What Voids Your Robot Battery Warranty Before the First Failure

Most robot battery warranty claims are lost before the pack ever fails. Water past the rated IP level, a charger nobody approved, a pack parked at the wrong state of charge through a three-week shutdown, a housing opened on a maintenance bench with a torque driver: those are the events the warranty document lists as exclusions, and the failure analysis report is what ties the field event back to that list. A pack can be genuinely defective and still land entirely on your balance sheet, because the denial is decided by what happened around the pack, not by what happened inside it.

I have sat in that meeting. Failure analysis PDF open on the table, a photograph of the pack's connector pinned to page four, and a service manager across from me explaining that the log shows eleven charge cycles outside the approved profile before the cell group went soft. The cells were bad. The claim was still dead. Nobody in the room was arguing about lithium chemistry. We were arguing about a hose and a charger.

That is the whole game. So I stopped reading warranty documents as legal documents and started reading them as operating instructions.

The four scenes that produce most denials

When I walk a site after a claim, I am not looking at the robots. I am looking at four places, in order, because those are where the exclusions live.

The wash-down bay

Food and pharma sites are the worst offenders, and it is not the operators' fault. Someone writes a cleaning procedure that says the floor gets hosed at the end of shift. The AMR either parks in that zone or crosses it. The pack is rated IP54. A hose at forty degrees from a foot away is not what IP54 was tested against.

Stand in that bay with me for a second. There is water on the floor, a nozzle hanging on the wall, and a robot sitting in a puddle because that is where its charge contact is. The exclusion clause is one line. The scene produces it every single night.

What you do about this is not exotic. The pack spec in the RFQ needs an IP rating chosen against the cleaning procedure, not against the marketing category. If the hose is non-negotiable, the enclosure rating is a procurement requirement, and the supplier should be asked what spray pattern and duration the rating was validated at. If the answer is a shrug, you have found the clause you will be arguing about in fourteen months.

The charger bay

The charger is the most common self-inflicted wound I see. The OEM's charger goes on a fourteen-week lead time, somebody finds an equivalent on a shelf, the connector fits, the robot charges, and the fleet runs fine. Until the claim.

Charging outside the approved profile, or on a charger that is not on the approved list, is one of the cleanest exclusions a supplier can write, and it does not require them to prove the charger caused anything. There is also the less obvious version: an approved charger run at the wrong amperage because someone bumped the setting to shorten a turnaround, or opportunity charging pushed into a duty cycle the pack was sized for at a gentler rate. Same clause, different cause.

I handle this with boring mechanics. A named charger model list in the purchase agreement. A physical lockout on the charge contacts so a foreign connector cannot mate. A site rule that any charger change goes through whoever owns the warranty, written into the contractor handbook rather than into an email nobody reads.

The shelf

Storage is the exclusion people forget exists until it bites them. Packs sitting at low state of charge in a warm spare-parts room over a summer shutdown. Packs stored below the temperature floor in an unheated mezzanine. A spare that sat on the shelf for a year and went out of the SOC window before it ever turned a wheel.

None of that is a field failure. It is a receiving-and-layup procedure that nobody wrote. The fix costs almost nothing: a stated storage SOC window and temperature range from the supplier, a rotation schedule so spares cycle through, and a logging step at goods-in so you can prove the pack was inside the window when it arrived and inside the window when it went out. Two columns in a spreadsheet, honestly.

The bench

Unauthorized disassembly is the one that feels unfair, because it usually is. A technician cracks a housing to look at a swollen cell, or a third-party shop does a connector repair, and the whole pack warranty is void even if the failure turns out to be a weld defect unrelated to anything the technician touched. Unrelated failures do not resurrect a voided pack.

The lever here is training and tamper evidence, not legal language. Seal integrity on the housing, a documented service boundary that says where your technicians stop, and a return path that gets a suspect pack back to the supplier without anyone opening it. I have watched a five-thousand-dollar pack get denied over a two-dollar gasket, and the technician was trying to help.

Fade is not a defect until the contract says it is

This is the piece that gets missed in almost every warranty conversation I have been part of. Cells fade. That is not a failure, and a supplier who never defined the difference will treat any capacity complaint as normal aging and point at the duty cycle.

The warranty schedule needs a fade curve: how much loss is expected at what point under what duty, and where defect coverage begins. That curve has to be tied to a duty cycle, because a pack on a single shift in an ambient warehouse and a pack on a three-shift freezer aisle are not the same product experience, and the supplier knows it.

If the number is not in the document, the number is whatever the supplier says it is on the day you call. I would rather negotiate a fade curve I do not love than walk into a claim with no curve at all.

The matrix, laid out against your duty cycle

This is the artifact I actually build before a purchase order closes, and it is the thing that changes how the warranty reads. Every row is a real operating scene from the site, matched to the exclusion it triggers.

Operating scene	Exclusion it triggers	Who carries it if unmanaged	Procurement lever
Nightly wash-down zone	Water ingress past rated IP	You	IP rating set against the cleaning procedure; spray validation evidence
Opportunity charging at boosted rate	Charging outside approved profile	You	Named approved charger list; connector lockout; rate settings locked
Summer shutdown, spares on shelf	Storage outside SOC or temperature window	You	Storage window spec; rotation schedule; goods-in logging
Technician opens a swollen pack	Unauthorized disassembly	You	Service boundary training; tamper-evident seals; return path
Freezer aisle, charge at low temp	Use below the temperature floor	You	Heated charge position; charge inhibit in the pack spec
Three-shift duty over 24 months	Capacity fade past the guaranteed curve	Shared, if the curve exists	Fade schedule tied to duty cycle in the warranty

The pattern in that last column is not an accident. Almost every exclusion that lands on you has a procurement answer that costs a fraction of the pack. The reason it lands on you is that nobody mapped the scene to the clause while there was still leverage to change the spec.

The claim you lose without a single failed cell

Here is the part that surprised me the hardest. A meaningful share of denials are not driven by an excluded event at all. They are driven by a customer who cannot prove what happened.

Suppliers can require BMS fault logs, charge histories, and maintenance records to demonstrate that a claim is not excluded. That is reasonable, and it is also a trap. If your fleet software retains telemetry for thirty days and you file the claim in month four, you have lost the argument by default. Not because the failure was excluded. Because the evidence window closed before you noticed the problem.

So before the first robot ships, I want four things settled in writing: what data gets captured, how long it is retained, who owns it, and how fast it can be exported in a format the supplier's failure analysis team will accept. I have seen a supplier's cloud hold charge history the customer's own platform never recorded. I have also seen the reverse, and the customer was the one with the defensible file.

Decide which side of that you want to be on, and decide it during procurement, because you cannot decide it during a claim.

What a voided warranty actually costs you

The number most teams use is the price of a replacement pack. That is the small number. The real exposure is the fleet that keeps running on a pack family that just lost coverage, the third-party pack you now have to qualify mid-program, the downtime while an integrator re-validates, and the capacity-fade argument you are having annually instead of once.

I do not try to eliminate exclusions. No supplier will sign that, and a buyer who demands it gets a price that hides the risk instead of removing it. What I want is a matrix where every exclusions row has a named owner on my side, a named owner on theirs, and a lever I already pulled. Then the failure analysis meeting is a short conversation, because the only question left is whether the pack was actually defective.

That is the whole point of reading a warranty like an operating procedure. It stops being a document that arrives with the product and becomes a list of things your site does every day.