How Battery Suppliers Price the Risk Behind a Capacity Fade Guarantee

A capacity fade guarantee is a priced risk transfer, not a free promise. Any supplier quoting one has already estimated how many packs it expects to fail or fall below threshold over the term, multiplied that by what a replacement costs to build and ship, and folded the result into the unit price. The real question in an RFQ is not whether a fade guarantee is available. It is how much risk is being bought back, what the supplier charges to carry it, and whether the guarantee's test conditions describe your robot or the average one in their installed base.

I have spent years on the buyer's side of that quote sheet, and the moment it became obvious was the second tab.

There is a pack quoted for an AMR fleet on the table between us. The first tab is what everyone looks at: cells, BMS, enclosure, connector, first-article test. The second tab took two emails to get, and near the bottom there is a line labeled "warranty" with a number against it, and above that number, four rows of small print about cycle counts, ambient temperature, and how capacity will be measured. The engineer beside me wants to argue about the CAN protocol. I want to stay on that line, because it is the only place on the whole sheet where the supplier has written down what it thinks is going to go wrong.

That is what a fade guarantee is. It is a failure forecast with a price on it, sold back to you.

Where the reserve hides

Ask for the number behind the number. What you want is the accrual: expected replacement rate across the term, times the loaded cost of a replacement, divided by units shipped. Loaded means materials, labor, freight in both directions, diagnosis time, and the administrative cost of processing a claim. A supplier with a real quality system can walk you through how it builds that. A supplier that cannot tell you anything about it is either pricing off a gut feel or pricing off somebody else's fleet.

Either way, the number is in the price. When a warranty line is not shown separately, it is in the margin and it is bigger than it looks, because the supplier is also reserving against the failures it did not model. That is the part buyers forget: the reserve covers the uncertainty around the defect rate, not just the defect rate.

And there is a second layer they will rarely say out loud. The reserve is priced to a pool. The supplier cannot model your warehouse in July or your duty cycle with a thirty-second opportunity charge at every stop. So it prices to an average application and an average ambient. If your fleet runs gentler than the pool, you are subsidizing somebody else's hard duty cycle. If it runs harder, you look cheap at purchase and expensive at claim time, and the correction arrives at renewal in the form of tighter terms or a different number on that line. Making your duty cycle legible to the supplier before they price is the whole game.

The dials they turn instead of the price

When a supplier wants to win your business without cutting the number, it moves one of four things. Watch for these in the order they usually show up.

The threshold

Ninety percent of nameplate sounds better than eighty percent of beginning-of-life measured capacity, and it is a weaker promise, because nameplate is a nominal figure the actual pack may have beaten on day one. A threshold against a beginning-of-life reference performance test measures degradation. A threshold against nameplate measures a marketing number. I want the threshold written against the measured capacity of that specific pack on the day it shipped, and I want to know how that measurement was taken.

The test conditions

This is where most guarantees lose their value. Capacity measured at what C-rate, at what temperature, across what voltage window, after what rest period, on what equipment, by whom? A pack tested slowly at 25 degrees C in a lab will show capacity the machine never sees at 40 degrees C under a real load profile. If the guarantee does not pin the test to something close to your operating envelope, you have bought a number about a lab, not about a robot.

The throughput limit

A five-year guarantee capped at a cycle count or an energy throughput is not a five-year guarantee. It is a cycle guarantee that happens to mention a calendar. Run the arithmetic on your own fleet: cycles per day, days per week, and how long it takes to reach the cap. I have watched a fleet hit its cycle cap in well under half the stated term, which means the back half of the guarantee was decorative.

The exclusions

Ambient temperature band, charge rate limits, depth of discharge, storage conditions, and the clause about damage from misuse. Read the temperature band as a sentence about your building. If the exclusion is written for a range your mezzanine exceeds every August, you bought a guarantee for a different application.

Two quotes for the same pack, at two terms

Here is the technique I lean on most, because it does not require the supplier to reveal anything confidential.

Ask for the same pack, same volume, priced at three terms: a short one, the one you actually want, and a long one. Everything else held constant. Then look at the slope between them.

That slope is the supplier's own failure estimate, priced. If the difference between a two-year and a five-year guarantee is small, they genuinely believe the packs will last, or they have drafted the terms so few packs will qualify. If the difference is steep, they are telling you what they think of the design. Either way, you have extracted their forecast without asking them to hand over their data.

Ask also for the same pack at two capacity thresholds. That second slope tells you what they think about the tail of the distribution, which is the part that actually costs you a truck roll.

What the money does not buy until you write it in

A fade guarantee covers one failure mode, slowly. Everything else happens fast, and it is governed by the other clauses.

The DOA window matters more than it looks. A pack that arrives dead, or that fails in the first weeks, is a freight and scheduling problem, not a capacity problem. Who pays freight in, who pays it back out, and how many days the supplier has to turn a replacement. If they diagnose for three weeks while your robot is down, the guarantee did nothing for you.

RMA terms are the same question in a different place. Who owns the diagnosis when the pack comes back and the supplier finds no fault? That phrase, no fault found, is where a fair amount of money and goodwill quietly goes. Get the answer in writing: if the returned unit passes their test, who eats the freight and the labor.

Then escalation. When the same failure shows up on five units in a month, what happens? I want a response time, a containment step, and a root cause report with a date attached, not a promise of a thorough investigation. The root cause report is what protects the rest of the fleet, because it tells you whether to change the duty cycle, the thermal design, or the supplier.

A longer guarantee is not a braver supplier

It is tempting to read a long term as confidence. Sometimes it is. But it is equally consistent with a high risk premium, with terms drafted so tightly that few claims qualify, or with a supplier whose commercial plan does not extend as far as the guarantee does. Length on its own tells you nothing. The informative comparison is always length against price against terms, held on the same pack.

What I ask for now

One page in the RFQ, every time. Show the reserve as a line if you can. Quote the pack at three terms. Quote it at two thresholds. State the capacity measurement method, the equipment, the temperature, the C-rate, and whether the reference is beginning-of-life measured or nameplate. State the cycle and throughput caps as numbers, not adjectives. State DOA days, RMA freight, no-fault-found responsibility, response time, and root cause reporting. State the temperature band, then compare it to the actual temperature log from the site.

When the numbers come back, do not try to negotiate the reserve to zero. It will not go to zero. It will go somewhere, and if it leaves the price it comes back as slow claims, tight interpretations, and a supplier who is no longer interested in your account. What you can do is change what they are forecasting. Better test conditions, a threshold tied to real beginning-of-life data, exclusions that match your building, and a duty cycle they have actually seen. That is how you move the number honestly.

The guarantee is not a promise the supplier makes to you. It is a forecast they make about you, and they charge for it either way.