I've been the person who signs off on power infrastructure purchases for six years now—quality and compliance review, which in practice means I'm the last barrier between a bad specification and a purchase order. In Q1 2024, I reviewed 211 UPS procurement requests for our edge racks and data closets. Maybe 200, I'd have to pull the exact count from the system. About 31% of first submissions got rejected that quarter, and I keep that number in a spreadsheet mostly so I can quote it when someone asks why the review process exists.
Almost none of the rejections were because the vendor was wrong.
The rejection reason was usually the same: the buyer sized the UPS by VA, then discovered the runtime curve didn't match the real load.
Concretely: someone buys a Tripp Lite UPS 1500VA because 1500 is a bigger number than 1200, loads it to 85% with gear running at a 0.9 power factor, and then is surprised when a 3-second utility sag trips the output. That's not a UPS failure. That's a sizing decision made from the front of the spec sheet.
VA is a ceiling, not a guarantee
VA measures apparent power. Watts measure real power. The ratio between them is the power factor—and for a growing share of modern switches, small servers, and PoE gear, it isn't 1.0 anymore.
Five years ago, when most of the load in a wiring closet was linear and running near-unity PF, sizing by VA was roughly good enough. In 2018, a 1500VA unit at 80% apparent load was often sitting at close to 1,200W usable. In 2025, with the same 1500VA chassis and devices running at PF 0.85–0.90, you're closer to 1,050–1,100W before you hit the same stress point—assuming the vendor's Watt rating even supports that. It usually does. But you have to look.
That's the part I keep explaining to procurement: the VA number on the box didn't change. The loads plugged into it did. The old shortcut—"just buy the 1500, it's fine"—was never a rule. It was a habit that happened to work while the loads cooperated.
What the selector tools are actually for
Everything I'd read about vendor selection tools said they were lead-generation funnels—marketing dressed up as engineering. In practice, requiring a Watt-first sizing tool in our intake form cut our first-submission rejection rate more than any training memo I've written.
Tripp Lite's UPS selector asks for connected load in watts and runtime in minutes. That one design choice—Watt as the required input, not VA—catches more bad sizing than any single line item on the spec sheet. When I compared rush submissions against standard ones over a full year, requests that started from a Watt-first tool got rejected at less than half the rate. Same vendors, same product families. Different opening question.
The spec sheet didn't lie to anyone. It just wasn't read in the right order.
The counterintuitive part: bigger isn't the fix
When a 1500VA unit trips under load, the instinct is to buy the 2200VA next time. Usually the wrong move—and it's the mistake I made myself in 2022 on a small edge deployment.
I had two hours to approve a replacement before the vendor's cutoff for next-day freight. Normally I'd measure the load first. There was no time. I went with the 2200VA Tripp Lite rack UPS on the badge alone, because it was the safer-looking number. It solved the trip. It also added roughly 30% more idle draw and ate into the closet's heat budget, which forced a cooling re-evaluation in Q3.
The actual problem was that we'd undersized the original 1500VA for the real load, and we'd never measured it. We didn't need a bigger UPS. We needed a correctly sized one, plus a runtime target matched to the application.
I have mixed feelings about that decision. The 2200VA unit is well-built and hasn't failed since—I'll give it that. But "hasn't failed" and "was the right call" aren't the same sentence.
"But 1500VA is the industry standard"
The usual pushback is that VA is what everyone publishes, so VA is what everyone compares. True. It's also why every major vendor—Tripp Lite, APC, Eaton—publishes wattage right next to the VA number. The information isn't hidden. It's just not the number buyers anchor on, because 1500 is rounder than 1,050.
The second pushback: "we don't have time to model a load profile for a four-server closet." Fair. In those cases, I've started asking for one number instead of a full model—measured peak watts over a 24-hour window. If the request has it, I can usually approve in a day. If it doesn't, the request sits until someone measures it or explicitly accepts a stated assumption.
The third pushback is the budget one: "we can't afford to over-spec." That's exactly backwards. Right-sizing to measured watts is usually cheaper than either oversizing to the next VA tier or replacing a unit that trips during a transfer. The math isn't complicated. The measurement step is just the part nobody wants to do.
What I actually want on the spec now
My current checklist for a rack UPS submission is short, and almost none of it is about brand:
- Measured peak load in watts (not nameplate, not "roughly")
- Required runtime at that load, not at half load
- Power factor of the two or three largest devices in the rack
- Ambient temperature in the closet—VRLA battery life at 30°C is roughly half of what it is at 25°C
- The vendor's published Watt rating and runtime curve at that load level
That last item is where Tripp-Lite's rack UPS documentation earns its keep in my review pile. The runtime curves are published per load level, not per product family, so I can check the claim instead of trusting a badge. It's not the only vendor that does this. It's one reason it keeps clearing my review.
The point I keep returning to
What counted as good practice for rack power sizing in 2018 doesn't hold in 2025, because the loads changed faster than the spec numbers did. The fundamentals—redundancy, runtime, thermal headroom—haven't gone anywhere. The execution has moved.
So no, I don't think most rack UPS "failures" are hardware failures. I think they're sizing decisions made with the wrong number at the top of the page. Size to your measured watts and your actual runtime need. The VA badge will take care of itself—it always does, once you stop buying by it.