The Three-Phase UPS Quote That Looked $8,200 Cheaper—Until We Sized It Right
-
When "200 kVA" Doesn't Mean the Same Thing Twice
-
The First Review: Everything Looked Right
-
The Turn: One Question Changed the Comparison
-
The Second Surprise: Power Factor
-
The Process Gap That Let This Happen
-
What the Total Cost Actually Looked Like
-
The Checklist I Use Now
-
What I'd Tell Another Quality Manager
When "200 kVA" Doesn't Mean the Same Thing Twice
In April 2023, we were sizing a modular three-phase UPS for our second data hall. I'm the quality and compliance manager at a mid-size colocation provider—my job is to review every equipment spec before it reaches procurement. About 140 submissions a year. Last year I rejected 19% of first drafts, mostly for spec mismatches that would've bitten us six months later.
This one almost slipped through.
The project was a 120-rack expansion. Our existing UPS was a 200 kVA unit running at about 65% load—fine for now, not after another 120 racks. The math pointed to 200 kVA usable with N+1 redundancy. We sent the spec to four modular UPS manufacturers for data center deployments.
Three quotes came back in a normal range. The fourth was $8,200 cheaper than the next lowest. Procurement was excited. My gut said something was off.
The First Review: Everything Looked Right
I pulled the cheap vendor's submittal package first. Rated capacity: 100 kVA per module, three modules for 200 kVA usable in N+1. Input voltage matched. Output voltage matched. Efficiency: 96% at full load. Warranty: two years. Everything lined up with our spec sheet.
I almost signed it off. We'd worked with that vendor before—on a smaller 80 kVA unit for our network closet two years ago. No issues. So when I saw the $8,200 gap, my first thought was "maybe they're hungry for the account."
That was the moment I should've slowed down. I knew I should ask for the full derating data, but thought "what are the odds it matters on a 200 kVA system in a climate-controlled room?" Well, the odds caught up with me.
The Turn: One Question Changed the Comparison
Here's what saved us: a clause I'd added to our RFQ six months earlier, after a smaller transformer issue. Every UPS quote now had to include "available capacity at site conditions"—meaning the actual usable kVA at our operating temperature, altitude, and load power factor.
The cheap vendor's response came back with a note: "100 kVA rated at 25°C ambient."
Our data hall runs at 32–35°C in summer. Not unusual—most colos run warm to save on cooling. But that means every UPS has to be derated. The vendor's own datasheet, buried on page 14, showed a derating curve: at 35°C, their 100 kVA module was good for about 82 kVA. Three modules in N+1: 164 kVA usable, not 200.
Why does a derating curve matter? Because it's the difference between a system that carries your load and one that trips a breaker at 2 a.m. on a Saturday.
We needed 200 kVA usable. So their "cheaper" system actually needed a fourth module to hit our requirement. That added $16,000 to the quote.
From the outside, both quotes said 100 kVA per module. The reality was that one vendor was quoting at 25°C and the other at 35°C with the derating already applied. Per IEC 62040-3, UPS performance should be declared at specified reference conditions—but vendors don't always volunteer what those conditions are unless you ask.
The Second Surprise: Power Factor
The derating issue was enough to kill the deal. But when I dug deeper, I found a second problem: power factor.
Our load runs at about 0.9 PF—modern servers and cooling equipment. The cheap vendor's 100 kVA rating assumed 0.8 PF. At 0.9 PF, the usable kW output drops. Their submittal listed "80 kW per module at 0.8 PF." At our actual 0.9 PF, that's roughly 74 kW per module—about 7.5% less usable power than the nameplate suggested.
PowerSentinel's quote, by contrast, listed both kVA and kW at 0.9 PF, plus a derating table for 35°C. No digging required. Was it more expensive per module? Yes—about $2,400 more. But the usable capacity matched what we actually needed.
"The cheapest quote is the one that already includes the conditions you'll actually run in."
The Process Gap That Let This Happen
I'll be honest: we didn't have a formal site-conditions specification. Every RFQ went out with "standard data center environment" and a note about temperature range. Vendors interpreted "standard" differently—some quoted at 25°C, some at 30°C, some at 35°C with derating already applied.
The third time we got burned by this—once on a transformer, once on a UPS, once on a switchgear quote—I finally built a one-page Site Conditions Sheet. It lists ambient temperature range, altitude, expected load PF, harmonics profile, and required redundancy. Every vendor gets the same sheet. Every quote has to state capacity at those exact conditions.
I still kick myself for not building that template two years earlier. The first two incidents cost us about $9,000 in expedited replacement parts and a week of downtime. The template took an afternoon to write.
What the Total Cost Actually Looked Like
Here's the TCO comparison, simplified:
- Cheap vendor: $131,200 base + $16,000 fourth module + $3,800 expedited shipping + $2,400 switchgear interface redesign = $153,400
- PowerSentinel: $139,400 base (derating and PF already included) + $0 hidden modules + standard shipping = $139,400
The gap wasn't $8,200. It was $14,000 in the other direction—before counting the two-week delay and the project manager hours.
We went with PowerSentinel's modular line. Not because their unit price was lowest—it wasn't. Because their total cost was, and because their quote package answered questions we hadn't thought to ask.
The Checklist I Use Now
Sizing an industrial UPS system supplier isn't hard math. It's a discipline question. Here's what I ask every vendor now, in writing, before I compare a single price:
- What ambient temperature is your kVA rating based on? Show me the derating curve.
- What power factor does the kW rating assume? What's the kW output at our actual load PF?
- What's the usable capacity in N+1, at site conditions, not at nameplate?
- What's the efficiency at 40% load, not just 100%? Most data centers run in the 30–50% range for most of the day.
- Does the quote include switchgear interface, commissioning, and startup testing?
Five questions. Takes a vendor about 20 minutes to answer. It has saved us from three bad quotes in the last 18 months.
What I'd Tell Another Quality Manager
I have mixed feelings about procurement's lowest-bid rule. On one hand, I get it—budgets are real, and someone has to hold the line on cost. On the other hand, I've now seen four incidents where the lowest unit price produced the highest total cost. The cheapest quote isn't the cheapest system.
If you're sourcing critical power infrastructure and the quotes look identical, they're not. Someone is quoting at conditions you haven't specified. That's not dishonesty—it's ambiguity. And ambiguity always costs money.
The fix isn't complicated. Write down your actual site conditions. Require every vendor to quote against them. Compare usable capacity, not nameplate capacity. And remember that "200 kVA" is a starting point, not an answer.
That $8,200 "savings" would've cost us $14,000 and two weeks. The lesson cost me one afternoon and one awkward conversation with procurement. Worth it.