Adding a Second Pack Supplier Without Paying Two Minimums
Splitting an annual pack volume across two assemblers only pays for itself when the second source owns its own part number or its own slice of demand, not a percentage of the same part number. Two plants building the identical pack, each carrying its own annual minimum, turns a modest program into a program you buy twice: the split drops both plants below their commitments, pushes both out of the tier where they were pricing you, and duplicates the tooling and first-article work once while the minimums repeat every year. The structure that avoids it is to give each assembler whole work, whether that is a whole SKU, a whole service channel, or everything above the incumbent's committed ceiling.
I have had the two quote sheets squared up on my desk more than once, unit prices close enough that the decision looked like a coin flip. It never was. The number that decides it is never in the unit price column. It is in the notes block underneath, where MOQ, annual commitment, tooling amortization, and NRE recovery sit in that small font suppliers use for the parts of the quote they would rather you skim.
So here is what I do with it.
The split that does not work
Take a program that needs 14,000 packs a year, one part number, one housing. The incumbent will commit to 10,000 a year at the price you already have. The candidate will commit to 8,000 a year to take the business.
The numbers below are made up to show the shape, not to quote a market. Every buyer I know has run some version of this table.
| Plant | Annual minimum | 50/50 split | Short |
|---|---|---|---|
| Incumbent | 10,000 | 7,000 | 3,000 |
| Candidate | 8,000 | 7,000 | 1,000 |
| Total | 18,000 | 14,000 | 4,000 |
You have 14,000 units of demand and 18,000 units of commitment. That 4,000-unit gap does not disappear because the split looked fair in the meeting. It comes back as one of three things: packs you buy and store, a shortfall you pay, or a conversation you have in the fourth quarter when the forecast is already locked. At a unit price in the nineties, that gap is roughly the cost of a new hire, sitting on a rack.
And that is only the first hit. The second one is quieter. The incumbent was pricing you as a 14,000-unit customer. At 7,000 you are a different customer on their rate card, one tier down, sometimes two. So you lose the tier at the plant you already trust at the same time you are paying a premium at the plant you do not trust yet.
Where the second minimum actually hides
The annual commitment line is the obvious one. The expensive ones are the adders that scale with the inverse of your volume.
Tooling amortization is the clearest. A fixture set and an end-of-line test jig for a mid-size pack is a capital number. Put thirty or forty thousand dollars into a fixture set against 8,000 units and it is a few dollars a pack. Cut the second plant's volume to 3,000 and the same fixture is suddenly a large fraction of a small unit price. The fixture did not get more expensive. Your base got smaller.
Setup is the same trap, and it is the one people miss. If you keep monthly releases but each release is half the size, you run the same number of line changes over half the units. A line-change charge that was under a dollar per pack at a thousand-unit lot is nearly two dollars at five hundred. Nobody quotes that as a minimum. It just shows up in the price when you ask for a small lot.
Then there is NRE recovery: programming the BMS against the second plant's test bench, building the harness, writing work instructions, the first-article package, the pilot run. Some of that arrives as a one-time number. Some of it is quietly amortized into the first year's unit price, which is why the second source's quote looks competitive at full volume and stops looking competitive when you model your actual split.
The duplicate list you fund once
The minimums repeat annually. The qualification bill hits once, and it is bigger than most buyers budget for.
- A second fixture set, built to the second plant's line, not the first plant's.
- A second end-of-line test jig, calibrated, with its own gauge R and R.
- A second first-article inspection package and process approval file.
- A second pilot run, including the engineering days to sit on it.
- A second quality system audit, and another one every year after.
- A second set of incoming inspection plans, records, and AQL decisions.
The hardware is the visible part. The engineering weeks are the part that hurts, because they come out of the same calendar you were going to spend on the next program.
Two lot chains instead of one
The moment the second plant ships, you own two traceability chains for one finished pack. Two sets of lot codes, two sets of incoming records, two answers needed when a field return comes back and you have to know which building welded it.
The cheap fix is a plant code in the first characters of the lot number. Then a warranty return tells you the plant in five seconds instead of an afternoon of email. I have watched a small thing like that save a week of investigation, and I have watched its absence cost a week.
Budget the inspection bench time too. Two suppliers means two receipts, two sampling plans, two sets of paperwork. It is not hardware cost. It is a person, or part of one, every week.
Alternating lots
There is a middle path that looks like it dodges the whole problem: alternate production lots between the two plants so neither goes cold. You do not duplicate everything at once, and both suppliers see real volume.
What you trade for it is qualification speed. If the second plant builds every other month, it takes twice as long to accumulate the process history that tells you whether they can hold the spec. Long gaps between lots are exactly the conditions where drift hides. A process that runs twelve lots a year can be read. A process that runs six, irregularly, mostly tells you that it ran.
Alternating also means carrying enough finished stock to cover a switch, because no plant ramps on a dime.
What warm actually costs
Keeping a source warm with a couple of small lots a year is a real strategy, and it has a real price. Those lots fall under the minimum, so they get quoted at the low-volume tier, often a meaningful premium over where the same part prices at volume. On a couple of thousand packs a year, that premium is a five-figure line item in some cases. Call it insurance, then compare it honestly to the alternatives: six weeks of buffer stock, which is working capital plus a compliant lithium storage area you may not have, or the revenue you lose if the primary plant goes down for a month.
One caveat I put in front of anyone considering this: a supplier that is qualified but idle is worth less than the certificate suggests. The clock on their readiness starts when they run production lots, not when the audit closes. If the last real lot was fourteen months ago, you have a name on a list, not a source.
Give each plant whole work
The structure that actually avoids two minimums is not a discount. It is a different division of labor.
Split by SKU. If the program has two part numbers, say a high-volume standard pack and a low-volume variant with a different enclosure or connector, give each plant a whole part number. Each supplier's minimum is then met by demand that exists on its own. Nobody is short. The incumbent keeps the volume leader. The candidate gets the tail SKU they can build profitably at a smaller annual number, and they get genuinely good at it because it is the only thing they run. I have seen this work better than any split-percentage deal, because it removes the argument about whose units count.
Split by growth. Give the second source everything above the incumbent's committed tier. Their minimum is satisfied the moment your program grows, and until it does you are not paying for phantom units at the incumbent. The risk is obvious: if growth does not come, the second source is still short. This only works if you are honest with them about the forecast and they price for it.
Split by channel. Service and spare-parts demand is genuine demand. It is usually steady, it often arrives in smaller lots, and a smaller assembler is happy to run it. Hand the second source the service channel and their minimum is real work.
When the honest answer is no
Below a few thousand packs a year, the arithmetic usually does not close. Duplicated qualification can cost a year's worth of margin on that program, and the second source's premium eats most of the rest. That does not mean you accept single-source risk. It means you buy the risk down with something cheaper.
Hold the drawings and the validated manufacturing package so a new plant can quote from a complete package instead of starting from your CAD. Own the tooling. Keep a producing buffer in finished packs, sized to the recovery time you can actually tolerate. Qualify a second cell or BMS source, which costs a fraction of qualifying a second pack assembler and covers the failure mode that causes most shortages in the first place. And keep the incumbent aware that a second source exists, because that conversation changes behavior long before it changes volume.
What I hand to the contracts person
The one thing I never do is try to write the commitment wording myself. What I do is write a one-page statement of intent and hand it over: what volume each plant should see in a normal year, what counted as committed volume, what happens in a down year, what happens if a plant misses on quality or delivery. The words are the contracts person's job. The arithmetic behind them is mine, and if the arithmetic is not in that memo, the wording will be clean and the deal will still be wrong.
Then I read the final terms back against my model and check one thing every time: whether the commitment is counted against the program or against the plant. Everything else in a dual-source arrangement can be fixed in operations. That one gets fixed before you sign, or it does not get fixed at all.
