Where Unit Cost Actually Lives in a Custom Battery Pack BOM

Unit cost in a custom lithium-ion pack sits in six BOM lines, and they do not carry equal weight. The cell sets the floor, the BMS sets the ceiling on anything smart, and the enclosure, connectors and interconnect, and assembly plus end-of-line test fill the space between. The design choices that actually move unit cost are the ones that change cell count, series count, current path, sealing method, or the volume of per-pack test data the buyer demands. Everything else is noise until those four things are settled.

I keep the pack BOM open in one window and the drawing in another, and I read down the lines in that order, because the lines talk to each other whether or not the quote does. A supplier will hand you a single piece price and a tooling number, and that number hides the whole conversation. My job is to get the conversation back out into the open, one line at a time, so that when the price moves I know which design decision moved it.

The cell line sets the floor, but not in the way the spreadsheet suggests

Chemistry, capacity, format. Those three words are the whole cell line, and everything downstream inherits from them.

Chemistry is the first fork. An LFP pack generally lands lower on the cell line for a given amount of energy, and it also tends to be more forgiving on the thermal and safety hardware side, which quietly pulls cost out of the BMS and out of the enclosure's thermal provisions. An NMC pack buys you energy density and pays for it at the cell line and again in the thermal and protection hardware. I have watched buyers fight hard for a cell price concession on an NMC design when the real money was sitting in the thermal hardware that chemistry choice had dragged in behind it.

Format is the second fork, and this is where I see the most expensive wrong instinct. Someone decides a larger-format cell will be cheaper because there are fewer of them. Sometimes that is true. Often it is not, because the arithmetic is not just cell count. Fewer, larger cells do cut interconnect hardware, weld joints, and the number of placements the line has to make. But series count follows pack voltage and per-cell voltage, not capacity, so the BMS channel count does not budge when you move to a bigger cell within the same chemistry. And if the taller can pushes the pack height, you have just re-opened the enclosure, the tooling, and the end-of-line fixture. I have seen a cell swap that saved on the cell line cost more in a re-cut housing tool than it ever gave back.

Capacity is the third fork, and the honest way to read it is as a trade against everything that scales with cell count rather than a straight price-per-cell comparison. When a supplier quotes two capacities, I ask what happens to the interconnect count, the weld count, and the enclosure height before I look at the delta.

The BMS is priced by three numbers and one ownership question

Channel count is the first number. It follows series count, and series count follows pack voltage. If someone on your team wants to add a parallel string, that adds cells and interconnect, not channels. If someone wants a higher pack voltage, that adds channels, and channels are the meter on the BMS line.

Current rating is the second number, and it is the one people over-spec. Continuous rating and peak rating are different purchases. Peak is what sizes the switching and the current path, and peak is usually a number pulled from a load profile that nobody re-examined. I have seen a pack carrying a peak requirement two or three times what the application ever drew, and the BMS line was paying for that margin in hardware that also ran hotter and needed more room. That is not a design lecture, it is a line item. Ask who owns the peak number and whether anyone has looked at it since the requirements document was first written.

Communication is the third number. Protection-only boards and smart boards with CAN or SMBus are not the same product category. Adding CAN means a transceiver, often isolation, and a connector with more pins, and that connector change ripples into the enclosure cutout and the sealing method around it. It is never just a chip.

The ownership question is firmware. Whether the BMS firmware is developed in-house, licensed, or built by the pack supplier as NRE changes what you pay up front versus what you pay per pack, and it changes who you call when a pack comes back from the field behaving strangely. I want that answer in writing before the quote is final, because it determines whether the BMS line in year three looks anything like the BMS line in year one.

The enclosure is priced by the seal, not by the plastic

Material matters, but less than people think. ABS versus PC/ABS versus glass-filled nylon versus aluminum is a real cost conversation, and aluminum brings machining or extrusion plus a different set of fastening and grounding questions. But the driver that dominates is the sealing method, and the sealing method is dictated by the IP rating.

An IP54 requirement and an IP67 requirement are two different enclosures, two different assembly processes, and two different end-of-line test regimes. A gasket, an ultrasonic weld, and a potting approach each carry their own tooling, their own cycle time, and their own scrap rate. When a product manager casually upgrades the spec to IP67 because it sounds better in the marketing copy, the cost does not move by a few percent on the enclosure line. It moves the sealing method, the fixture, the test time, and the reject rate.

And any custom molded housing carries tooling as a separate one-time number. That tooling is a real decision, not a formality. At low volume, a machined or off-the-shelf enclosure with a modified lid is sometimes the cheaper path even though the piece price looks worse, because you are not amortizing a tool across a quantity you may never reach. I always ask what the tooling amortization assumes and what happens to the piece price if the volume forecast misses by half.

Connectors, busbars and wiring grow with current, and they grow fast

The lines I see underestimated most often are the ones carrying the current. Connector cost rises with current rating through the contact system, the plating, and the housing, and it rises again when the connector is a custom part with its own tooling and minimum order quantity. Busbar cost rises with copper cross-section and with plating. Wiring cost rises with gauge and with the labor to route and terminate it.

There is a second cost hiding here that never shows up in the piece price: sourcing risk. A custom connector can lock you to one supplier with one lead time, and that lead time becomes a cost the day you need to ramp.

This is also where the peak current spec from the BMS conversation comes back around. Every amp of peak you did not need to specify is copper you are buying, connector family you are restricted to, and enclosure volume you are reserving. The three lines move together.

Assembly and end-of-line test is a line item, not overhead

Buyers treat the labor and test portion of the BOM as a residual, and suppliers are happy to let them. It is not a residual. It is a function of how much of the process is automated.

Cell sorting and matching takes time, and how much time depends on how tight the buyer's matching requirement is. Tab welding is a process choice with different capital, different cycle time, and different scrap characteristics. Inline inspection adds stations. None of this is exotic, but all of it lands on the piece price, and the buyer who never asks about it has no way to tell whether the quote is efficient or padded.

The bigger lever is the record-keeping you require. If you want a serialized end-of-line test record per pack, retained for a defined period, tied to cell lot traceability, that is a real cost per unit and it does not shrink with volume the way people assume. It is also frequently the right thing to require. I would rather pay knowingly for the traceability than discover during a field issue that nobody can tell me which cell lot went into which pack. But I want it priced as a line, visibly, so I can decide whether the requirement is worth what it costs.

Where the lines interact, and why single-number quotes mislead

Here is the thing a line-by-line read gives you that a total never will. The expensive moves are rarely inside one line. They are the ripples.

A cell format change touches the enclosure height, which touches the tool, which touches the end-of-line fixture. Adding CAN touches the pin count, which touches the connector, which touches the enclosure cutout, which touches the sealing method and the IP test. Raising the peak current rating touches the busbar cross-section, the connector family, and the internal volume, which pushes back on the enclosure. Every one of those is a design decision that arrives on the BOM as three or four separate cost increases, and if you are reading a single number you will attribute all of it to whatever you happen to be negotiating that week, which is usually the cell price.

That is the trap. Cell price gets the attention because it is the biggest line and the easiest to benchmark. Meanwhile an over-specified peak current and a marketing-driven IP rating sit there quietly costing more than the concession you flew across an ocean to win.

What I actually put in front of a supplier

When I send an RFQ, I ask for the BOM broken into these buckets: cells, BMS, enclosure and sealing, interconnect and connectors, assembly, and end-of-line test. One-time costs go on a separate page from piece price, and the tooling amortization assumption is stated explicitly.

Then I ask three questions that do more work than any price negotiation. Which specification drives each line. What happens to that line if the spec relaxes. And what else on the BOM changes if that line changes.

That last question is the one that separates a supplier who knows their own cost structure from one who is reading it off a template. The answer tells me whether I am buying a pack or buying a guess.