A Capacity Fade Guarantee Can Cut Pack Replacement Cost If You Model It

A capacity fade guarantee lowers fleet pack replacement cost only after the retained-capacity percentage is converted into dollars per unit per year, and that conversion almost always shows the buyer holding more of the cost than the headline number suggests. The percentage pays for the cell. Labor, freight, and the hours the machine is out of service stay exactly where they already were.

I keep that conversion on a tab I have rebuilt more times than I want to admit, and it is open right now. Page four of a supplier's warranty attachment is on the left monitor. Table 2, "Performance Guarantee," four lines of text: not less than 80% of rated capacity at 2,000 cycles, measured at 25 C, 0.5C charge, 1C discharge, per the supplier's published test method. It reads like a strong commitment. It is four lines.

The right monitor is the one that decides the purchase. Column A, unit number. Column B, what that pack does all day — tug, lift, transport, dock. Column C, cycles we actually accumulate in a year, which is never the same as what a test bench accumulates. Column D, the retained capacity at which the unit stops being useful to us. Column E, the supplier's trigger. And then the two columns everything hinges on: years until the guarantee fires, and years until we would replace the pack anyway.

When E sits to the right of D, the guarantee never pays. Not once. That is the first thing I check, and it has ended more negotiations than price ever has.

That is the whole trick, and it is not a trick. A capacity fade guarantee is a financial instrument with a strike price, and the strike price is the trigger point. If the trigger is set where the pack is dead and your fleet quits on packs that are merely tired, you bought a document, not coverage.

The trigger point is a runtime threshold, not a dead pack

A pack does not fail in service. It fades, and at some point it can no longer do the job it was bought for. On an AMR running a ten-hour shift with a mid-shift opportunity charge, that point arrives when the unit starts returning to dock early, or starts refusing the last two tasks in a wave, or drops into limp mode in the last hour. Nobody in operations calls that a battery failure. They call it a slow robot, and then they call me.

The threshold moves with the duty. A low-draw inspection robot tolerates a lot of fade. A high-cycle tug pulling loaded carts up a ramp does not. In my models the number lands somewhere in the seventies or low eighties for most of the fleets I have worked with, but I would not hand that to anyone as a rule. You get it by watching runtime over state of charge as the packs age, not by picking a round number.

The supplier's trigger is usually set later than that, and set on a test bench, not on your floor. This is not malice. It is just two different definitions of "worn out." The guarantee says worn out. Your operation says can no longer complete the duty cycle. The gap between those two sentences is the cost you are underwriting.

Price the lines the guarantee never mentions

Here is how I build the line. For one replacement event on one unit:

• Pack price, times the buyer's share after proration and any remedy discount.
• Labor to remove and replace. On an AMR this is usually not a ten-minute job; there is a cover, a connector with a locking mechanism, sometimes a lift assist, sometimes a thermal pad or a strap that does not go back on the same way. It is also a scheduled appointment with a technician who is otherwise doing revenue work.
• Freight both directions. A new pack inbound is straightforward. The failed pack outbound is not, because it is Class 9 hazmat and the carrier will want to see the paperwork. If the pack is swollen, or has been over-discharged and will not accept a charge, you may find yourself with a pack nobody will take and a supplier asking you to return it before they ship. That argument costs weeks.
• Downtime. Hours out of service, times what an hour of that unit is worth to the operation. This is the line that usually dwarfs everything above it, and no capacity fade guarantee I have seen pays a dollar of it.

Add those four, multiply by expected events per unit over the coverage period, divide by years, and you have cost per unit per year. Now run the same line with the guarantee removed entirely and see the difference. That difference is what the guarantee is worth to you. Not the percentage. The difference.

Most of the time it is a smaller fraction of the total than the buyer assumed going in, because downtime does not care about the remedy. That is the moment when the smarter move stops being "negotiate harder on the percentage" and becomes "negotiate the replacement logistics instead." A supplier who keeps form-fit-function packs in a regional depot and commits to shipping in two business days is worth more to the fleet than three extra points of retained capacity that will not trigger for another year.

Proration and cycle counting: two definitions that decide the payout

Proration is a sliding scale, and it is designed to feel fair while doing the opposite. Read the curve, not the endpoint. Month twenty-four of a thirty-six-month term is not half. It is whatever the schedule says, and the schedule is not always linear. Replacement-only remedies compound it: the supplier ships a new pack, you pay the labor, you pay the outbound freight, you eat the downtime, and you are now running a new pack with a fresh guarantee clock on a unit that has already been through one event. Two guarantees with the same percentage can be worth very different amounts to a fleet, and the difference is entirely in these clauses.

Then there is cycle counting, which is where fleets lose coverage without noticing. Ask the question plainly: does a partial charge count as a cycle, or is a cycle defined as amp-hour throughput divided by nameplate capacity? It matters enormously if your units opportunity-charge. An AMR that returns to dock four times a day and tops up may log four "cycles" a day under one definition and a fraction of one under the other. Same fleet, same guarantee percentage, wildly different coverage window. I have watched a program burn through a 2,000-cycle guarantee in under eighteen months on a definition nobody read until the first claim.

Settle capacity disputes with a protocol, not a conversation

The guarantee is only as good as the method used to prove a breach. If the warranty references "the supplier's test method" without attaching it, you have agreed to a dispute that will be settled by whoever has more patience. I want the protocol in the document, in writing, with enough detail to reproduce: charge rate, discharge rate, ambient temperature, cutoff voltage, number of preconditioning cycles before the measurement counts, rest period, and the tolerance on the test equipment.

The reason is not that suppliers cheat. It is that a capacity measurement is genuinely sensitive to conditions. Run the same pack at a different discharge rate or ten degrees warmer and you will get a different number. If both sides are arguing in good faith with different methods, the argument never resolves. If both sides are running the same written protocol on calibrated equipment, the argument is over in a day.

I ask for a third option as well: an independent lab both parties accept, with the cost allocation written down before anyone needs it. Agreeing this at signature is cheap. Agreeing it after a failed pack is expensive, and you are negotiating from a worse position.

The exclusions describe the failures your own fleet causes

Open the exclusions and read them as a list of things your operation does. Over-discharge is the big one. A unit parked over a long holiday, a contactor that sticks, a BMS in deep sleep that still draws a trickle — the pack goes below cutoff and comes back damaged. That is a fleet-side event, and it is excluded. Charging at high ambient temperature is another. A warehouse that hits 40 C in August, a charger in direct sun on a dock, a pack that charges immediately after a hard shift while it is still hot — excluded. Unauthorized BMS changes are the third, and this one is the quietest, because a fleet-wide firmware push is usually a good idea made by someone who has never read the warranty. Change the charge profile over CAN and you may have just voided your coverage across the entire fleet in one afternoon.

None of that makes the guarantee worthless. It makes it narrower than the buyer assumes. If your operation has any of these patterns — and most do — then the honest way to price the guarantee is to discount it by the fraction of failures you expect to be self-inflicted. I usually model that as a haircut, sometimes a large one, and I would rather haircut it myself than learn the number from a denied claim.

A second source often transfers more risk than a stronger clause

There is a version of this that does not involve negotiating at all. Qualifying a second pack supplier who can hit the same form factor, the same voltage and CAN protocol, and the same mounting envelope means that when a wave of replacements hits, you are not waiting on one vendor's allocation. It also means a single supplier's line problem, cell supply issue, or quality escape is not a fleet-wide outage.

It is not free. It costs engineering time, a qualification build, a pilot, and usually some amount of inventory you would not otherwise carry. But a dual source addresses the part of the risk that guarantees structurally cannot: the concentration risk. One supplier's guarantee covers one supplier's failures on one schedule. A second source covers the schedule itself.

The clustering nobody puts in the model

One more thing, and it is the one that turns a reasonable spreadsheet into a bad quarter. Packs built from the same cell lot, running the same duty in the same building, will fade together. Your model shows replacement events spread evenly across eight years. Reality delivers them in a clump, because the fleet was commissioned in the same month and everything crossed the threshold within the same six weeks.

That clump hits the guarantee too. A warranty that promises a replacement within two business days is a promise about a single event. A fleet of two hundred units arriving at the threshold in the same month is a different problem, and no clause I have read solves it. What solves it is having packs on your own shelf, or a depot within a day's ground shipping, or a second source, or a staggered commissioning plan that spreads the aging out on purpose from the day you deploy.

So the model has a row for that. Concentration risk, replacement lead time, and the number of events I expect in the worst quarter rather than the average quarter. It changes the answer. It usually changes it more than the percentage does.

The guarantee is worth modeling. It is just worth modeling as one line among six, not as the headline it is sold as. Convert it to dollars, put it next to the labor, the freight, and the downtime, and let the arithmetic tell you whether the clause you are fighting for is worth more than the shelf you could build instead.