Your Battery Warranty Says Cycle Life, but What Does It Guarantee?

A cycle life figure in a battery warranty guarantees almost nothing until the clause it sits in names four things: the cycle count at which capacity is measured, the retained-capacity threshold, the test protocol that produces that measurement, and the remedy when the pack misses it. Most pack warranties that reach a procurement desk carry the number and skip all four.

I keep one afternoon in mind because it's the one that changed how I read these documents. A 48-volt pack off an AMR, lid on, terminals cabled to a load bank, the fans on that load bank running, and beside it a laptop with two things open: the fleet's discharge log and the supplier's warranty, page four. Page four had the sentence that got quoted in the sourcing meeting — "3,000 cycles to 80% of rated capacity." The discharge log showed that aisle's newest packs sitting around 82% at 1,400-odd cycles. So I went back through the page line by line. It never defined a cycle. It never said at what temperature or depth of discharge the 3,000 was generated. It never said who decides a pack has failed, or what I receive if it has. We had a number. We did not have a guarantee.

Everything below is what I do instead, and what I'd ask you to do with the next warranty that lands in front of you.

Read the test conditions before you read the count

The 3,000 came from somewhere, and the somewhere is what you're actually buying. Cycle life is a test result. A test result without its conditions is a slogan. Four conditions do most of the work, and every one of them should be written into the warranty clause rather than left in a report you never see.

Depth of discharge. A full 100% DOD cycle and a 30% partial cycle do not age a cell the same way. Many lab protocols cycle to 80% or 100% DOD. If your robot opportunity-charges at every idle moment, it may accumulate two or three shallow events per shift, and the relationship between those events and a lab cycle is not one-to-one in either direction. A pack tested to 2,000 full-DOD cycles has not been tested against your duty — it has been tested against a different duty that happens to produce the same word on the page.

Temperature. Cells cycled in a controlled 25 C chamber tell you about a chamber. A pack working a trailer bay in Phoenix in July, or a refrigerated aisle, or cycling right after a fast charge that warmed it, is living somewhere else. Heat accelerates fade, and cold charging can plate lithium and cost you capacity permanently. Get the test temperature in the clause, and get the derating expectation you should apply outside it.

Charge rate. If the protocol charged at a gentle fraction of C and your fleet charges at 1C between picks, the test understates the stress. Same story on discharge: the difference between an average draw and the peak draw during a lift or a hard turn is real wear.

End-of-discharge cutoff. Where the test stopped discharging changes what "capacity" means. A pack run to a conservative cutoff reports a different number than one pushed to the voltage floor, and the two packs are not equally degraded.

If the warranty cites a protocol, read the protocol. If it cites an internal one, ask for the document, not the summary slide. If it cites one of the IEC 62660 series or an equivalent automotive standard, at least you know the shape of the test and can ask which of the available conditions the supplier chose to run.

"Cycle" has at least three meanings, and your BMS picks one

This is the part that costs people real money. Ask a supplier what a cycle is and you'll get one of three answers.

A full equivalent cycle by amp-hour integration: total amp-hours discharged divided by rated capacity, so partial events accumulate. A counted event: any discharge-and-recharge transition the BMS logs as one cycle, however shallow. Or a test cycle in the protocol: one discharge at the specified DOD and rate, followed by the specified charge.

The first two are produced by the BMS, and the BMS was configured by the pack builder. A counter set to count shallow events individually inflates fast on an opportunity-charged robot. A counter set to integrate amp-hours tracks physical wear more closely. Neither is wrong. The problem is that the warranty says "cycles" and your fleet's counter and the lab's counter are not the same instrument, so the two of you can look at the same pack and disagree about whether the guarantee has been breached.

So the clause has to separate them on purpose. What I ask for now: cycles accumulated on the BMS define when you're allowed to file a claim. An independent capacity discharge test defines whether the pack has actually failed the guarantee. The BMS counter is a trigger. The test is the evidence. If the warranty lets the BMS state-of-health estimate serve as the failure evidence, you have handed the decision to a firmware algorithm that can drift, get recalibrated in an update, or reset when a pack goes through service.

Decide who measures, who pays, and what happens to the fleet meanwhile

A claim has a physical cost that almost never appears in the warranty text. To prove a pack is below threshold you generally have to take it out of service, ship it, and have it discharged on calibrated equipment. Lithium packs move as Class 9 dangerous goods with UN 38.3 test evidence attached, which means paperwork, a qualified carrier, and lead time. That machine is not moving pods while this happens.

Before any of that, you need a control at the other end: a baseline. "80% of rated capacity" only means something if you know what the pack delivered when it was new and in your hands. Nameplate ratings are optimistic, and rated capacity itself is defined at a rate and temperature that may not match yours. I have watched a program argue for weeks over whether a pack had lost 19% or 22%, because nobody measured it at commissioning.

Two cheap things fix that argument. First, a commissioning capacity test on a sample: pull one or two packs from the first shipment, discharge them at the agreed rate and temperature, record delivered amp-hours, and get the supplier to sign off on that result as the baseline. The guarantee then runs against a measured starting point rather than a printed one. Second, a written claim procedure. Who performs the verification test — a third-party lab, the supplier's lab with your witness, or your own facility? Who pays if the pack passes, and who pays if it fails? What turnaround is acceptable from claim to result? And while the pack is gone, do you get a loaner, or does the machine sit?

Exclusions are where guarantees go to die

Every capacity warranty excludes operation outside specified temperature and current limits, and it should — abuse is abuse. The failure mode is when the exclusion is drawn tighter than the duty cycle the robot actually runs. Three versions of this I have argued about.

A peak discharge limit written as a continuous rating. Robots draw in bursts. A lift or a hard acceleration can pull several times the average current for a few seconds, and your own data log will prove it. If the warranty's current limit is a continuous number and the exclusion reads "operation outside specified limits," the supplier has a file of logs showing you outside the limit on every shift.

A charging window that doesn't cover your site. Charge temperature ranges often start at 0 C or above. An unheated warehouse in a cold climate, or a freezer-adjacent aisle with no place to warm, can put routine charging outside the window without anyone abusing anything.

A storage condition that catches normal idle. If a machine sits unplugged over a weekend and the clause requires a storage state of charge or temperature range nobody can practically hold, that exclusion swallows the guarantee.

The fix is to write the operating envelope into the warranty as the same envelope you specified in the RFQ, and to require that any exclusion match those numbers rather than a separate, tighter set. If the supplier's duty-cycle limits differ from your specification, that is not a warranty problem. That is a pack sizing problem, and it will show up as fade no matter what the clause says.

Ask for arithmetic, not a promise

"Repair or replace at supplier's discretion" sounds generous and budgets like nothing. It gives you no way to predict what a fading fleet costs in year three, and it gives the supplier every reason to find the pack within specification.

A pro-rata remedy is dull and much better. The shape of it: the warranty states the warranted cycle count and threshold — say 3,000 cycles to 80% of rated capacity, measured per the named protocol. If a pack drops below threshold before reaching that count, the credit is proportional to the shortfall. If the pack reaches 2,400 cycles at the threshold instead of 3,000, you received 80% of the warranted life and the remedy covers the missing 20% against a defined replacement price. You know the number before you need it, which is the whole point.

Two details carry that arithmetic. The price the credit is computed against should be fixed at purchase, not floating with list price two years later. And the exclusions from pro-rata — wear items, physical damage, out-of-envelope operation — should be itemized, so the credit doesn't quietly shrink through a clause nobody read.

I also want to know where the clock starts. Date of shipment and date of commissioning can be months apart on a project with a slipped install. If the clock starts at shipment and the pack sits in a warehouse for a quarter, you paid for coverage you never used.

What goes in the RFQ now

A short set of asks, and I've stopped accepting a warranty that doesn't answer them.

The warranted cycle count and retained-capacity threshold, with the protocol named down to depth of discharge, temperature, charge and discharge rate, and cutoff.

A sentence stating that the BMS cycle counter is a claim trigger only, and that failure is established by an independent capacity discharge test.

A commissioning baseline measured on your packs and signed by the supplier.

A named verification lab, a stated turnaround, a rule for who pays in each outcome, and a loaner or spares provision so a claim doesn't idle a machine.

Exclusions that match the envelope in your specification, with peak current handled as peak current.

Pro-rata remedy arithmetic against a fixed replacement price, itemized exclusions, and a start date tied to commissioning.

None of that makes a weak pack strong. What it does is turn a number on page four into something you can measure, prove, and budget — which is the only thing a cycle life figure was ever supposed to do.