What a Pack Line Audit Should Check Beyond the Certificates

Pick one pack off the finished-goods pallet, write down its serial number, and ask for every record that number touches, from the day its cells were issued to the moment it passed end-of-line test. That thread is the audit. The certificate wall in the lobby is a story about the past; the serial thread is a story about the present, and it tells you whether the process is documented or merely practiced. When a pack comes back from a customer at eighteen months, the only thing that saves you is being able to say which other serials went through the same cell lot, the same weld schedule, on the same shift.

I take the serial off a pack that is already banded to a pallet, not one sitting on a bench mid-build. The palletized ones are the ones nobody has had time to groom. Then I walk it backwards through the building, and the quality engineer walks with me carrying a laptop and a three-ring binder, and the first twenty minutes of that walk tell me more than the next four hours of the audit agenda.

You can hear the floor before you see it. The pop of a resistance welder, the click of a torque driver hitting setpoint, the hiss of an air line. Smell of isopropyl and flux. Heel grounders scraping on an ESD floor. I am not listening for anything in particular. I am listening for whether the sounds change when a supervisor walks by.

Read the scope line before you read the room

ISO 9001 registration covers the quality management system for whatever the scope statement says, and only that. "Assembly of lithium-ion battery packs" is not the same registration as "Design and manufacture of lithium-ion battery packs," and neither of them is the same as a scope that quietly carves out cell procurement. I have stood in buildings where the framed certificate listed a corporate address two states away and the scope language was broad enough to cover the plant, a warehouse, and a service depot at once. That is not fraud. It is just a document that answers a different question than the one you are about to ask.

So before the tour: find the certificate's site address and compare it to the building you are standing in. Read the scope paragraph twice. Then set it down, because everything useful from here on is downstream of it.

What ISO 9001 does not tell you is whether the weld schedule was ever validated, whether the torque tool was in calibration last Tuesday, or whether the operator can read the traveler she is signing. The standard asks whether you have a process and whether you follow it. It has no opinion on whether the process is capable. That gap is the entire reason you drove here.

The IATF question, and when to stop asking it

IATF 16949 was written by and for automotive OEMs. It brings PPAP, APQP, PFMEA discipline, control plans, MSA on the measurement systems, capability requirements, and a change-notification habit that is genuinely hard to fake. It has teeth because the OEM holds the supplier's business hostage to it.

Robotics is not automotive. A pack that strands an AMR in a cold aisle at two in the morning, in a facility where a technician has to put on a coat and walk out to it, is a real cost, and if the fleet is large that cost can justify asking for automotive-grade process discipline. But it will shrink your supplier pool and it will move the piece price. I have watched buyers default to "must be IATF registered" in an RFQ because it felt like the safe answer, and get back two quotes instead of nine, from suppliers who were not necessarily better at welding tabs.

My position: ask for the specific disciplines you actually need — control plans, first-article approval on changeover, documented weld validation, MSA on the EOL tester — and let the registration be a tiebreaker. If you do require IATF, require that the certificate covers the pack line specifically. A corporate registration that includes a stamping plant does nothing for you.

Walking the line at the speed of a pack

Cell grading and matching is the first stop, because it is where a pack's destiny gets set. The assembler measures open-circuit voltage and internal resistance on incoming cells and bins them. The audit question is not whether the equipment exists. It is whether the grading results are recorded per cell or per lot. Per lot is common and it is close to theater: if the whole incoming lot is treated as one grade, the matching step is a story people tell about their process.

Then stacking and welding. Welding is where packs live or die, and it is where I slow down. What is the weld schedule, and is it locked by part number or can an operator dial it? Does the welder log energy and resistance per weld, and does that log carry the pack serial or only a timestamp? Is there a destructive coupon — typically a tab pull on the first weld of a shift — pulled to a force target, and is the coupon's result tied to the schedule and the machine that ran it?

BMS mounting and busbar torque come next. I ask to see the torque tool, and I ask what happens when it fails to reach setpoint. If it is a click-type wrench and the answer is "the operator feels it," that is a different level of control than a transducerized driver writing a value to a log.

Potting or sealing, if the pack is IP-rated. Mix ratio, pot life, cure time, and the batch record for the resin. Seal test results, if there is a seal test.

End-of-line test last. Insulation resistance and hipot, charge and discharge, capacity, internal resistance, the communication check that confirms the pack talks to the charger and the host. The interesting part is not the test. It is the disposition logic around a failure — can a failed unit be re-run, how many times, and what triggers escalation instead of a retry?

The records I actually pull

Four record families, and the same question for all of them: does this key on the pack serial, or on time and station?

Weld pull and peel data. Torque logs. In-process inspection results. End-of-line test logs.

A lot of EOL testers write a CSV named by the timestamp of the run, and if that is all you have, you can still reconstruct a pack's history — as long as you trust that nobody moved a pallet or ran two units out of order. I have done that reconstruction. It takes an afternoon and it gives you an answer you cannot defend.

The technique that cuts through this is to ask for a record from a specific day. Last Tuesday. Not a blank template, not a sample from the file marked "examples." If it takes twenty minutes and two phone calls to produce, that is not a paperwork problem. That is the finding.

When cells arrive through a distributor

Plenty of assemblers buy cells through distribution, and when they do, the distributor's certificate of conformance is a document about documents. What I want is the assembler's own incoming inspection record: sample size against the lot, the date code captured, what they measured, and the disposition on any failures. If they are doing nothing but counting boxes and filing the CoC, say so plainly and price the risk.

The harder issue is lot capture. If the cell lot number never gets recorded against the pack serial, then traceability starts at the pack, not at the cell — and a cell-lot problem becomes a fleet problem. That is the worst sentence you can write in a field failure report, and the time to prevent it is on the audit, not in the recall.

Changeover is an audit question, not a scheduling detail

Most of these lines run two or three SKUs. A changeover means fixtures swap, the weld recipe changes, machine programs change, and the bins get cleared. Line clearance is the step people fudge, because it is pure discipline and it produces no visible output.

Ask for the most recent changeover record and the first-article approval that followed it. Then ask who signs that no cells from the previous build remain in the feed or in the stacker. If the answer is "the operator," go ask the operator. You will learn in one sentence whether it is a real check or a signature.

What the serial thread is actually for

Everything above produces one deliverable: a containment boundary. If that pack fails at eighteen months, which other serials go with it? Which cell lot, which weld schedule, which shift, which torque tool, which changeover window? If the answer is "all of them," you have learned something worth the trip, and it is better to learn it now with the supplier's quality engineer sitting across from you than on a call with a customer's operations director.

That is also why the nonconformance log is the most revealing document in the building. Not the certificate wall — the NC log, the MRB dispositions, the rework travelers, and if you can get a look at it, the scrap bin. Reworked packs that get re-serialed or not re-serialed. Deviations that were signed off and shipped. That is the culture, written down in the handwriting of whoever was on shift.

None of this substitutes for a quality agreement. The audit does not create obligations, it tells you what to put in them. What it gives you is a sense of how fast and how narrowly this supplier could pull packs out of the field if they had to, and whether the person sitting next to you at the conference table is the person who would actually be doing it.