The 7-Step Office Power Reliability Checklist I Wish I'd Had in 2020

Who this checklist is for

This is for the person who inherits "the power situation" at the office—usually an office administrator, sometimes IT, sometimes both if the company is small enough. You probably have at least one UPS in a closet somewhere. You probably can't tell me its rated capacity without walking over to look.

I manage procurement for a 220-person company across two offices, roughly $180,000 a year of facility and IT spend. I'm not an electrician. But I've been the one answering the "why is the server down" text at 7 a.m. enough times to have a checklist I actually trust now.

Here's the seven-step version. It runs about 45 minutes per site if nothing is broken. Longer if it is.

Step 1: Confirm what's actually dead before you assume anything

Nine times out of ten, when someone tells me "the UPS failed," it's actually one of three things: the outlet lost power, a breaker tripped without visibly tripping, or the UPS did its job and nobody's sure what happened after. You can't diagnose any of that by staring at the LEDs.

Get a basic multimeter—a $30 unit is fine, you don't need a Fluke unless your electrician says otherwise. Set it to AC voltage, range 200V or higher (or just set it to auto-range if yours supports it). Then:

• Test the wall outlet feeding the UPS: hot to neutral, hot to ground, neutral to ground
• Test at the PDU if there's a PDU between the wall and the UPS
• If the UPS has accessible outlets, test the output with nothing plugged in

Normal readings in a U.S. office sit around 118–122V line to neutral. Line to ground should read the same. Neutral to ground should be near zero—a few millivolts is fine, anything above a couple of volts is worth a call to your electrician.

The part I missed the first three times I did this: a tripped breaker doesn't always look tripped. Some Square D and Cutler-Hammer breakers sit in the middle position when they trip, not the "off" position. You have to cycle them fully off and back on. If the multimeter reads dead at the outlet but the panel shows the breaker "on," flip it off and back on before you buy a new UPS.

Step 2: List what can't go dark—in watts, not "units"

Everyone writes this list in headcount. "Two servers and the switch." That's useless when you're sizing a UPS because a 2U server might pull 400W or 1,200W depending on what's inside it.

Walk to each device and read the label on the power supply. The wattage printed on a server PSU is the maximum it can deliver, not what it pulls under normal load. A reasonable rule of thumb: assume a server draws 50–60% of its PSU rating in steady state. Add the switch, the firewall, the NAS, the KVM, the phone system head unit—anything that loses business data or door access when the lights go out.

For us the list looks like this: two core switches (180W combined), one firewall (60W), the door access controller (spec sheet says 40W but I've never fully trusted that number—call it 50W to be safe), the SAN (900W rated, we assume 550W draw), and the two server cabinets (1,400W combined at max, roughly 850W typical).

Total it, then add 20–30% for headroom. That's your minimum VA target. Keep the wattage number too—you'll need it in the next step.

Step 3: Right-size the UPS—but not against the number you just calculated

This is where I'd argue most offices get it wrong. They calculate 1,500W, go buy a "1,500VA UPS," and end up running it near its maximum output. VA and W aren't the same number. VA is apparent power, W is real power, and the ratio between them is the power factor. On modern Tripp Lite units that's typically 0.9. On cheaper units it's often 0.8.

So a 1,500VA unit with a 0.9 power factor gives you 1,350W of real capacity. That 1,350 is the number to compare against your load list, not 1,500. If you sized your load at 1,500W and bought a 1,500VA unit, you're over the rated load and the unit is going to trip.

Two other things I'd tell anyone doing this for the first time. Run the UPS at 80% of rated capacity in steady state, not 100%—it ages batteries faster and the unit runs hotter at full load. And splitting the load across two smaller UPSes is often cheaper than one big one, plus it isolates failures. When one of our two cabinets was pulled for maintenance last year, the network didn't blink.

When our server-closet UPS needed replacement in 2023, I had approval up to $2,500 and a realistic replacement need closer to $3,500. I ended up buying the 3000VA online unit instead of the 5000VA, cutting the runtime spec from about 18 minutes to about 11. That was my call. The 11 minutes covers a generator startup with room to spare, so it turned out fine—but it was a decision I had to be willing to defend if it hadn't.

Step 4: Put servers behind an online double-conversion UPS, not a line-interactive one

Online double-conversion is the design where the load always runs off the inverter—power goes AC to DC and back to AC. The wall is only there to charge the battery. The load never sees mains power directly, so there's no transfer time when the utility drops out.

Line-interactive UPSes (Tripp Lite's SmartPro line, for instance) are fine for workstations, network gear, and routers. They switch over in a few milliseconds and do the job. But that brief transfer window, where the load runs off battery at slightly different waveform characteristics, is harder on some switch-mode power supplies than others. Two years ago we had a blade chassis lose both PSUs within a month of a line-interactive unit taking over during a brownout.

Honestly, I'm not sure whether the PSU failures were causally the UPS. My best guess is the brownout itself was the main problem. But we moved the servers onto a Tripp Lite SmartOnline unit and haven't had the issue since, and the logic of online double-conversion for anything that can't lose a power supply is solid. The cost difference versus line-interactive has narrowed enough that I wouldn't buy anything else for a server rack.

Step 5: Consider lithium UPS where the battery is a pain to reach

Battery lifecycle is the single biggest hidden cost of a UPS. Standard VRLA batteries last 3–5 years in a climate-controlled room, less in a hot server closet. Then you're buying replacements, paying someone to install them, and occasionally discovering a swollen battery wedged into the tray.

Tripp Lite's lithium-battery UPS units trade off in a few specific ways. On the plus side: roughly 8–10 years of service life instead of 3–5, better heat tolerance, and significantly lighter—which matters if you're the one lifting a 3U unit into a rack. On the downside: the upfront cost is typically 30–50% higher, and the runtime at a given VA isn't automatically longer. Lithium energy density is better, but the watt-hours packed into a given chassis don't always change as much as the marketing implies.

My take: use lithium anywhere the battery is hard to service. The AV cabinet UPS behind the conference room console that requires moving furniture to reach? Lithium. The unit in a branch office closet that needs a drive to swap? Lithium. The UPS in the well-visited server room you're already in every week? VRLA is probably fine and you'll save money.

Step 6: Know what the building does past UPS runtime

Your UPS gives you 10–30 minutes, usually. That rides out a brownout or a clean shutdown. It isn't enough for a four-hour outage, which happens more often than people expect at commercial sites outside major metro grids.

This is where you talk to the building owner or facilities manager. The question is simple: does the building have backup generator capability, and is your suite wired to it? Every site I've managed since 2020 has needed this answered, and the answer is usually "partially."

If you don't have generator coverage and can't get it, the UPS runtime is your hard limit. Size for it directly—external battery packs or a bigger chassis. There's no way around the physics of it.

If the building does have a generator, the equation changes. Now the UPS is just covering the 30–60 seconds between grid loss and the generator spinning up. That's a much smaller job. But you need to know that gap. Most commercial auto-start generators stagger-start within 10–30 seconds. If they're on a scheduled test cycle, or if the first start attempt fails and the second one takes longer, you can see 60 seconds easily. Your UPS has to cover the longest realistic gap, not the shortest.

We had one site that was a converted house—1,200 square feet, residential until 2019. The landlord had installed what homeowners call a whole home backup generator, and the commercial tenant lease treated it as building infrastructure. Two things I learned: "whole home" means nothing to a commercial tenant until you confirm your circuits are actually wired to the generator panel, and you want to know exactly where the transfer switch is before an outage, not during one. Ours was in a utility closet behind the water heater. Nobody had thought to tell the tenant that.

Step 7: Don't forget the small loads that still matter

This is the step most published checklists skip, and it's the one that's cost me real annoyance twice.

When power comes back and you're walking the site doing damage assessment, you need to actually see what you're looking at. That means light and a charged camera. If you're like us, the site documentation kit lives in a drawer and includes a decent LED flashlight and an older Nikon D90 with its battery charger—because our insurance process wants genuine photos, not phone snaps.

Last March we had a regional outage on a Sunday morning. The D90's battery was at about 15% because the last person who used it took it home and didn't bring the charger back. We ended up photographing water damage on a phone at 11 p.m. with the flash on, and our claims adjuster flagged two of the images as too blurry to use. Not a catastrophe. Just avoidable.

So: put a small UPS or a dedicated circuit under the charging shelf. We dropped a 350VA unit under the shelf where the chargers and flashlight dock live. It'll run them for hours, and it means that if an outage hits overnight, the documentation tools are ready by morning.

The other categories in this bucket, depending on your site: the phone system head unit if it's still physical hardware, the front door badge reader, and any line-of-business terminal at reception. None of them need a big UPS, but all of them need something.

Common mistakes to spot on the way out

• Sizing to VA instead of W. Read the watts line on the UPS spec sheet, not the VA line, when comparing against your load list. It's the most common error I see in inherited setups.
• Never testing under load. A UPS that lights up green and answers a ping is not the same as a UPS that can run your rack. Pull the input breaker, watch the actual load, and note the runtime. If it's half the spec sheet, the batteries are halfway gone.
• Assuming batteries are fine because they're 3 years old. Three years is when VRLA starts to fall off. Test at year 2 and expect to replace at year 4.
• Not knowing where the transfer switch is. If the building has a generator, the transfer switch is the piece that makes it work. Find it before the outage, not during.
• Skipping the small loads. The charger shelf, the badge reader, the phone head unit. No big UPS required, but they still need to be on something.

That's the list. Forty-five minutes per site if you're lucky, a half-day if you're not. The first pass is the worst one; after that you're mostly verifying that nothing changed. Which, honestly, is most of what facility work is once the setup is right.