What Your Custom Robot Battery Warranty Leaves Out Before You Sign

A custom lithium-ion pack warranty for a robot, AMR, or AGV is a risk-allocation document, not a lifespan promise, and the order you read it in decides whether you find the risks or skim past them: start date and duration first, then the capacity fade guarantee together with the test conditions attached to it, then the exclusion list in full, then the claim and RMA mechanics, and only then the pass-through question of whose warranty you are actually holding. The big number on the cover page is the least informative thing in the file.

I have read a lot of these documents, and the one that sticks with me ran to about thirty pages, with the sentence that decided the whole deal buried in a footnote in the last third of it. So pull the file up. I want you to go through it in a different order than last time.

Most people scroll to "24 months" or "36 months," nod, and move on to price. I skip the duration line until I have found the start date, because the start date is where coverage either begins or quietly evaporates.

Find the start date before you look at the duration

Months from shipment. Months from installation. Months from first power-up or first use. I have seen all three of those formulations land in the same procurement cycle, and the difference is not cosmetic.

Picture a pack that ships in the first week of January, sits in a container and then in customs and then on your receiving dock for six weeks, and then waits another month in kitting while the robot build slips. Under "24 months from shipment," that pack has spent roughly a quarter of its coverage before it ever took a load. "From date of manufacture" is worse, because the clock started before you placed the purchase order.

If you cannot get "from installation" or "from first power-up," get a documented ship date and a defined maximum storage period that you control. Then check the storage exclusion. Storage outside specified temperature or humidity conditions is a standard carve-out, and it is exactly the clause that can convert a slow program into an uncovered one without anyone making a decision about it.

The fade guarantee is a set of conditions, not a percentage

Almost every pack warranty splits manufacturing defects from normal capacity fade, and the two do not share claim rules. The defect side usually has a defined window and a remedy. The fade side has a threshold, and that threshold is where the real negotiation lives.

"80% retained capacity at 1,500 cycles" reads like one number. It is four or five numbers wearing a trench coat: the cycle count, the temperature at which that count is measured, the depth of discharge, the charge rate, and the definition of a cycle. Ask what a cycle is. If your fleet opportunity-charges during every idle window, your operation produces a great many shallow partial cycles, and whether those aggregate into full-equivalent cycles, and how, determines how quickly you arrive at 1,500.

Then ask the question most people skip entirely: how is retained capacity measured when you file a claim? At what temperature, at what rate, on whose equipment, and who pays to ship the pack there? A fade guarantee tested on the supplier's bench after the pack comes back is a very different instrument from one tested on site against a protocol both sides signed before the first unit shipped. Neither is automatically better. The second one is easier to actually use, and usability matters more than the percentage.

Ask about the remedy at the threshold, too. Some fade clauses are binary, some prorate. A prorated credit toward a future purchase is not a replacement pack, especially if you are at the end of a program with no future purchase to apply it to.

The exclusions are the actual warranty

This is where I stop reading promises and start reading the list. Every exclusion is a probability claim about how these packs fail in the field. The usual set:

• water ingress beyond the rated IP level
• unauthorized firmware or configuration changes
• over-voltage and under-voltage events
• physical or impact damage
• storage outside specified temperature or humidity limits

For an AMR that gets washed down in a food plant, or one that works a cold-storage aisle and forms condensation every time it crosses the threshold, the IP exclusion is the one that will be argued about at claim time. For a program where your integration team touches the CAN configuration during commissioning, "unauthorized configuration change" is a landmine, and it is far better to define in writing what counts as authorized before anyone connects a laptop.

Read over-voltage and under-voltage as pointer clauses. They point at the charger, the charging strategy, and the protection settings, most of which you or your integrator specify. If your charger does not match what the pack's protection was designed around, you have created a failure mode the warranty explicitly does not cover, and the supplier will be right about that. That is a system-level gap that a warranty cannot close for you.

Physical damage is the ambiguity I push hardest on. Dents happen on AMRs; it is a warehouse. Ask who determines whether a dent is cosmetic or damaging, and whether the answer is a person, a photo, or an inspection report.

DOA windows, freight, and what "remedy" actually means

Out-of-box terms are short, specific, and worth arguing about, because they cover the claims you are most likely to file in the first ninety days. What starts the window: delivery, installation, or first power-up? Who pays return freight, and who pays freight on the replacement? What is the remedy, repair, credit, or a new unit? A credit only helps if you have somewhere to spend it, and on a small fleet you may not.

Then the clause almost nobody reads: does a replacement restart the warranty, or inherit the remainder of the original term? Both are defensible positions. But if a pack fails at month twenty of twenty-four and the replacement inherits four months, you have effectively been told the second pack carries no coverage at all.

Pass-through is the question that decides how much of this you own

A pack supplier is usually an assembler. The cells come from someone else. The important question is not "do you warranty the pack," because everyone says yes. It is whether the assembler can bind the cell manufacturer's terms to you, or whether the cell warranty runs only to the assembler and stops there.

If it stops there, a cell-level defect is a claim against the assembler's own balance sheet and goodwill rather than against the cell maker. That can be perfectly fine. It can also mean the assembler's commitment is capped at whatever they are willing to absorb, and that cap is not written in the document you signed. Ask directly: what happens internally if the cell maker declines a claim? Ask what cell lot traceability they keep and for how long, because without it, any claim that depends on identifying a bad lot becomes a conversation about somebody's memory.

I do not expect every supplier to hand over a flow-down warranty. I do expect to know which one I am buying, because it changes how I price the residual risk and how much spare capacity I hold across the fleet.

Your claim will be decided by records you are not keeping yet

Most claim processes require evidence: cycle counts, charge and discharge history, temperature exposure, and lot information. That evidence lives in your fleet telemetry, not in the supplier's file. The moment to decide what you retain is commissioning, not the week the first pack underperforms. Sit with whoever owns your fleet data and read the claim form out loud, line by line, and ask whether you can produce each item for a specific serial number going back two years. If any line comes back as a no, either fix the retention or accept that your first claim is going to be a negotiation.

Change control and end of life

Robots outlive cell generations. Ask what happens to coverage when the cell or the BMS revision changes. A like-for-like replacement that keeps the same part number usually preserves the warranty. A substitution that is "form, fit, and function equivalent" but carries a new part number often does not, and the first time you find out is when you file.

Get the end-of-life terms on the record as well: how much notice before a part is discontinued, how long a last-time-buy window stays open, and whether spares bought in that window carry the original warranty or a shortened one. For a fleet you intend to support for eight to ten years, the EOL clause is a continuity term as much as a warranty term.

What I do before signing

I put warranty terms into the RFQ as a table rather than a paragraph, so every supplier answers the same questions in the same order: start date trigger, duration, fade threshold and its test conditions, remedy type, DOA window, freight responsibility, replacement warranty treatment, cell pass-through position, required claim evidence, change control treatment, and EOL notice. Then I score them side by side. The exercise exposes the suppliers who write broad promises and narrow mechanics, and it usually surfaces one term that matters more than price at the margin.

The honest caveat: you will not win all of these. Some suppliers will not move on pass-through. Some will not accept an on-site fade test. Some will not restart the clock on a replacement. That is not a reason to sign blind; it is a reason to know what you are self-insuring. The trade I will make nearly every time is a shorter, cleaner warranty with a defined start date and a remedy I can actually use, over a longer one with an ambiguous clock, an untestable fade number, and a claim process that ends in the supplier's favor by default.

The number on the cover is the headline. The document underneath it is the deal.