What Does a Second Pack Source Really Add to Your Unit Cost?

A second pack source adds cost in three places, and most programs only count one of them. Splitting annual volume between two suppliers drops the volume tier at both, so the resulting price step lands on the entire program rather than the half you moved. On top of that sits a one-time block for the second supplier's tooling, qualification builds, and design-level certification testing, plus a recurring load of duplicated audits, change control, and documentation. On a program in the tens of thousands of packs a year, those three together commonly land somewhere between a few cents and several dollars per pack, and the volume-leverage loss is usually the largest single piece. None of it means much until the interruption it is supposed to cover has a number attached to it.

The price step lands on every pack, not half of them

There's a printout on my desk right now. One page, two columns, the same pack quoted at 40,000 units a year and at 20,000 units a year. The gap is eight dollars.

Everyone in the room reads that as: the second source costs eight dollars more per pack on the half we move to them. That's wrong, and it's wrong in the direction that makes dual sourcing look cheaper than it is.

When you go from 40,000 a year at one supplier to 20,000 and 20,000, neither supplier is quoting the 40,000 tier anymore. The whole program reprices. Forty thousand packs times eight dollars is $320,000 a year of leverage handed back, and the second source only builds 20,000 of those packs. If you charge the entire leverage loss to the second source's half — which is exactly what a controller will do when they ask what the second source actually costs — you get sixteen dollars per second-source pack, not eight.

That's the number to put on the whiteboard before anyone starts talking about resilience.

It depends on whether your suppliers publish real tier breaks. Some quote a flat price and recover the leverage at the annual negotiation instead, so the loss is real but shows up once a year as a worse quote rather than a line item on a sheet. Either way it's there. And tier curves usually aren't linear — the step between 20k and 40k tends to be steeper than the step between 40k and 80k, which is why splitting mid-range volumes hurts more than people expect. If your annual volume is large enough that you sit comfortably above the last published break, the leverage loss shrinks and the rest of the arithmetic starts to dominate.

The one-time block, amortized over the second source's volume

The second supplier needs its own production tooling. Weld fixtures for the tab and busbar joints, a potting or overmolding tool if the pack uses one, end-of-line functional and HiPot jigs, a programming fixture if the BMS gets flashed at the pack house, plus the handling and shipping fixtures you specified for the first supplier. Some of that copies straight out of the design package. Some of it gets rebuilt from scratch, because the second supplier's laser welder or test rack isn't the same model and the original programs won't load.

Then the qualification builds. A-samples, B-samples, DV, PV — whatever your stage gates call them. Those packs come off a line that isn't earning production money yet, at lower first-pass yield, with more rework and more MRB and more of your engineers standing on the floor. The first few thousand saleable packs out of a new line cost more than the quote suggests, and that delta is real even though it never shows up on a purchase order.

Then certification. Transport testing is design-specific, and a second assembler is not automatically covered by the first supplier's test summary. If the cells, the BMS revision, or the protection scheme differ in any way that matters, you're paying for the test series again. Depending on your end product there may be additional listing or EMC work that has to be repeated on the second build.

Stack it up and, for illustration, call it $150,000 to $250,000 for a moderately complex pack: new tooling, a full qualification cycle, one round of transport testing. If that second source will build 60,000 packs over three years, you've added roughly $2.50 to $4.00 to every pack it makes, before anyone discusses price.

The ongoing load is the part that never ends

Two approved suppliers means two quality systems to watch. Two audits a year instead of one, plus the corrective actions out of both. Two first-article packages to keep current, two incoming inspection plans, two streams of end-of-line data to review, two scorecards, two sets of cell traceability records you'd need if a recall ever landed.

On paper: two quality agreements, two sets of material declarations, and two transport test summaries that have to be revised whenever the design moves.

My own rule of thumb is to budget a quarter of a quality engineer's year for the second source's audits, data review, and paperwork during steady state. That's before corrective actions, and it's before the first year, which is always worse.

Every change after launch costs twice

Here's the one I see underestimated most.

Identical specification sheets do not produce identical packs. Two suppliers build to the same drawing, and within a year the two builds have drifted — one substituted a connector because of lead time, one is running an older BMS firmware because the flash file didn't get pushed to their programming station, one quietly changed a torque spec. None of those were malicious. All of them were avoidable, and all of them cost money to find.

The control you need is boring and expensive: one change control board, both suppliers on the same ECN distribution, both sites re-validating every change. That means every engineering change after launch gets qualified twice. If you historically ran four ECNs a year on this pack, you now run eight qualification cycles, and the second one is never free.

Firmware is where it gets sharpest. If the pack house flashes the BMS, you need the flash file version-controlled on both sides, read-back verification at end of line, and a record tying serial number ranges to firmware revisions — because when a field issue shows up in eighteen months you need to know which supplier and which revision is in the unit in front of you. And if the host device checks a pack identifier or a protocol string, the second source's BMS has to present the same one. When it can't, you're shipping two SKUs.

Two SKUs means a second service part number, a second spare stocking decision, a second set of field documentation. That cost lands on service, not procurement, which is why it almost never makes it into the sourcing business case that justified the split in the first place.

Check what the second source actually buys you

Two pack assemblers is not two supply chains.

Before you pay for the split, ask the candidate for its cell part number, its cell manufacturer, its BMS vendor, and its distributor. If the answers match your primary supplier's, you have bought assembly redundancy. That covers a fire, a flood, a labor action, a plant-level quality shutdown. It does not cover an allocation event or a sub-tier shortage, because both sites will be short the same cell at the same moment.

That may still be worth buying — a plant-level failure is the most common interruption I've dealt with. Just price it as assembly redundancy, not as full supply continuity, and don't let the business case quietly assume protection it doesn't have.

Put a number on the thing you're insuring against

Now the other column.

Replacement tooling lead time: if your only supplier's line is destroyed, how long to rebuild the fixtures? For pack tooling I plan on ten to sixteen weeks, longer for anything involving a custom mold.

Requalification: even with a complete design package in hand, a new site needs a build and a validation cycle. Add eight to sixteen weeks on top, overlapping the tooling if you're lucky and running them in parallel if you're willing to spend more.

Lost production: your line-down cost per week, times the weeks. That's the exposure before probability.

Then multiply by a probability, and here's the honest limitation — that probability is a guess. Nobody has a real number for the likelihood that a specific supplier's plant burns down. Write the guess down anyway. Writing it down turns a shouting match about risk appetite into an argument about one number, and that number can be revised every year as you learn more.

Annualize it. A once-in-twenty-years event that costs $4 million is $200,000 a year of expected loss, and that's what it has to beat.

Why minimum share locks the cost in

The structure I see most often is a primary and secondary split with a floor for the second source — 80/20 or 70/30 — so the second source stays warm, keeps its operators trained, and stays production-viable. That floor exists for a reason. Below roughly ten or fifteen percent of your volume, most suppliers will quietly de-prioritize you, and the qualification you just paid for decays into a piece of paper.

The floor is also what converts an occasional cost into a permanent one. You aren't buying the option to dual source; you're buying a standing commitment to sit in the lower price tier forever, and to carry the audit and change-control load for as long as the program runs.

The comparison, in one line

Annual carry cost equals the per-pack price delta times full program volume, plus amortized NRE, plus the ongoing quality and documentation load.

Compare that to the annualized expected interruption loss. If the carry cost is larger and the sub-tiers overlap, dual sourcing is buying something you already assumed you had.

The alternatives are worth pricing in that case: keep one supplier and fund a spare tooling set that you own, hold a strategic buffer of finished packs sized to your actual recovery window, and reserve capacity for surge. Those often cost less than a live second source and cover the same failure mode — with two caveats. Buffer stock does nothing if the supplier goes bankrupt. Owned tooling does nothing if the real problem is a sub-tier shortage at a supplier you don't control. Match the mitigation to the failure you actually fear, and price the second source against that, not against a vague sense that two is safer than one.