Qualify Your Second Battery Supplier Before You Need One

Qualifying a second battery supplier means redoing two things and inheriting everything else. The site audit runs against the second source's own building, and the pack-level validation runs against the second source's own build. Cell qualification data, the specification set, the drawings, the design record, the firmware, and the compliance file all transfer from a qualified first source, provided the second source is pinned to the same revision of every one of them. What does not transfer is anything that attaches to a configuration: a different BMS, a different enclosure, a different connector, a different cell orientation inside the same holder. Those reset the pack-level test work, and no document from the first source changes that.

The bench that explains the whole job

I'm standing at an end-of-line test station in a pack shop we'd shortlisted, and the BMS board in the pack on the bench is not the board on our released drawing. Same cell. Same holder, same nickel strip, same 6S3P layout, same label. The engineer running the station tells me it's "the same pack." I ask for the test file, and he hands me the UN 38.3 summary report from our first source.

That's the entire problem sitting on one bench. The second source read our specification, found a BMS that met the functional requirements, and built a pack that works. It is not the pack we validated. The cell-level data in that report is still good, because it's the same cell from the same cell manufacturing site. The pack-level results are worthless here, because the assembled thing is different. External short circuit, overcharge, forced discharge, impact, thermal, vibration, shock, altitude — all of it was run on one assembly and is now being cited for another.

So I stopped asking about price and started asking about configuration. That's the shift that makes second-source qualification tractable instead of overwhelming. You are not auditing and validating a second supplier from scratch. You are auditing a building and validating a build. Everything else you already own.

What carries over

The specification set carries over. The drawing package carries over. The design record, the DFMEA, the compliance file, the firmware binary, the pack-level requirements — they all go to the second source as a package, at a revision, with a transmittal you can point at later.

The cell test summary carries over, with one condition I've watched people skip: the identical cell from the identical cell manufacturing site. Same part number out of a different plant is a different conversation, and you want the cell maker to tell you that in writing rather than assume it.

What does not carry over is anything the first source performed with its own hands. The site audit. The process FMEA as executed on that floor. The control plan as the operators actually run it. First article inspection. Process capability on the welds. And the whole pack-level validation stack.

The reason I split it that way is that the inherited package is a transfer obligation, not a permission slip. The second source has to be able to build from the same documents, which means you owe them the documents, and you owe your first source a conversation about what's being handed over and under what terms. If that conversation hasn't happened, the package doesn't really exist.

The delta audit

You can audit a second source on the delta. Not the whole thing — the parts that are different.

To do that, you need something current from the first source: a process flow diagram that reflects today's line, not the one from the launch review. If the flow is two revisions stale, your delta is wrong and the audit depth is pointed at the wrong stations. I've made that mistake. I marked a laser weld as identical when the first source had moved to ultrasonic eighteen months earlier, and the audit walked right past the operation that mattered.

Once you have a current flow, mark each step: same as the first source, or different. Shared steps get a walk-through and a small sample — enough to confirm the claim that they're the same. Delta steps get the full treatment. Capability, gauge R&R, calibration status, operator training records, first-off approval, reaction plan when something drifts.

The deltas I see most often are boringly predictable. A different weld technology on the busbar. A different reflow profile at whichever shop runs the BMS assembly. Different cell binning and matching equipment. A different end-of-line tester, which matters more than any of them because different test limits mean the two packs leave the building under different acceptance criteria. And a different enclosure sealing process, which is where you find out whether anyone has actually measured cure time or just written down what the sealant vendor recommended.

Confirm the site before you confirm anything else

Second-source searches attract a specific kind of company: the one with a sales office, a website, and no assembly line. Sometimes it's a trading company. Sometimes it's a pack shop that subcontracted the BMS assembly to a third party and didn't think to mention it. Either way, the audit has one job before any technical question gets asked.

Is the address on the quote the address on the door?

I check it against three things that are hard to fake: the address on the quality system certificate, the address on the UL file for the assembly, and the badge the receptionist hands me at sign-in. If the certificate says one industrial park and I'm standing in another, the qualification is for a company, not a factory, and I'm not interested in qualifying a company.

If the BMS is built by a subcontractor, that shop is in scope. Not as a courtesy — as a site. The pack-level performance you're about to pay to test depends on how that board was reflowed, and you can't audit what you haven't visited.

Capacity claims and what to do with them

A stated monthly output figure is the least useful number a supplier will give you. It costs nothing to say.

What costs something is the arithmetic. Ask for the shift pattern. Ask how many people are assigned to the line on each shift, and get names or at least headcount by roster. Ask how many end-of-line test stations exist, because on a pack line that station is very often the bottleneck and one station caps the ceiling no matter what the assembly cells can do. Then ask for the cycle time of the slowest station and the OEE they actually achieve, not the one on the slide.

Then ask to see the test log. Every pack that goes through end-of-line generates a record — a pass/fail row with a timestamp and a serial number, sitting in whatever MES or spreadsheet the shop uses. Pull the last four months for that station and count. That's the only capacity number I trust, because it's a byproduct of work that already happened rather than a forecast of work that hasn't.

One part number, two suppliers, one revision

This is the piece that gets treated as a follow-on task and shouldn't be. Change control is part of second-source qualification, because the moment you have two qualified suppliers, you have two paths for a revision to drift.

Every engineering change notice goes to both. Firmware gets verified by checksum on receipt at both, not by label. A deviation request from one supplier gets assessed against the other, because granting it to one and not the other is how you end up with two different products under one part number.

And here's the part that surprises people: any change that touches the pack configuration re-triggers the pack-level test work at whichever site is affected. Not the cell data. The pack data. If you don't internalize that, you'll approve a small BMS component substitution in a change board, ship it to both suppliers, and never notice that your safety file no longer describes anything you're building.

Second source versus dual source, and why it changes the bill

A second source is an alternate you switch to. A dual source splits volume continuously. Different arrangements, different qualification load.

With a dual source, both suppliers are current at all times, which means change control is continuous and revalidation happens on a shared clock. With a second source, you qualify once and then keep it warm.

That last part is where programs fail quietly. A second source that hasn't built your product in fourteen months has lost the fixtures, rotated the operators, and let the work instructions go stale. If you need it on a Tuesday, it isn't there. Run a trickle of production on a schedule you can defend, re-audit on a defined cadence, and treat the occasional small order as the cost of having an option.

I'll also say the thing that doesn't get said enough: if the cell itself is single-sourced, a second pack assembler buys you assembly continuity and not much else. Same cell, same cell plant, same allocation problem. You've second-sourced the box, not the supply chain, and it's better to know that going in than to discover it during an allocation crunch.

What actually matters on the day you switch

When I go back and look at the programs where the second source worked, the pattern is always the same. The inherited package was transferred cleanly and version-controlled. The audit went deep on the deltas and shallow on the rest. The pack-level testing was run on the second source's own assembly, at the second source's own site, and the report has that site's name on it.

And when I look at the ones that went badly, the second source had a test report that belonged to somebody else, and everyone along the way had agreed it was close enough.