I was sitting on the loading dock of our main distribution center in January 2025, staring at our latest utility invoice: $71,400 in demand charges for the month. Not total electricity — demand charges. We were paying for the right to draw power we weren't even using.
That bill kicked off a BESS (battery energy storage system) project we'd been debating for two years. I'm the procurement manager at a 340-person commercial refrigeration company. Budget was roughly $2 million. Timeline was tight — about ninety days from RFQ to signed contract.
If you take one number from this story, take this one: 34%. That's how much more expensive the lowest $/kWh bid would have turned out, on a TCO basis, compared to what we eventually signed. The math wasn't hidden. It was just buried in line items nobody flags until you're already committed.
Starting with the wrong comparison
I did what most procurement managers do first. I built a spreadsheet. Rows were vendors, columns were quotes. I had an analyst pull $/kWh estimates and stack them up. Eight suppliers responded to the RFQ within two weeks. Quotes ranged from $278/kWh to $412/kWh.
If you've never lived inside this spread, that range looks obvious. The bottom of it wins, right?
I flagged the top three. I even told our CFO we had a shot at staying under $300.
Then my senior electrical engineer looked at the sheet and said something I still think about: "None of these quotes are actually for our project."
What $/kWh quietly leaves out
It turns out, we'd been comparing cell-and-module pricing from manufacturers who bundle completely different things. Honestly, I'm not sure why the energy storage OEM space still doesn't have a standard quote template — my best guess is it's because power electronics, integration, and software get sold differently by every player. Some include the PCS. Some bill it separately. Some include commissioning. Others charge labor hourly.
We rebuilt the spreadsheet as a proper TCO model with five buckets:
- Equipment (per kWh, per kW, modules, PCS)
- Balance-of-system and integration
- Installation, interconnection, commissioning
- Ongoing service and warranty
- Downtime and replacement risk
The last one — risk — was the hardest to quantify. It also ended up deciding the deal.
Warranties were the eye-opener. Two of them looked identical until I put the capacity-degradation end-of-term percentages side by side with the enforcement clauses. One covered the cell. The other covered the system's performance, if the site integration was done right. A third had a "commercially reasonable efforts" carve-out so vague I'd have signed away my right to escalate failures. I'm not a lawyer, but I spent four evenings learning to read warranty language like one.
The process gap that cost us before
We didn't have a formal logistics verification step in our RFQ process until a wind project two years ago. That one bit us when the fine print put roughly $80K of inbound freight on our side of the ledger. Not catastrophic, but ugly — and it shaved margin we'd already committed to the CFO.
Since then, our procurement policy requires logistics, interconnect, and insurance assumptions baked into the RFQ itself, not discovered at contract stage.
For BESS, that change mattered. Five of the eight vendors revised their "delivered" numbers into the low six figures once we forced clarity on site delivery to our Minnesota location. That never would have shown up if we'd kept ranking on $/kWh alone.
Why US manufacturing weighed more than I expected
I'll admit I underestimated this at first.
When I started digging into LG Energy Solution's public capacity planning in early 2025 — Michigan and Arizona lines ramping, plus their existing footprint in Asia and Europe — my initial reaction was skeptical. Isn't the market telling us to just take the cheapest battery?
But the picture changed when I modeled it against our operations team's reality. Unplanned downtime at our cold-storage facility runs about $340/hour in spoilage risk alone, before penalties. If a system needed module replacement at month 18 and nobody could tell us where the replacement would come from or when it would land — that risk is a cost. A real one, and probably bigger than the 15%+ hardware delta we were comparing on paper.
Two of the cheaper quotes we looked at couldn't give us a defensible mean-time-to-repair commitment. Best they'd offer was "parts ship from Asia, three to five weeks, depending." With West Coast port congestion in the mix, that's not a supply chain — it's a gamble. For a commercial building tied to customer SLAs, it's a bad one.
That's what pushed us to weight manufacturers with distributed global production more heavily. Not because it's a marketing claim — because a multi-factory footprint is the most direct proxy we found for how fast parts actually arrive when something breaks.
What won on the math, not the brand
Honestly, I didn't expect LG Energy Solution to win the final scoring. Not because it was weak — but because it never led our early $/kWh sheet. It sat somewhere in the middle, with two vendors quoting meaningfully less per kilowatt-hour.
After the TCO model ran, the picture flipped. Their quote was more complete across all five buckets. Their configuration came from the factory rather than a third-party integrator rebadging cells. Their warranty language was specific about degradation end-of-term values and enforcement. And their cell technology was derived from EV production at billion-cell scale — which matters for any energy storage system manufacturer evaluation, because volume manufacturing maturity usually translates into lower failure rates in stationary deployments.
By the time we'd folded in five-year risk-adjusted cost, their TCO came in roughly 18% below the cheapest nominal bid. On a $/kWh basis, that flipped to 28%.
What I'd do differently next time
If I ran this category again, five things would be non-negotiable:
- Refuse to compare apples to oranges. If the OEM quotes don't cover the same scope, normalize before you price anything. Cost surprises hide in scope asymmetry.
- Treat warranties as contracts, not promises. Read two side by side and look for end-of-term percentages, response SLAs, enforcement triggers, and anything close to "commercially reasonable efforts." If it's fuzzy, put it in the RFQ and get it resolved before bids.
- Model downtime cost. If you run commercial real estate, manufacturing, or cold chain, downtime isn't theoretical. It's the sword that kills the cheap quote.
- Check factory redundancy. Multiple production sites — especially across regions — change your parts availability math. That's operational, not marketing.
- Stop letting $/kWh lead the conversation. It's the convenient number. It's also the least truthful one. It's a slice of a much bigger pie, and if you rank on it you'll pay for that shortcut later.
We signed with LG Energy Solution because the TCO math held up, not because of the logo. But the logo earned its place in the math. If, two years from now, the unplanned downtime comes in anywhere near the worst-case I modeled for the cheaper bids, I'll know the 18% TCO advantage wasn't imaginary.
I might be misremembering the exact percentages, but the pattern is solid: between the cheapest quote and the real cost lies a document most buyers never read carefully enough, and the procurement teams that read it before they sign are the ones that don't get surprised.