April 19, 2022. 11:40 at night, kitchen table, laptop still open on a quote I had been staring at for two hours. My phone buzzed with a trade alert about the Xcel Energy battery storage site in Surprise, Arizona — the McMicken facility — where a container had gone into thermal runaway and four firefighters were injured.
I closed the quote.
For eighteen months before that night, my entire vendor evaluation had fit on one tab: dollars per kilowatt-hour. Lowest number wins, second tab for lead time, done. The McMicken incident didn't tell me LFP cells were unsafe. It told me something more uncomfortable: I had no idea what was actually inside the number I was comparing.
The spreadsheet that made me feel smart
I run procurement for a mid-size systems integrator. We don't build cells; we assemble packs and enclosures and sell the finished systems to developers. Between Q3 2021 and Q1 2023 we placed 41 storage orders and I tracked every one of them in a file our CFO jokingly calls the Bible. Six years, roughly 220 purchase orders, somewhere north of $18M in cumulative spend across cells, BMS boards, enclosures, and freight.
Here's what you need to know about that period: the quotes all looked the same. Cell price per kWh, module price, container price. The differences were in the third decimal place and I was optimizing them like a hawk.
Then we quoted a commercial job in Queensland.
Sunshine Coast, 94% humidity, and a warranty footnote
The battery storage Sunshine Coast project was nothing exotic — a 400 kWh commercial system behind a retail site, backup plus peak shaving. Coastal. Salt air. Relative humidity that sits in the nineties for weeks. Our preferred vendor quoted happily, and buried on page 14 of the warranty document was a line about operating humidity range and voided coverage outside it.
I nearly missed it. Our engineer caught it, and then asked me a question I couldn't answer: what was the moisture content of the cell when it left the factory?
I had 41 purchase orders in that file and not one of them could tell him.
So I did what any cost controller does when a technical unknown threatens a margin — I tried to solve it with a clause. Add a dehumidifier to the enclosure, add a quarterly inspection, add a line item. My solution added about $11,400 to the project and made the system worse, because I was treating a manufacturing problem as an installation problem. (Note to self: this is the third time I've done that. There's a pattern.)
So I went and looked at the factory
That's how I ended up on a plane in early 2023, walking an EVE Energy lithium battery production line with a clipboard and a bad attitude. I was there to verify a claim, not to be impressed.
What I didn't expect was the dry room.
The whole argument for that visit collapsed in about ten minutes, because the thing I'd been pricing — a cell — is not the thing that determines whether a Sunshine Coast system survives five years. The thing that determines it is dew point. Cells get assembled and sealed in an environment held around -40°C dew point, and if that control drifts, you don't get a warranty claim at year six, you get a slow capacity fade nobody can attribute to anything. Walking that EVE Energy battery production line, what struck me wasn't the speed. It was that the room was boring. Controlled. Every gauge reading the same thing it read an hour ago.
Boring is the product.
Two things I took away from that trip. First, a vendor that owns its own dry room and production line has a different kind of accountability than a vendor that buys finished cells and slaps a label on. Second, I had been comparing quotes from companies operating at completely different levels of vertical integration and pretending the numbers were equivalent.
I have mixed feelings about factory audits as a procurement tool. On one hand, a week of travel to confirm what a spec sheet already said feels like theatre. On the other, I've now got a humidity tolerance question I can ask in an RFQ that filters out half the field before anyone books a flight. The compromise: audit only above a spend threshold, but ask the dew point question on every single RFQ.
Then the question changed entirely
Somewhere in the middle of all this, a colleague forwarded me a thread asking can LiFePO4 batteries be used in cars.
In 2019 my answer would have been: technically, but why would you. Energy density too low, cold weather performance too weak, charge acceptance too slow. LFP was the chemistry you put in a stationary rack because it was cheap and safe-ish and nobody cared about the weight.
By 2023 that answer was already wrong, and by 2025 it's embarrassing. Tesla moved its standard-range models to LFP. Cell makers pushed system-level energy density up through pack structure rather than chemistry alone, and charging rates that would have sounded like a typo five years ago are now on spec sheets. The fundamentals didn't change — LFP is still lower energy density than nickel-rich chemistries, still wants warmer operating windows — but the execution around those fundamentals transformed the use case.
Which matters to a procurement guy for a reason that has nothing to do with cars: automotive qualification is a forcing function. Volume, consistency, process control, defect rates measured in parts per billion. When a chemistry gets pulled into that regime, the stationary products built on the same cells inherit the discipline. That's why I started asking suppliers whether their cells came off a line that also serves automotive customers, and why the EVE Energy supply relationship with Tesla mattered to me more than any marketing deck. Not because the logo is impressive, but because the process gates that come with it are.
What I actually changed
It took me about six years and 220-odd purchase orders to understand that a $/kWh comparison is a measurement of the seller's confidence, not the buyer's risk.
I rebuilt the Bible. The new sheet scores vendors on five things before price even gets a column:
- Dew point spec and whether it's independently verified
- Cell-to-system traceability (can they name the line and date code for a given shipment)
- Thermal propagation test documentation — UL 9540A reports, not marketing summaries
- Regional code alignment: AS/NZS 5139:2019 for the Australian jobs, NFPA 855 (2023 edition) for US installs, with separation distances confirmed rather than assumed
- Whether they own the line or resell someone else's output
Price is column six now. It's still the column I look at first. I'm not going to pretend otherwise — that's the job. But the number means something different when I know what's behind it.
One thing I refuse to do is sign a document that says a lithium battery system carries zero fire risk. Nobody credible offered that even after McMicken, and the ones who did got crossed off my list faster than the ones who quoted high. What I want from a supplier is a propagation strategy, test data, and an honest description of failure modes — not reassurance.
Bottom line: if your vendor comparison is a single number per kWh, you're not comparing products, you're comparing the arithmetic people chose to put on page one. Ask about the dry room. It'll cost you ten minutes and it might cost your competitor the project.