Why a Pack Serial Number Alone Cannot Prove Cell Traceability

A pack serial number is a label, not a ledger. It identifies the finished assembly in front of you and says nothing about what is sealed inside it. The link between that serial and the actual cells exists only if the assembler's work order ties them together: a cell consumption record, the weld or fastener parameters used at assembly, and the end-of-line test results, all keyed to that same number and all retrievable on request. Without those records sitting behind it, the serial is just a good-looking barcode.

I have made that request enough times to know what the reply means before I finish reading it. A pack comes back from a field teardown, or a cell supplier issues a notice on a specific lot, and I send the assembler an email with one serial number and a photo of the laser etch. Then I wait.

The good reply is a packet. Work order number. Build date and shift. The cell lot or lots consumed on that work order, with quantities. The welding program revision or the torque values that were in force at that station that day. The end of line record: capacity, DC resistance, charge and discharge traces, the pass flag. If the unit was reworked, the rework record too. When I get that packet, a suspect cell lot collapses to a finite list of pack serials, and containment becomes twenty units on a shelf in a warehouse instead of a container held at a port.

The bad reply is a build date and a lot code. That is not stonewalling, usually. That is simply what their records contain. And that distinction matters, because it changes what you do about it.

Batch-Level Consumption Is the Quiet Failure

At a lot of shops, cells are issued to a work order by the bin, not by the unit. An operator pulls from a tray holding two hundred cells from one supplier lot, builds two hundred packs, and the record says "lot X, two hundred pieces consumed." From the assembler's side that looks complete. Every pack on that work order got cells from that lot, and the lot is either good or bad.

From my side, it depends entirely on the failure mode. If the hazard is the whole lot, batch-level is genuinely adequate and I will say so. If the notice is narrower, a coating irregularity traced to one coating run, a tray that sat on a dock over a weekend, a weld issue from a single upstream shift, then "lot X" is a range, not an answer. I can tell you that a thousand packs might be affected. I cannot tell you which forty actually are. That is the difference between a targeted field action and a full shipment hold, and it is decided years earlier, by whether somebody scanned a barcode at the point of consumption.

That scanning step is the unglamorous mechanism that makes per-unit traceability affordable at low and medium volume. Three scans per pack: the pack serial entering the cell loading station, the reel or tray label at the moment cells come off it, and the pack again at end of line. I am not describing a software rollout. I am describing three trigger pulls. What matters is that the scans land against the same identifier and that the resulting rows survive the next system migration.

Ask for a Live Pull, Not a Certificate

During a supplier audit, the traceability section is usually a slide with a screenshot and a sentence about full lot control. I do not care about the screenshot. I hand the quality manager a serial number off a pack that is physically in the room, and I ask them to produce the packet while I watch. Then I ask for one built six months ago and one built last week.

Three things come out of that exercise, and none of them are on a certificate. How long the pull takes, because a trace response is a containment response and a slow one turns a targeted action into a guess. Whether the older record is as complete as the recent one, because systems get better over time and the pack you are worried about is already eighteen months into service. And whether the person doing the pull knows what a complete packet looks like, because if they have to go ask someone, the process lives in one person's head.

I also ask them to do it for a pack that went through rework. That is where most programs quietly fall apart.

Where the Chain Actually Breaks

Rework is the failure I see most. Two cells get cut out and replaced after an end of line failure, or a busbar gets re-welded, or a connector is swapped. If that work is not recorded against the same pack serial, the original consumption record is now wrong, and wrong is worse than missing. It is wrong with confidence. The pack in the field carries a trace that points at cells that are not in it. So I ask for the specific case: when a cell is replaced after end of line, does the removed cell's lot get flagged as removed, and does the new cell's lot get added to that serial? If the answer is a shrug, that is my answer.

Scrap is the second crack. Cells issued to a work order that never made it into a pack. Cells installed but never scanned because the reader was down. Ask what happens when a scanner fails. Is there a paper traveler, and who enters it, and when, and against which serial? A traceability claim that only holds while the hardware is working is a claim with a hole in a predictable place.

And there is a structural problem with cell markings that buyers tend to skip past. Requiring the cell supplier to laser mark every cell is expensive and, on its own, buys nothing. A marked cell only helps if the assembler reads that mark and writes it into the pack's record at the moment of installation. Even then, the marking often does not survive to the point where you need it. Teardown of a welded, potted, or encapsulated pack is destructive. By the time you have a cell in your hand with the marking facing up, you have usually destroyed the assembly you were trying to verify it against.

Which is the point under the point. The cell marking is a fact about a cell. The pack serial plus the work order is a fact about a build. Only the second one tells you which cells are in the unit sitting in your customer's facility.

Certificates belong to the same category. A UN 38.3 test summary, an IEC 62133 report, a UL 2271 listing, an ISO 9001 certificate with clause 8.5.2 identification and traceability ticked off: these establish that a design was tested and that a quality system exists. They do not tell you what happened on a Tuesday in March. Certificates prove the design. Records prove the build. Buyers who accept one in place of the other find out which they had at the worst possible moment.

What Goes in the RFQ

None of this is hard to ask for, but it has to be asked before the purchase order, because afterward you are negotiating with someone who already has your money and a production schedule.

I want a definition of the record, not a promise of traceability. On request, the supplier returns the full assembly record for a given pack serial, containing the work order number, cell lot identity and quantities consumed, welding or fastening parameters, end of line test data, and any rework entries. I want a response time for routine requests and a shorter one for a containment event, because those are different obligations and should be staffed differently. I want it written down that batch-level consumption is acceptable only where the buyer has agreed in writing for a defined program. And I want it to survive a change of ownership at the supplier, which means it lives in the supply agreement and not in a slide deck.

Here is the tradeoff, and I will not pretend it away. Per-unit consumption records cost cycle time. At low volumes, a scan at the loading station is nearly free and the payback is obvious. At hundreds of thousands of packs a year, reconciling every reel against every pack adds real station time and real infrastructure, and there are programs where batch-level records plus tight lot control upstream is the honest engineering answer. That is a decision to make deliberately, with the failure modes in front of you, not a default you fall into because nobody asked.

The reason I keep pushing on this is simple and a little grim. The day you need the record is the day you have no leverage left. A cell lot goes bad, a field unit does something it should not, and the only thing standing between a targeted containment and a full hold on everything you have shipped is a row in a table that somebody either captured or did not. A pack serial number is a fine place to start. It is a terrible place to stop.