The Number Everyone Anchors On
When I built our first BESS sourcing spreadsheet in 2022, it had two columns. Column A was dollars per kilowatt-hour. Column B was lead time. That was the whole model. Two columns for an asset that would sit on our books — and our customers' sites — for the better part of two decades.
Four years and roughly $4.6 million in cumulative battery and balance-of-plant spend later, that spreadsheet has 23 columns. The $/kWh number is still in there. It's column M now.
If you're in energy storage system sourcing right now, you probably know the feeling. You've got quotes back. The spread between the highest and lowest is wide enough to matter. The lowest one looks like an easy win for your budget, and your job is not to overpay. So you push on price, you negotiate lead time, and you sign.
That's the problem you think you have. It isn't the one that'll cost you.
The Problem Underneath the Problem
I didn't fully understand what "compliance included" meant in a supply agreement until a 2.4 MWh order sat in our warehouse for eleven weeks waiting on a single document. Nobody was lying to us. Nobody was even being sloppy, really. We just hadn't defined the word.
That order is where this article comes from. Because the gap wasn't in the price. It was in everything the price was supposed to cover.
Specifications that aren't specifications
A one-page sheet listing nominal capacity, cell chemistry, and an IP rating is not an energy storage system specification. It's a business card.
What's usually missing:
- The C-rate at which capacity was measured. A 5 MWh nameplate at 0.5C and a 5 MWh nameplate at 0.25C are not the same product, and they don't behave the same on a hot afternoon.
- Auxiliary load. HVAC, controls, and thermal management can pull 2–5% of nameplate. Almost no summary sheet lists it where you'll see it, and it runs every hour of every day for fifteen years.
- The degradation curve, and the assumptions underneath it — ambient temperature profile, depth of discharge, cycles per year.
- Round-trip efficiency measured at beginning of life, midpoint, or end of warranty. All three get called "efficiency" in the same conversation.
Everything I'd read about evaluating BESS vendors said to compare round-trip efficiency first. In practice, the number that moved our ten-year cost model more than anything else was auxiliary load — and it was the number we had to ask for twice. Twice from the same vendor. That should've told me something.
Compliance scope, which nobody scopes
This is where energy storage sourcing diverges from almost any other equipment category I've bought.
In the US market you're dealing with UL 9540 for the system listing, UL 9540A for thermal runaway test data, UL 1973 for the modules, NFPA 855 on the installation side, and IEEE 1547 on interconnection. Export deals pull in their own stack — IEC 62619 and IEC 62933 show up constantly.
The certificates are issued to a configuration. That's the part that got us. If the cell supplier changes, or the enclosure changes, or a module gets a revision, the listing may not travel with it. So the question worth asking isn't "are you UL certified." It's "what exactly is listed, and who owns the paperwork when it changes?"
Very few suppliers volunteer that answer unprompted. The ones who do tend to have already thought through their own BESS compliance requirements at a level most of their competitors haven't.
Integrated versus component sourcing — and why it isn't obvious
There's a second layer here that took me two cycles to see. Buying a container, a PCS, and a rack of modules separately can look cheaper on paper. Sometimes it is. But then you own the integration risk, and the system-level listing becomes your problem, not the vendor's.
The trade-off isn't price. It's where you want the grey area to live. Some teams are set up to manage that. Most distributors aren't, and neither are most project developers at the scale where they're buying five or ten units a year.
The warranty you didn't read closely enough
Battery warranties in this category are usually throughput guarantees, not capacity guarantees. That's fine. It's also where the money hides.
Things to pull out of the fine print: the augmentation schedule and who pays for it, whether the truck roll is included, the availability guarantee — 95–98% is normal, and the gap between those two numbers is real revenue — and what voids coverage.
We saved about $48,000 on paper in 2024 by taking the bid that stretched augmentation from year four to year six. That decision cost us roughly $165,000 when the mid-cycle augmentation landed anyway, plus crane time, plus re-commissioning, plus three weeks of a site running at reduced capacity. The word "recommended" in that augmentation clause did a lot of quiet work in that contract.
What This Actually Costs You
I'm not going to pretend I have one clean number for the total damage. Some of it's soft. But here's what four years of our procurement records show.
The delay tax. Eleven weeks of a 2.4 MWh system sitting in our warehouse is storage, insurance, and a customer relationship taking damage in real time. We ate liquidated damages on that one.
The re-procurement penalty. When something fails at the AHJ stage, you're not renegotiating from your original position. You're buying spot, on a deadline, from whoever can move. That's the most expensive way to buy anything.
The soft costs nobody invoices for. Engineering hours chasing documentation. Two of our people on calls with a supplier's compliance team for six weeks. That labor is real even though it never shows up in a comparison sheet.
And the one that's hardest to see: we passed on a bid that was 9% higher in 2023 because the $/kWh looked wrong for our segment. That project commissioned on schedule. The cheaper alternative we went with instead had a compliance package that took four months longer to close out. I think about that 9% differently now. Granted, 9% is 9% — budgets are real and I'm the one who signs the PO. But I've stopped treating the sticker as the whole story.
What We Changed
The fix isn't complicated. It's just unglamorous, which is exactly why most sourcing teams skip it.
We built a 22-line specification template and send it out with the RFQ. Vendors respond to our sheet instead of handing us theirs. That one change killed most of the apples-to-oranges problem in the first round.
We turned the compliance document set into a gate rather than a follow-up. Substantiation matters here — the FTC's guidance on advertising claims is written for marketers, but "show me the evidence" is a reasonable standard for a spec sheet too. If the UL 9540A summary and the listing configuration aren't provided up front, the vendor doesn't get a second meeting.
We built a five-line TCO model: capex, ten-year auxiliary load, augmentation spend, expected availability shortfall, end-of-life handling. It's not sophisticated. It's just more honest than column A.
And we added a scope map. One page per vendor: what they do, what they don't, who they hand off to. When we rebuilt the shortlist in 2025, Powin was one of the names that cleared it — and not because of the number on the sheet. The documentation package answered fourteen of our twenty-two line items without a follow-up call, and the compliance scope was described in plain terms instead of marketing language.
That scope map has been worth more than any price sheet. The vendor who told us "this isn't our strength — here are two firms who do it properly" is the vendor we've kept. Not because they're cheap. Because they told us where the edges were, and that made everything inside those edges easier to trust.
A supplier who can't describe their own limits probably hasn't found them yet. You don't want to be the one who finds them for them.