How to Win a Capacity Fade Guarantee from a Resistant Supplier

The capacity fade guarantee gets won before anyone drafts a warranty clause — by narrowing what the supplier actually promises to a bounded duty cycle, a stated capacity retention floor, and a claim process with a pre-agreed cost split, then paying for that narrowing with volume commitment and a shared verification bill. Suppliers that refuse "a capacity fade guarantee" in the abstract will usually sign a narrow one, because the resistance is almost never about the cells. It's about open-ended liability they can't price.

I keep a scan of one refusal in a folder, because it is the cleanest example I own of a warranty that guarantees nothing.

Second meeting, a supplier's application engineering office. Four of us at a table built for twelve. Sample packs lined up along the edge, that smell of shrink wrap and warm electronics that never quite leaves those rooms. The quality manager slid a two-page warranty sheet across. Defects on page one. Cycle life on page two. I put my pen on the line: 3,000 cycles.

"Three thousand cycles to what?"

"To end of life."

"Which is what?"

"Eighty percent. Typical."

There it was. Two pages, a number that looked like a commitment, and one adjective that made the whole thing a decoration. He wasn't lying to me. He was telling me exactly what he could defend, and what he could defend was a range.

"Typical" is the entire negotiation

I've stopped treating that word as sloppiness. It's a signal. When a supplier writes "up to 3,000 cycles" or "≥80% retention, typical," they are telling you they measured something on a fixture, at a temperature they chose, and they are not willing to put a number on a population of cells that will be doing something they've never seen. Push on the adjective and you get a definition. Push on the definition and you get to the real question, which is: what profile?

That's where most buyers lose the deal. They ask for a stronger guarantee without changing the information they've given the supplier. The supplier hears "guarantee 80% at 2,000 cycles" with no idea whether you're running an AMR at 0.5C with long rests and overnight charging, or an AGV that pulls a 90% depth of discharge twice a shift in a hot warehouse behind a charger that overshoots. Those are two different businesses. One of them the supplier can warrant. The other one, they shouldn't.

Show up with a duty cycle they can price

So before the next meeting I build the duty cycle file. Not a spec-sheet summary — the actual week. Current profile sampled at short intervals, peak discharge, average, the regen spikes if the drive throws them back. Depth of discharge per cycle. Rest periods, and whether they're real rest or just idle with the BMS awake. Cell-level temperature from a thermocouple buried in a pilot pack, not the ambient reading off the wall unit. Charge regime including the tail — constant-current to constant-voltage with what cutoff, at what ambient, on the charger you already own.

Then I compress it. A shift that runs eight hours with three charge events becomes a loop I can describe in one paragraph and test in a few weeks. Pick a loop that over-weights the hard parts — the deepest discharge, the hottest hour, the fastest charge — and say so out loud. A supplier who watches you stack the deck on purpose trusts the resulting number more, not less.

Some years back I sat in a subcontract lab watching forty-eight cells from a pilot lot run that loop on two racks of cell testers, chamber held at the warehouse's summer temperature. Every few days another discharge capacity number populated a column. The point of sitting there was never the data. It was that I could describe the test, the fixture, and the failure mode in the next negotiation without saying "we think."

That's the trade. You hand them a bounded problem, and in exchange you can demand a bounded promise. "Eighty percent retention at 2,000 of these loops, on this profile, at this chamber temperature" is a sentence a quality manager can carry into his own engineering group. It has edges.

Give away the half of the risk they can't control

Every fade conversation gets stuck on the same fear, and it isn't fade. It's the field failure they'll be blamed for. A pack that cooked because the enclosure had no airflow. A charger floating above the per-cell limit. A BMS that let a string run too low under load. A robot left on a dock in July.

So I hand them the exclusion list before they ask for it. Thermal events traceable to the enclosure we designed. Over-discharge beyond the cutoff our BMS should have enforced. Charging outside the profile we supplied, or a profile we changed without telling them. Water ingress. Physical damage. Anything downstream of our integration, we carry.

This costs nothing to concede and it moves the room. What's left on the table afterward is the thing you actually want: the cells lose capacity faster than agreed, under conditions both sides signed off on. That's a narrow claim, and narrow claims get signed.

The other half of the risk transfer is time. A supplier writing a three-year guarantee on a product that's been in the field eight months is guessing, and they know it. If they have return data, failure analysis reports, or field capacity checks from a comparable program, ask to see the shape of the curve — not the numbers, the shape. Whether fade is linear or has a knee decides whether your guarantee should be a single endpoint or a step-down.

Pay for the guarantee in commitment, not in price

The lever that actually moves a resistant supplier isn't unit price. It's schedule certainty. A guarantee is an insurance policy, and insurers need a book of business before they write one.

What I offer is a firm twelve-month take-or-pay tranche against a rolling twenty-four-month forecast, sized to something I can defend internally. Not the whole program. Enough that the supplier can see a production run inside it. What I don't offer is an invented multi-year volume number thrown out to win the meeting — that commitment becomes a liability on my side of the ledger long after the guarantee has expired, and it's the concession I've most regretted making.

Price moves last. If I'm buying a guarantee with price, I've paid twice: once for the pack, and again on every unit for the rest of the program. Volume and schedule are one-time concessions. Price is permanent.

Write the claim before you write the remedy

A guarantee you can't afford to prove is decoration. I've watched a team win a retention floor and then discover that proving a claim meant shipping packs to the supplier's own lab, waiting eleven weeks, and accepting a determination written by the party being accused. The credit, if it ever came, would have been smaller than the downtime it was supposed to offset.

So claim mechanics get negotiated in the same session as the number. The drafting isn't mine — counsel papers it — but the shape of the deal is, and the shape is what gets argued later.

Sampling. How many packs are tested, from which lots, selected how. One pack failing proves nothing about a population, and a full-fleet test costs more than the remedy. A quarterly pull with a defined sample count, half tested and half retained, is the shape that usually survives contact with a real quality team.

The retained lot. Packs or cells held back from the first production run, stored at controlled temperature, untouched. Without a time-zero reference, every claim becomes an argument about the test equipment.

Who tests, and who pays. Name the lab in advance, ideally one neither side owns. Then make the cost two-way: the supplier covers the verification bill if the claim holds, I cover it if it doesn't. That single clause filters frivolous claims on my side and reflexive denials on theirs better than any amount of meeting time. Cap the exposure so neither party is betting the program on one test result.

The remedy. Replacement packs, credit against the next tranche, or a price adjustment. Decide which one you actually want before you sit down. On a fleet, replacement usually beats credit, because credit doesn't restore uptime.

The line between defect and degradation. These get conflated in supplier paperwork, sometimes carelessly and sometimes not. A cell that fails at month two is a defect. A population sitting at 71% at month thirty is fade. Different remedies, different proof standards. If the warranty doesn't separate them, your fade claim gets filed as a defect claim and slow-walked.

The ladder, for when they still say no

Not every supplier signs the first draft, and a no on the full guarantee isn't always a no on the risk. What I work through, in order.

A step-down instead of an endpoint — say 92% at 500 cycles, 85% at 1,200, 80% at 2,000. Suppliers find this easier because it matches how fade actually behaves and cuts the exposure at any single point in time.

A shorter term with a scheduled reopen. Sign twelve months now, then set the long-term retention number at month six using pilot and early field data. Write the reopen into the agreement with a date and a defined data package, so it isn't an option to renegotiate later — it's an event on the calendar. I've gotten better final terms this way than in any single sit-down, because by month six the argument is about data rather than confidence.

A pilot-lot guarantee. Full terms on the first N packs, standard terms on production. This is often the only path with a supplier who has never built for your application, and it isn't a bad outcome. You find out whether they can hold a number before you're committed to a fleet.

When a supplier won't sign anything

At some point the answer stops being a negotiating position and becomes information.

A supplier with a mature cell, field data from comparable duty cycles, and a process they trust will sign a bounded guarantee, because they expect to pass it. A supplier who won't sign any retention number at all — not stepped, not shortened, not on a pilot lot, not even against a profile they helped write — is telling you something about their confidence under your conditions. That's worth more than the clause. It belongs in the sourcing decision, not the contract file.

I walked from one program on that basis. Maybe the packs would have been fine. But a supplier who won't put a fade number on paper is also a supplier who will be very hard to talk to in year three, when the fleet is at 76% and you're deciding whether to replace the packs or the vendor.