When Your Only Lithium Cell Supplier Fails, Will You Have a Backup?

The only real backup cell source is one you qualified while your primary was still healthy, which means three things done in advance: a cell you have measured and built on your own line, a BMS handshake you have confirmed at temperature extremes, and reserved capacity for your specific part number in writing. A supplier you have never built with is not a backup. It is a name in a spreadsheet that will not survive the first allocation email.

I keep a shallow cardboard box on the corner of my bench with four cells standing in it, each in its own bag, each bag marked with a date and a supplier code. Nothing in that box is for building. It is my reminder that a cell is a physical object before it becomes a line on a purchase order.

The last time I pulled it down, our buyer was standing in the doorway with a printout. Our primary cell supplier had sent an allocation notice with no end date on it. She asked whether the second source was ready. I handed her a caliper and told her we were about to find out.

"Ready" turned out to mean several months of work, most of which we had already done badly once.

Start with the caliper, not the datasheet

The datasheet is a marketing document with tolerances attached. The cell on the bench is the truth. So the first hour of any second-source evaluation is quiet: incumbent on the left, candidate on the right, calipers and a scale and a magnifier, and nobody talking.

Diameter at three points, because a wrapped cell is not perfectly round. Overall length including the top contact, because a positive cap that stands a fraction of a millimeter proud will bottom out in a holder before the tab welds. Where the tabs sit and how wide they are, because a narrower tab changes your weld schedule and a wider one may not clear the spacer. Vent geometry. Sleeve thickness. Whether the can has a step or a shoulder.

Then I take one of each apart, Dremel and vise, gloves, on a tray, and look at what is inside. Not because I can evaluate a jelly roll, but because I need to know whether the safety devices are arranged the way my pack's fault analysis assumed. A current-interrupt device and an internal fuse in a different position, or the absence of a venting path where I assumed one, is not a drop-in. It is a new design review.

Here is the checklist I actually fill in, in the order I fill it in:

• Mechanical: diameter, length, tab width and thickness and position, weldable area, vent, sleeve, weight.
• Electrical: nominal and minimum capacity, charge and discharge cutoffs, AC impedance at 1 kHz, and pulse behavior at the temperature my pack actually sees.
• Chemistry and thermal: the same nominal chemistry is not the same blend; ask for the aging and thermal characterization, not just the cycle-life marketing curve.
• BMS: protocol, register map, thermistor type and beta value, shunt or coulomb-counter assumptions, and the state-of-charge versus open-circuit-voltage curve the gauge is calibrated against.

That last bullet is where most drop-in claims die. Two cells, both nominally the same capacity and both NMC, can have OCV curves that differ by tens of millivolts through the flat middle of the discharge. Your fuel gauge was calibrated on the incumbent's curve. Swap the cell and the gauge still reports a percentage. It will just be the wrong percentage, and it will be wrong in exactly the part of the curve your customers stare at.

Same story with the thermistor. A 10k NTC is a 10k NTC until it isn't; a different beta value reads a couple of degrees off at the ends of the range. That is the difference between a pack that derates charging near freezing and one that doesn't, and you will never see it on a bench at room temperature.

Second source and dual source are not the same program

I learned this one the expensive way. A dual source means you split volume between two qualified suppliers from the start: two PPAP packages, two incoming inspection plans, two sets of weld settings on the fixture, and a standing argument about which line runs which SKU. It buys you resilience every day and it costs you yield and attention every day.

A second source is dormant. It is a supplier you have fully qualified, that is not shipping to you at volume, that you keep alive with small sustaining orders and an annual re-qualification. Cheaper to run, slower to turn on, and it has a failure mode all its own: dormant sources rot. The supplier's process drifts, the cell gets a revision letter, your qualification package goes stale, and the account manager who cared about your modest annual order moves on to a bigger account.

Which one you want depends on how much revenue rides on that cell. If a line-down week costs more than a second qualification costs in a year, run a dual source. If your volume would make you a rounding error at a second supplier, you may not be able to buy a genuine dual source at all, and the honest answer is a dormant second source plus a deliberate buffer.

The reservation is the part everyone skips

Qualification is the visible work. The reservation is the work that decides whether the backup exists on the day you need it.

A supplier who has passed your audit and your pilot build, and who has no commitment from you, will allocate constrained output to the customer with the signed agreement and the rolling forecast. That is not a slight. It is arithmetic. So the deliverable of a second-source program is not a qualification report. It is a qualification report plus a capacity reservation agreement.

What I want in that agreement, in plain terms:

• A named part number, not a cell family or a "similar format."
• A rolling forecast window and a take-or-pay band around it, so both sides know what "reserved" is measured against.
• The ramp rate, meaning how quickly the reserved quantity can be called off, and what happens if I need it faster than that.
• Where the reservation sits: a specific line, or general allocation. The first is worth far more than the second.
• A defined re-qualification trigger for cell revisions, so a silent process change doesn't quietly invalidate everything behind it.

Expect a holding fee, a prepayment, or a minimum annual commitment above your current forecast. That cost is real, and it belongs in your landed-cost model next to freight and duty rather than being treated as an insurance premium you slip past finance. If you cannot get a reservation at all, that is information. You have a qualified supplier but not a qualified backup, and you should say so internally in exactly those words.

Prove it on your line, not on a bench pack

The validation that counts is not the one where an engineer hand-builds six packs. It is the one where second-source cells go through your actual weld fixture with your actual recipe, and you measure pull strength and resistance afterward.

This is where a second source usually forces a process change. Different tab material, different thickness, different plating, and the schedule tuned for the incumbent may not hold. That is fine, but it is a documented and validated change, and it takes a pilot run rather than a memo.

I like to run the pilot in a low-consequence window: a small-volume SKU, a separate shift, or the slow quarter, with a work instruction that names the second-source cell explicitly so nothing gets mixed on the floor. Then sample the pilot into accelerated aging at the temperatures the pack actually lives at, and put a handful of units in the field where you can watch them.

The trap is running that pilot during a crisis. If you are building your first second-source lots because the primary has already failed, you are doing a first article inspection under the worst possible conditions, with customers calling, and every quality escape gets pinned on the new cell whether or not the cell caused it.

Watch the supplier, not just the part

Second-source planning assumes the backup will still be there. So monitoring has to be continuous, and it has to look at the supplier rather than the cell.

I don't have a dashboard for this. I have a folder and a habit. Every quarter I look for the things that show up before an allocation notice: hiring freezes on the manufacturing side, capex that gets pushed a year, longer lead times on the supplier's own incoming materials, a customer mix drifting toward someone much larger than me, turnover in the account team, pressure to renegotiate payment terms. Quality drift is the other tell, a slow rise in incoming rejects, or a process capability number that used to sit comfortably above your limit and now doesn't.

None of that is proof of anything. It is a reading, and it changes how much weight I put on the reservation.

What this actually costs

Two qualified sources means two incoming inspection plans, two sets of safety documentation kept current, two sets of dangerous goods paperwork every time samples move, and two process settings on the line. On a low-volume custom pack, that overhead can exceed the risk it covers, and pretending otherwise is how second-source programs get killed in the next budget cycle.

When a true second source is not economically available, a custom format or a proprietary can or a volume too small to interest anyone else, the honest substitute is a documented single-source risk with a buffer sized to your real recovery time and reviewed on a schedule. That is a worse answer than a qualified second source. It is still better than a name in a spreadsheet.

The box on my bench gets a new cell about twice a year. That is the whole discipline, really. Qualify people before you need them, write down what you measured, and pay something now so the phone call later is a scheduling problem instead of an existential one.