Qualifying Your Second Pack Supplier Before the First One Fails

Second-source qualification for a custom lithium-ion pack is a schedule problem, not a paperwork problem. The sequence — spec freeze, drawing release, design review at the new supplier, tooling, first-article build, dimensional and electrical testing, first-article approval, pilot run, production release — takes as long as its slowest step, usually the tool or the test lab queue, and none of those days come back once a disruption starts. Two rules carry most of the weight. Start while the primary supplier is still shipping on time. Qualify the alternate against the same specification, the same drawing revision, and the same test limits as the primary. Miss either one and what you have is a second supplier, not a second source.

I was standing at a receiving bench one afternoon with two packs side by side, and that is the day the difference stopped being a phrase I used in meetings.

Same part number on both labels. Same customer. The left pack came off the primary's line, the right one off the alternate's. Both stickers said Rev C. I pulled the alternate's first-article folder and the dimensional report inside was signed off against Rev B of the drawing — the enclosure tool had never been re-cut after the revision that moved two mounting bosses. The pack had passed every test it was run against. It had just been run against the wrong drawing. Nobody had lied to anybody. The alternate built exactly what its tooling told it to build, and its tooling was a revision behind.

That is the failure mode I plan against now, and it does not look like a bad supplier. It looks like a good supplier, an old approval, and a revision letter that changed somewhere in a document control system while everybody was busy.

The clock you are actually buying

When I build a qualification plan now, I write the sequence out as a stack of queues instead of a list of activities, because the queues are what eat the calendar. Design review sits behind the supplier's engineering workload. Tooling sits behind the tool shop. The first-article build waits on sample cells, and the sample cells wait on an allocation decision made upstream weeks or months before anyone tells you about it. The test queue sits behind every other program in the lab. The pilot run waits for a slot in a plant that is already running.

Some of those you can buy down. You can pay to jump the tool shop queue, you can reserve lab time, you can fly samples. None of it changes the fact that a pack has to be built, measured, and aged before you know whether the process is capable, and that cycle has a floor. What starting early actually buys is optionality. When the alternate hits a problem — and it will, usually in the first-article dimensional report — the primary is still shipping, and the problem stays a supplier's problem instead of becoming your line-down problem.

So the trigger is not bad news. The trigger is comfort. Start the alternate qualification on the same day you start feeling comfortable with the primary. If you wait for a signal that something is wrong, you have already spent the schedule margin you needed.

Same revision, or it is a different part

The equivalence rule is where most second-source programs quietly fail, because it is not one check. It is a stack of them, and the part number on the label only covers the top one.

Qualifying the alternate to the same specification is not enough on its own. The specification is the requirement; the drawing revision is how that requirement was last implemented. Both have to match. When I walk a new alternate through this, I check the cell part number and generation, the BMS board revision and the firmware revision and the configuration file that goes with it, the connector part number, wire gauge and cut length, thermal interface material, label content and placement, and packaging. Same measurement method too — same fixture, same gauge, same limits, same sample size. If the primary's dimensional report was taken with a go/no-go gauge and the alternate's with a CMM at a different datum scheme, you are comparing two different numbers and calling it a match.

There is a legitimate version of this that is not a violation. Sometimes the alternate genuinely needs a different drawing — a different tool, a different assembly approach, a different busbar. That is fine, but it stops being a second source and becomes a second design. It needs its own qualification, its own change control, and an interchangeability check that proves fit, form, function, and interface before you treat one as a substitute for the other. Calling it the same part because the customer's part number is the same is how you end up with two packs that both meet spec and do not fit the same bracket.

A new site is a new first article, even when the design never changed

This is the part people push back on. The design is unchanged, so why does the alternate have to run a full PPAP-style submission?

Because PPAP is not a statement about the design. It is a statement about the process that produced the part, and the alternate is a different process — different tool, different fixture, different operators, different incoming inspection, different everything that turns a drawing into a pack. A first-article approval is site-specific by definition. Change the site and the approval is void.

So the alternate submits its own package: design records and the change documents that go with them, process flow, PFMEA and control plan, measurement system analysis, dimensional results, material and performance test results, initial process studies, sample parts, the master sample, and whatever checking aids the program requires. If the design is identical, the design records are copies. Everything downstream of them is new, and that is the point.

Keep the whole package tied to a revision and a date. An approval without a revision letter attached to it is not an approval. It is a memory.

The split is a commercial decision and a quality one

Dual sourcing means splitting volume, and splitting volume changes the economics on both sides. Tooling gets amortized over fewer units, or the alternate buys a lower-cavity tool with different process capability. NRE gets renegotiated. Minimum order quantities get worse. The alternate's operators get less repetition, and that shows up in the first-article data as variation that has nothing to do with the design.

The split percentage is a commercial decision. It is also a quality decision, and the two have to be made in the same room. A sliver of volume keeps the price bad and the process cold. A meaningful share keeps the tool warm and the operators sharp but costs you the primary's volume leverage. There is no percentage that is right for every program; there is only the percentage that matches the failure you are actually insuring against, and whether you can afford the NRE to get there.

Two suppliers, one constraint

The check that most second-source programs skip is the one below the pack. Map the sub-tier under both suppliers and find where the two chains converge. The BMS board is the usual one. So is the enclosure tool, if both pack houses bought from the same tool shop. So is the cell itself, which is the one nobody likes to hear about, because a second pack assembler buying the same cell out of the same allocation is not a second cell source. It is the same source with an extra step in the middle.

The continuity document should name the single-sourced items inside each supplier's chain, not just the supplier names. If both packs stop when one board house stops, you paid for two suppliers and got one.

Keeping the approval alive

An approved first article is a snapshot. It is valid for one drawing revision, one BOM, one tool, one line, and one sub-tier chain, and it starts going stale the day it is signed. A supplier that produced nothing for a long stretch is not a live second source — the people who knew the program have moved on, the tool has been moved to a shelf, and the approval in the file describes a process that no longer exists in that form.

Keeping it live takes a small standing order, a periodic revision cross-check against the primary's current drawing, and a re-run of the first article whenever the revision changes. That costs money, and it is the honest tradeoff: full second-source qualification is not free, and neither is maintaining it. Sometimes the right answer is a partial second source — a second line inside the same plant, or a second site of the same supplier — which is faster and cheaper and covers some failure modes and not others. A second line in the same building does not help you if the building is the problem. Which one you need depends entirely on what you are afraid of.

What I ask at the end of every one of these reviews is a single question: which revision is the alternate approved to, and what is the date on that approval?

If two people have to go look it up, you do not have a second source yet. You have a plan to look for one, and the clock is already running.