How to Evaluate BESS Manufacturers: 6 Checks From a Buyer Who Learned Them the Hard Way

If you want to cut a BESS manufacturer short before the first pricing call, ask for three things: a UL 9540A thermal runaway test report matching your exact cell and rack configuration, a warranty document that names who actually performs service, and an OEM or private-label agreement specifying where your brand goes on the unit. Everything else — the per-kWh number, the vague story about 'supplying global OEMs', the PDF render that looks like a factory — can wait until those three are confirmed.

I sit on the procurement side. Eight years, roughly 60 battery and storage purchases, 14 significant misjudgments, about $230K in waste by my own tally. Half of it was mine. Here is the checklist that came out of it.

The one that hurt most happened in September 2022. A commercial and industrial rack deployment. The supplier sent a stack of UL 1973 certificates on the first call. Looked serious. Our EPC asked for thermal runaway data and got six weeks of silence. When the report finally showed up, the spec did not match. The project slipped 11 weeks, and between demurrage, a rack redesign, and expedited retesting, we burned about $48K — plus a quarterly review I would rather forget.

1. Ask for UL 9540A, not 'UL certified'

UL 1973 covers cells and battery systems. UL 9540 covers the full energy storage system including power electronics. UL 9540A is the actual test method for evaluating thermal runaway propagation. They are not interchangeable, and most suppliers mix them together in the same sentence.

When someone says 'UL certified', what they usually mean is the cell passed. Not the rack. Ask three specific questions: which cell model the report covers, which rack configuration was tested, and what revision of the test protocol was used. Tier-1 manufacturers like Samsung SDI do publish cycle-life and cell-level data — but you still have to check which protocol revision the report corresponds to rather than assuming.

2. Read the warranty as a contract, not a promise

The clause that matters is not the coverage period. It is the execution path. How far away is the warranty holder? Is labor included? Who pays freight on replacement parts after a failure? Many terms quietly require service performed by their own certified technicians — which, if your site is far from their network, converts the warranty into a piece of paper.

We got caught on this in Q1 2024. Cheapest bid, signed, then noticed the certified-technician clause. We had none in the region. Paid $31K extra for third-party inspections — essentially just to keep the warranty valid.

One more distinction worth naming: '10-year warranty' and '10-year non-declining, capacity-percentage, conditional on annual remote audits' are not the same document. Read page by page.

3. OEM and private label are two separate conversations

OEM is product carrying their brand or an agreed model number. Private label is the product carrying yours. The fine print on the second one is where things go sideways. Precisely: are you labeling modules, the nameplate, or the whole cabinet? Whose name is on the warranty card? If a customer calls the number printed on your card, who answers?

What I mean is that these are not small details — they affect packaging, manual text, compliance labels, and every certification document an auditor will ask you to produce the first time something goes wrong, and if they are not written down at the outset, you will spend weeks re-working them later. Twenty minutes on this check saved us a three-week legal back-and-forth.

4. Cell specs are not system specs

Most buyers stare at cost per kWh and completely miss C-rate, depth-of-discharge assumptions, and the temperature derating curve — the three parameters that determine whether the 2 MWh you bought is actually 2 MWh or '1.6 MWh on a good day'.

If you are sourcing PHEV lithium battery cells rather than stationary storage, the test protocol is different, but the evaluation logic is the same. Ask what C-rate the rated capacity was measured at, and at what temperature. If the answer is vague or the conversation changes subject, the number is probably a lab best case.

5. Fold fire code and AHJ cost into the comparison

If you are deploying in the US — whether the source is Samsung SDI's US BESS channel or anyone else's — the AHJ is usually the party that decides whether the project energizes on schedule. A rack quote and a rack quote that already accounts for NFPA 855 separation requirements, suppression, and permit-review timing are not the same number.

Add that line to your comparison sheet and the cheap quote stops looking cheap.

Also: be suspicious of any supplier hinting that the fire review is a formality. It is not, and no one else can absorb that risk for you.

6. Verify environmental claims against the FTC Green Guides

The moment a supplier drops 'our batteries are recyclable' into a deck, the FTC Green Guides apply. Under 16 CFR Part 260, a recyclable claim generally needs substantiation — in practice, that means access to recycling facilities for at least 60% of consumers in the area where the product is sold.

Companies like Samsung SDI usually have documentation ready for this. Smaller suppliers frequently do not. Ask for the recycling partner, the geography covered, and the date. If they cannot produce it, the sentence was marketing, not a claim.

Where this order does not apply

To be clear — none of the above is universal.

On a small pilot (say, under 100 kWh), UL 9540A may not be mandatory unless your AHJ is unusually strict — Chicago and New York, for instance. If the supplier is already on your own approved-vendor list, the warranty-execution step can usually be skipped. If your team has strong storage engineering depth, I would actually reverse the order and lead with system integration and communications protocol.

Outside the US, IEC 62619 and IEC 62933 carry more weight than UL, so map the standards to the market you are deploying into.

One last thing: systemizing this — writing it down as a checklist instead of re-deriving it from memory each time — was the single highest-leverage change we made to our sourcing cycle. It has caught 47 potential issues in the past 18 months. Honestly, the best part is not the money saved. It is not having to clean up someone else's project because of a question I should have asked up front.