Late last March, the power in our building flickered twice, then went dark. The server room UPS clicked over, beeped for maybe twenty seconds, and shut down anyway. That moment cost us about $4,200. It also changed how I think about APC UPS units.
I manage procurement for a 40-person logistics company. I have managed an IT procurement budget of about $180,000 a year for six years, and I document every invoice in our cost tracking system. This was not a rookie mistake. I just ignored the UPSs because they were quiet, which meant nothing was happening.
The outage that started it
At first, the outage looked tiny. Lights flickered in the dispatch office, then the network dropped. Our file server went down hard. The database used by our order system needed a manual recovery process, and the contractor who handles our server work had to rebuild the file structure.
The cost added up fast: contractor time, overtime for dispatchers, and two expedited shipments because our phone queue and order portal were down during our busiest window. Total: $4,200. The server room APC UPS was not dead. It was old. The battery was over four years old, and the self-test feature had never been used. We had the capability to check it. We just did not.
I only believed in UPS battery testing after ignoring it and eating that invoice. The $4,200 outage was a tuition payment.
Batteries, new units, and the TCO question
After the outage, I got approval to fix every UPS in the building. We had twelve units. Seven were APC Back-UPS units on desks, one was a Smart-UPS 1500 in the server room, and four were older units collecting dust in a storage closet.
First, I compared replacement batteries against new units. For the older Back-UPS models, the cost of a genuine APC replacement battery was close enough to the price of a new unit that new units made more sense. New unit, new warranty, no old electronics inside.
Then I found third-party replacement batteries at roughly half the price of APC batteries. I almost ordered them. That is my job: find the lowest cost that gets the job done. But when I put the total cost into our procurement spreadsheet, the cheap battery lost. A third-party battery would save maybe $40 per unit. If it failed earlier, I would pay the same labor to swap it again. And if any warranty question came up later, that $40 saving was not worth the phone call. I ordered genuine APC replacement batteries, and for the older units I ordered new APC UPSs.
Why do genuine parts matter? Because the UPS is only valuable when it works on the day you need it. That is the entire product.
What I standardized on
Here is what the final setup looked like.
Workstations and call center desks
For people who handle live orders, I standardized on the APC BN650M1 Back-UPS 650 VA battery backup system. The manufacturer spec sheet lists 650VA and 390W, which is enough for a desktop, a monitor, and a VOIP phone. It also has enough runtime for somebody to finish the current screen, save what they are doing, and shut down cleanly.
If you are comparing a backup UPS 700 APC unit against the 650, the 700 gives you extra room for docking stations and larger monitors. For most standard desks, the 650 worked fine. For power users, the 700 was safer. It is not about being the best. It is about matching the load to the device.
Server room
For the server room, I went with an APC Smart-UPS, not a Back-UPS. The network switch and firewall have active PFC power supplies. Back-UPS output is simulated sine wave. Smart-UPS output is pure sine wave. Some equipment with active PFC does not like simulated sine wave, and it can shut down even while the UPS is running. That is the difference, according to APC by Schneider Electric, between the two lines: Back-UPS is designed for home and small business gear, Smart-UPS is for servers and network infrastructure.
The Smart-UPS cost more. It also protects the part of the office where a failure causes the most pain. That is not where I cut corners.
Surge protectors and whole-house protection
Why mention a whole-house surge protector? Because a lot of people search for how a whole house surge protector works when they start comparing UPSs. Here is the simple version: a whole-house surge protector installs at the main panel and clamps high-voltage spikes before they travel to branch circuits. It does nothing during a blackout. A UPS with a battery covers that gap. You can need both, but they solve different problems.
The maintenance lesson
Replacing the units was the easy part. Keeping them useful is the part I had skipped for years.
Our maintenance plan now has five items:
- Test every UPS using its self-test feature once a month, or run a controlled load test if the unit supports it.
- Replace batteries on a schedule, not when the beeping starts. For the small units, that means every three years in our climate.
- Write the battery install date on a label on the unit and in the same spreadsheet where we track purchases.
- Set a calendar reminder for battery replacement and warranty expiration.
- Check that the UPS is actually plugged in and is not being used as a power strip.
That last one sounds stupid. During the after-action review, we found one APC UPS in the dispatch area that was completely unplugged from the wall. Not damaged. Not failed. Unplugged. Nobody noticed because it sits under a desk.
Oh, and every unit now has a laminated tag that says: Test monthly, replace battery on schedule, no power strips. It sounds small, but it fixed our culture.
Five minutes of verification beats five days of correction. The checklist is the cheapest insurance I have.
What the numbers look like
The whole replacement project came in around $3,800, including new units, genuine replacement batteries, and the hour of labor to install everything. That is less than the $4,200 outage. For the first time, we have a spreadsheet that shows the status of every UPS in the building.
Since then, we had a 40-minute power bump in July 2024. Every unit did what it was supposed to do. The server room Smart-UPS carried the network for 27 minutes and then shut down cleanly. No lost sessions, no database recovery process, no emergency contractor invoice.
There is something satisfying about a green Pass next to every unit in the spreadsheet. After the March outage, I did not expect to get there that quickly.
To be fair, this is a small office setup, not a data center. If you are running a server farm, you need a different level of redundancy. But for a small business with a few critical workstations and one network closet, the answer is simple: buy decent equipment, test it, and replace the batteries before they fail. It costs less than the outage.