The call at 7:15 a.m.
In November 2024, a 32-chair dental clinic called me at 7:15 on a Tuesday morning. Power had dropped at 5:40. By the time the office manager reached me, the front desk phones were working, the network was up, and the treatment rooms were dark.
She wanted to know why her UPS had "failed."
I coordinate emergency power service for an electrical equipment supplier. I've handled 300+ rush service and replacement orders in six years, including same-day turnarounds for clinics, data rooms, and retail chains. So I've heard this question a lot.
Her UPS hadn't failed. The APC Smart-UPS in the server closet was doing exactly what a UPS does: covering a short outage while the load either rides through or shuts down cleanly. It had been doing that job for 18 minutes. Then it did what it was designed to do at minute 19—it stopped.
What failed was the assumption behind the purchase. Nobody had ever asked how long the building could be dark.
What people think they're buying
When someone types "apc ups 1350va" or "apc back-ups es 550" into a search bar, they're usually trying to solve a specific, visible problem: the desktop shuts down mid-edit, the POS terminal reboots during a flicker, the modem dies during a storm. That's a real problem, and the APC-UPS lineup solves it. A Back-UPS ES 550 is a fine answer for a home office with a router and a laptop. The 1350VA units cover a small workstation or a retail counter.
Where it goes wrong is the next step. The same person, having solved the flicker, assumes they've solved the outage. They haven't. They've bought a bridge. How long that bridge is depends on the load, the battery, and—this is the part nobody reads—the age of the battery.
I have mixed feelings about how UPS marketing handles this. On one hand, the run-time charts are honest if you actually read them. On the other, "up to 45 minutes" at 25% load is not the number anyone remembers when the lights go out.
The load nobody counted
Here's the deeper problem, and it's the one that actually causes the 7:15 a.m. calls.
Most UPS failures I deal with aren't hardware failures. They're load-list failures.
Someone specs the UPS around the server. Then the install electrician, doing what electricians do, puts the network switch, the phone system, the door access controller, and the camera NVR on the same circuit—because they're all in the same closet and it's tidy. None of those devices is large on its own. Together they cut your run time by a third. Sometimes half.
The smaller the device, the easier it is to forget. A building automation panel draws maybe 15 watts, but it's always on, and if it's on the UPS circuit, it's on your clock. I've seen the same pattern in mobile and marine installs, where a Tecma toilet control panel or a similar always-on controller ends up on the backup circuit because it was the closest outlet. Fifteen watts sounds like nothing until you're doing the math at minute 17 of a 20-minute bridge.
The UPS doesn't fail when the battery dies. It fails the day the load list was written and never updated.
There's a second layer here that catches even careful people. Nameplate wattage and startup wattage are not the same thing. A device rated at 300W can pull 600W or more for a second or two at startup. If your UPS is sized against nameplate, you're sized against the wrong number. Motors, compressors, laser printers, and anything with a heating element are the usual suspects.
This is the oversimplification that trips up most DIY sizing: "total up the watts and add 20%." That rule ignores inrush current entirely. To be fair, most installers aren't trying to cut corners—they're working from a rule of thumb that was probably fine for desktop loads in 2010.
The generator question that comes next
When the clinic called back the following week, the conversation had moved on. They wanted a generator. Understandable. That's usually the moment someone starts searching "home generator installation near me."
Here's the thing about generators: they don't eliminate the gap. They shorten it, after a delay. A standby generator needs anywhere from 10 seconds to a minute to start and transfer, depending on the unit and the transfer switch. Some need longer in cold weather. Your UPS has to cover that window—and only that window—if the generator is sized and wired correctly.
The failure mode I see most often is a generator sized for the building's average load, not its peak. Then the transfer happens, the generator picks up the load, and the UPS sees a voltage sag during transfer that looks a lot like an outage. The UPS does its job. The generator trips on overload. Now you're back on battery, and the battery is the only thing standing between you and a dark building.
The same logic applies to hybrid inverters, which is what a lot of people are asking about right now. What is a hybrid inverter? Broadly, it's a system that manages solar input, battery storage, and grid or generator input in one unit. It's genuinely useful for extending run time and shaving peak demand. What it is not is a substitute for a UPS's transfer time. A hybrid inverter's switchover is measured in milliseconds to seconds; a UPS's is measured in milliseconds. If you have equipment that can't tolerate even a brief interruption, the inverter is not the last line of defense. The UPS is.
Granted, this is more complex than most people want it to be. That's the point. The complexity is why the failures happen.
What it actually costs
I won't quote a generic "downtime costs $5,600 per minute" figure, because those numbers come from surveys of large enterprises and they don't translate to a dental practice. What I can tell you is what I see in service tickets.
In the clinic's case, the outage was 3 hours and 40 minutes. The UPS covered 18 of those minutes. The remaining 3 hours and 22 minutes cost them a full day of scheduled procedures—call it $14,000 to $18,000 in deferred revenue, or maybe closer to $12,000 once you net out the patients who rebooked. I'd have to pull the ticket to be sure.
The UPS itself was fine. The battery was 14 months past its recommended service life, which nobody at the practice had been told to track.
The standards exist for a reason. NFPA 110, which governs emergency and standby power systems for life safety and critical facilities, requires monthly inspection and annual load testing. IEEE 446 lays out recommended practice for commercial and industrial standby power. UL 1778 covers the UPS itself. None of these directly apply to a small dental office—but the underlying logic does. A system nobody tests is a system nobody knows the limits of.
And the battery is the part that ages silently. APC's documentation for the Back-UPS and Smart-UPS lines generally puts battery service life in the three-to-five year range under normal conditions, shorter in hot rooms. Nobody schedules that. The UPS keeps humming, the green light stays on, and everyone assumes it's ready. It's ready for a shorter outage than last year.
The five-check list
This is the part I'd put on a single page and tape inside the server closet door. It takes about 40 minutes once a year.
- Write down every device on the UPS circuit. Not the ones you meant to put there—the ones that are actually plugged in. Walk the closet. Check the panel.
- Measure, don't estimate. A $30 plug-in meter gives you real draw at idle and under load. Nameplate is not measurement.
- Find the inrush devices. Anything with a motor, compressor, or heating element. Those need headroom, not just capacity.
- Date the batteries. Write the install month on the unit with a marker. Put a reminder in your calendar for month 36.
- Test under real load once a year. Pull the breaker, not the plug. See what actually stays up, and for how long.
Five minutes of verification beats five days of correction. I've written that list into service reports for 14 clients over the past two years. Nine of them had at least one device they'd forgotten was on the UPS.
There's something satisfying about a system that behaves predictably during an outage. Not impressive—predictable. After enough 7:15 a.m. calls, that's the only standard that matters.
