something you can hold someone to: a written trace response within a fixed number of business days of the request, delivered as a document with named fields. Pack serial. Work order number. Build date. Cell manufacturer. Cell lot or date code. Whether cells came through distribution, and if so, the distributor's lot reference. Firmware version at ship. And the list of sibling serials built in the same run.

That last field is the one people forget to ask for, and it is the only one that matters for containment.

Walking forward is a decision, not a report

You get the sibling list and for about a minute it feels like relief. Then you read it.

Some of those serials are in your fleet. Some shipped to a different customer, and the supplier will not tell you who. Some are in the supplier's own stock or in a distributor's warehouse. What you have bought with a good trace is not a recall and not a cure. You have bought the ability to make a narrow decision instead of a blunt one.

In practice I sort the list into tiers. Units in the same duty regime as the failed pack, running the same firmware, in the same high-cycle application, get pulled and inspected first. Units from the lot that are sitting in stores get quarantined, tagged, and held, which costs almost nothing. Units from the lot in light-duty service get flagged in the maintenance system and watched at the next scheduled service rather than yanked off the floor. And I write down why each tier was chosen, because in six months someone will ask.

That sorting is the whole job. A lot-specific containment is disruptive but bounded. A broad containment shuts down routes, strands assets, and forces you to explain to operations why every AMR in the building is parked because of one pack you cannot identify.

When the chain breaks, and it will

The honest version is that a fair number of traces fail, and not because anyone is hiding anything. The label is destroyed. The pack was rebuilt or repaired by a third party and the serial no longer maps cleanly. The cells came from a broker during a shortage and the lot on the invoice is a distributor's internal code with no relation to the cell maker's date code. The pack is a one-off prototype that never got a work order at all.

In those cases you go to the physical evidence, and you go to the build-date range, and you accept a wider net. What you should not do is accept a verbal answer. "It was probably the same cells we always use" is not a trace, and it will not survive the meeting where you have to justify the containment scope.

One more thing that catches buyers: the trace gets you a lot identifier, not a list of every pack in the world that shares it. Other customers of your supplier may be running the same cells with no idea. That is the supplier's problem to manage, and you will usually not be told how. Your containment decision is yours, made on partial information, and it should be scoped to what you control.

Do not ship the evidence away

The default RMA flow wants the failed pack back. The moment it goes on a truck, your ability to answer anything about it goes with it, and you get a report instead of a unit.

So I keep it longer than is convenient. Before anything leaves, the serial, the lot or date code, the firmware version, and the full service history get tied together in one record, in one place, with the photos attached. If the pack must go back for cell-level failure analysis, I ask for it in writing that the unit and its identifiers are retained and that I get access to the teardown findings, not just a conclusion.

That single record is what turns one failure into a data point. Without it, you have an anecdote and a replacement. With it, the third pack from the same lot becomes a pattern you can take to a supplier with a straight face, and containment stops being guesswork and starts being a number.

What I actually put in the documents

Nothing elaborate. A trace clause with a response window and named fields. A retention clause that keeps the failed unit identifiable. A requirement that the serial label be backed by a secondary, permanent marking that survives heat, because the one on the sticker will not. And a question in the audit that has to be answered with a walk, not a description: show me how you get from this pack to its cells.

If the supplier cannot walk it in the calm of an audit, they will not walk it at 6 a.m. on the morning your fleet goes down.