The UPS Wasn't the Problem: What Actually Fails When the Grid Drops
When a facilities team starts shopping for backup power, the conversation is almost always about the box. How many kVA, Galaxy versus Easy UPS 3S, efficiency at partial load, price per VA. Someone builds a spreadsheet. Someone else asks whether the network card has a default password, and it goes on a commissioning checklist that may or may not get followed.
I've spent about twelve years in critical power — commissioning, field service, and the 2 AM calls that come with both. I've handled calls ranging from a single edge closet with a 3 kVA unit to multi-hundred-kVA rooms, and the failure pattern doesn't change much. In my experience, the UPS itself is rarely what lets people down. The hardware is usually fine. What fails is the layer around it: the configuration, the transfer path, the batteries, and the assumptions about how long any of it will actually run.
What People Think They're Buying
Most of the buying energy goes into the UPS. That's understandable — it's the expensive part, the part with a spec sheet, the part procurement understands how to compare. Schneider UPS lines cover a huge range now, from small single-phase units up through the Galaxy data center systems. The Schneider Electric Easy UPS 3S is pitched at the commercial and edge market: three-phase, modular, designed to be deployed quickly. 'Quickly deployed' is usually a good thing.
It's also how you end up with a production UPS that still has the factory login three years later.
Here's the thing: the spec sheet tells you what the box can do under test conditions. It doesn't tell you whether the network card was configured, whether the transfer switch was ever switched under load, or whether the battery string is still delivering the capacity it had at commissioning. Those are all setup decisions, and they don't show up in the spreadsheet.
What Actually Fails
It's basically a configuration problem wearing a hardware costume. Three layers tend to fail, usually in combination.
The network card that's really just an unpatched server
Modern UPSs ship with network management cards. They're useful — remote monitoring, alerts, graceful shutdown. They're also small computers with web servers, SNMP, and, out of the box, default credentials. Schneider's own documentation for the Network Management Card lists those defaults and instructs you to replace them during commissioning. The instruction exists because the step gets skipped.
If you search 'Schneider Electric default password UPS,' you'll find forum threads going back years. The problem isn't unique to Schneider — every major vendor has the same issue — but it's the one people ask about, which tells you something about how often it surfaces.
NIST SP 800-63B discourages default and shared credentials for exactly this reason. A Shodan search for common UPS web interfaces still returns thousands of results, so the guidance isn't theoretical. Most of those are not sophisticated exposures; they're cards on a management VLAN with a port forward someone added over a weekend because they wanted remote access without a VPN.
I'll be honest — the first time I found a production UPS with the factory login intact, I assumed it was a one-off. It wasn't. We didn't have a commissioning checklist for network cards, and that's how a production unit kept its factory login for three years. That's when I started requiring a timestamped photo of the settings screen before any UPS went live.
The transfer path that exists on paper
This is where residential and small commercial setups get interesting. A house power transfer switch is a solid solution — if there's a trained person standing next to it when the grid drops. In practice, the plan lives in someone's head. The generator is in the garage. The cord is somewhere. Nobody has actually run the sequence since the install.
Commercially, the same issue wears a different hat. An automatic transfer switch gets installed, tested at commissioning, and then ignored until the annual fire inspection. If it hasn't been exercised under load in eighteen months, it's a hypothesis. Mechanical parts stick. Control relays fail. The generator starts, the ATS doesn't transfer, and the UPS carries the load until the battery runs out.
NEC Article 702 governs optional standby systems, and 702.11 requires transfer equipment that prevents inadvertent interconnection of normal and alternate sources. That requirement exists partly because people still use double-male 'suicide cords' to backfeed a panel. They work right up until they energize a line a utility worker thinks is dead.
I assumed a site's ATS had been tested because it had a maintenance sticker on it. It had — in 2021. When we finally exercised it in 2024, it transferred fine. We were lucky, not prepared, and I don't like depending on luck.
The battery that passes the voltage test
VRLA batteries are the quiet failure mode. IEEE 1188 is the standard maintenance reference, and the commonly cited rule of thumb in battery maintenance literature is that every 10°C above 25°C roughly halves service life. A battery room sitting at 32°C is quietly eating its own design life.
The trap is that voltage checks feel like maintenance. A four-year-old string can read 13.2V at rest and still deliver less than half its rated capacity under load. Voltage tells you the battery isn't open-circuited. It doesn't tell you how long it will run the rack.
Maintenance chargers — the cen-tech battery charger you pick up to keep spare strings topped up, for example — have a role. They keep idle batteries healthy. They are not a capacity test. I've seen sites where a trickle charger was the entire battery program. It isn't one.
Then there's load growth. Nameplate runtime is calculated at a specific load and power factor. Loads grow. Nobody recalculates. A UPS sized for fifteen minutes at commissioning is delivering eight minutes three years later before you factor in battery aging. The UPS is doing exactly what it was told. The plan was wrong.
The Bill You Don't See Until the Lights Go Out
In February 2024, a client called because their UPS had 'failed' during a ninety-minute utility outage. It hadn't. The battery string was four years old and had never been capacity-tested. The load had roughly doubled since commissioning. Actual runtime was eleven minutes. The outage was ninety. They spent the next six hours restoring a storage array that should have ridden through on a generator that — it turned out — had never been wired to the UPS panel.
So the hardware cost of that night was a battery string and a service call. The real cost was the six hours, the recovery work, and the fact that the client had believed they were covered for three years when they weren't.
This is the part people miss about emergency power: the premium you pay in an emergency isn't really for speed. It's for certainty. A guaranteed four-hour battery replacement at 11 PM is rational money. Paying that premium because you skipped the annual capacity test is just expensive. Cheap and uncertain is more expensive than expensive and certain — always, once you count the recovery.
I've watched companies pay rush fees, overtime, and expedited freight to fix things that a commissioning checklist would have caught. The rush fee isn't the problem. The missing checklist is.
What the Setup Should Look Like
Look, none of this is exotic. It's just work that doesn't happen because it isn't urgent until it is.
- Commission the network card like a server. Replace default credentials, disable unused services, put it on a management segment with no direct internet exposure, and document the firmware version. For an Easy UPS 3S, that's a twenty-minute job during install and a very long night if you skip it.
- Exercise the transfer path under load. Quarterly for an ATS, annually at minimum with a load bank for the generator. For a manual house power transfer switch, write the procedure on a card, tape it to the panel, and actually run it once with someone who isn't you.
- Replace voltage checks with capacity tests. Annual discharge test per IEEE 1188 guidance. Log install dates and battery room temperature. Use a maintenance charger for spares if you like, but don't confuse it with a test.
- Recalculate runtime against current load every year. Use measured load, not nameplate. Include the aging factor. Write the number on the UPS.
- Plan the generator connection before you buy the generator. Where to buy portable generator is the easy part — big-box stores, equipment dealers, online. The hard part is the inlet box, the compliant transfer switch, fuel storage, and the fact that someone has to be there to start it. Buy for the connection you can actually support.
My experience is mostly mid-size commercial and edge deployments, 10 to 100 kVA. Very large data centers have their own commissioning culture, and I can't speak to that with the same confidence. If you're running hyperscale infrastructure, your baseline is different from what I see. But for the smaller facilities — the ones where one person wears the facilities hat alongside three other jobs — the pattern is remarkably consistent.
The UPS is the easy decision. The setup is the one that decides whether 2 AM is a minor event or a very long week. Buy the certainty, not just the box.
