I spent two hours on a video call last week with a facilities manager who was convinced his Eaton UPS was defective. He had the event logs, the alarm history, even a photo of the LCD screen showing "Load Off." Every piece of evidence pointed to one conclusion, he insisted: the UPS had failed mid-transfer and killed his rack.
I asked the obvious question. "What happened to the load at the exact moment of transfer?"
Silence.
Nobody had checked the transfer switch.
This is the single most common blind spot I see in my work. I'm a quality and compliance manager at a power protection company. I review every UPS system and transfer switch that goes out our door—roughly 200+ units a year. I've rejected about 8% of first deliveries in 2024 due to contact resistance issues, loose terminations, or documentation gaps. After four years of doing this, I've developed a reflex: when someone tells me their UPS dropped the load, my first question is about the transfer path. Not because I doubt the UPS—though I've seen genuine failures too—but because the transfer path is where power protection plans go to die.
The Eaton 9390 UPS is a solid three-phase unit. I don't say that as a marketing person; I say it as someone who has watched these units handle real grid disturbances in production environments. But no UPS, regardless of brand, can compensate for a transfer switch that fails to close.
Here's what I see in site audits over and over: facilities spec their UPS meticulously. They compare runtimes, maintenance contracts, efficiency curves. They follow the Eaton UPS news today—lithium-ion upgrades, connectivity features, the latest firmware. And then they pick a transfer switch based on "it was in stock at the distributor."
A DC transfer switch deserves far more scrutiny than it gets. What most people don't realize is that DC switching is fundamentally different from AC switching. On AC, current crosses zero 50 or 60 times per second, which naturally extinguishes the arc when contacts open. On DC, there's no zero crossing. The arc persists. So a contact design that works fine for AC can erode quickly under DC load. That's not a niche concern—DC distribution is showing up in more facilities every year, especially in telecom and edge computing.
Here's something vendors won't tell you: contact resistance specifications are not always published. You have to ask. And when you ask, the answer is often a range so wide it's practically useless.
I rejected a batch of transfer switches last year because the contact resistance measured 30% above the documented spec on three of fifty samples. The vendor argued it was "within industry tolerance." Maybe. But industry tolerance doesn't keep a critical load up at 2 AM when a distribution transformer fails.
So how do you catch a degrading transfer switch before it fails? You measure.
A few months ago, a client with a 10-circuit 30 amp manual transfer switch kit installed in their server room told me they'd "never had a problem" with it. I asked when they last measured contact resistance across the poles. The answer was a blank stare.
You don't need specialized equipment for this. A standard multimeter is enough. The procedure I train our field engineers on looks like this:
Here's how to check resistance with a multimeter in a way that actually tells you something: test quarterly, write every reading down, and look at the trend. A single high reading might be a dirty contact. Two or three high readings across different poles suggest a pattern—and patterns are how failures start.
What you're watching for is creep. If the reading climbs a few milliohms each quarter, that switch is telling you something. Heat from resistance accelerates oxidation, which increases resistance, which generates more heat. Left alone, that feedback loop ends at the worst possible moment.
I assumed "same specifications" meant identical results across vendors early in my career. It doesn't. Each manufacturer interprets contact resistance specs differently, and some don't test at all before shipping. That's a lesson I learned after a $3,000 order came back with readings that didn't match the data sheet—and I had to catch it before it reached a customer.
In 2022, I supervised a quality audit that uncovered a $22,000 problem. A client's Eaton 9390 UPS performed flawlessly during a grid disturbance—the event log confirmed it. The load still dropped. Why? The automatic transfer switch upstream had been installed with underspecified contacts, and the resistance had degraded to the point where the switch couldn't hold under load.
The client's first reaction was to blame the UPS. Their second reaction was to spec a bigger UPS. It took two weeks of conversation to convince them that a larger UPS would have changed exactly nothing. The UPS was never the problem. The path was.
That's the expensive part of this issue: misdiagnosis. You replace or upgrade the wrong component, and the failure recurs at the next outage. I've seen facilities spend six figures on UPS capacity they didn't need while the real culprit was a $400 switch that had never been tested.
Uptime Institute's 2024 outage analysis found that more than half of data center outages cost over $100,000 in direct costs alone—before counting lost revenue and reputational damage (Source: Uptime Institute, 2024). Even a modest 10-minute outage can run into five figures for a mid-size operation. And in my experience, the costliest outages are the ones where the protection system itself was the weak link, because nobody expected it.
I'm not going to tell you to throw away your existing gear. I'm also not going to pretend that one brand solves every problem. One of the most useful things a vendor can say is "this isn't our strength—here's what you actually need." The vendors who say that earn my trust for everything else. The ones who claim to be perfect at everything? I've learned to read their spec sheets a little more carefully.
So here's what I'd do before your next outage, in order of priority:
1. Measure your existing transfer switch. If you own a multimeter, you own the tool. Check contact resistance on every pole. Write down the readings. Test again in three months. If you can't find the manufacturer's spec, that itself is a red flag—call the vendor and ask. If they won't give you a number, replace the switch with one from someone who will.
2. Spec the transfer path with the same rigor as the UPS. When you're evaluating a UPS like the Eaton 9390 series, ask the same questions about the transfer switch: rated current, contact resistance, number of mechanical operations expected, DC rating if the load is DC. A 10-circuit 30 amp manual transfer switch kit is a reasonable starting point for smaller rack loads—but only if it's documented and tested, not just installed and forgotten.
3. Consider whether automatic is actually better for your situation. Automatic transfer switches add convenience, but they also add control logic and motorized mechanisms. A manual transfer switch is simpler, and simple things fail less. I've worked with mission-critical facilities that deliberately chose manual switches because there's no control board to misconfigure. You have to have a person present to operate it—that's the tradeoff. For a facility with staff on-site, it's often the right one.
4. Know what your UPS can't do. This is the boundary that matters. A UPS conditions power, bridges gaps, and communicates status. It does not guarantee that the circuit upstream will conduct when it needs to. The moment you understand what a UPS can't do is the moment you stop misdiagnosing failures—and stop spending money on the wrong fix.
The honest truth is this: most UPS "failures" I investigate are not UPS failures at all. They're transfer path failures that were visible months or years in advance—if someone had bothered to look.
The good news is that looking is cheap. A decent multimeter costs less than one hour of avoided downtime. A quarterly test takes ten minutes. The bad news is that most facilities never do it, which means the next outage is already in motion.
Your call.