The old backup-power playbook is easy to summarize: buy a UPS, plug it in, and ignore it until the beeping starts. The current playbook is a different document entirely—right-sized units, scheduled battery decisions, PO-level spec control, and a plan for power that moves. I'm not going to pretend the old approach was stupid. In 2019 it was a reasonable use of a limited budget. But what was fine then is expensive now, and most of that expense hides in labor and downtime rather than the equipment invoice.
I run facilities procurement for a roughly 180-person company—around $310,000 in annual spend spread across 14 vendors, five of which touch electrical gear in some way. Here's the comparison I wish someone had handed me five years ago: five dimensions, old playbook vs. new, with a straight answer at the end of each one instead of a shrug.
Dimension 1: How You Size the UPS
The old playbook
One size fits all. Either you bought the biggest unit you could justify (because bigger felt safer) or you put a cheap Back-UPS on every desk that had a computer. There wasn't much in between, and nobody asked what the actual load was.
The current playbook
You size to the real load and the real runtime target. A small network rack with a switch, a modem, and two cameras is an APC Smart-UPS 1000XL conversation. A closet with three servers, a phone system, and a NAS is an APC Smart-UPS 3000VA conversation. The 3000VA is not automatically the "better" unit—it's the right unit for that load and that runtime, and it's the wrong unit for a 90-watt closet.
Where I got this wrong: I used to treat oversizing as free insurance. It isn't. An oversized unit runs at very low load, so you've paid for capacity you'll likely never use—and you'll pay for it again at battery replacement, because there are more cells to replace. I should add that the runtime charts on APC's own documentation will tell you this in about ten minutes. Nobody reads them.
Verdict: Right-sizing wins. The counterintuitive part is that the smaller, cheaper unit is frequently the more reliable choice—if it actually matches the load.
Dimension 2: What You Do When the Batteries Age Out
What we used to do
Wait for the beep. Run a self-test once a year if anyone remembered. When the runtime collapsed or the unit threw a fault code, call somebody and hope the replacement was in stock.
What we do now
Batteries go on a schedule, and when the date arrives we make a real decision rather than a reflex one: swap the batteries or replace the whole chassis. An APC Smart-UPS 1000XL battery replacement is a fairly straightforward job—the kit exists, the procedure is documented, and a competent tech can do it in under an hour. But it only makes sense if the rest of the unit has years left in it.
I went back and forth on this one for weeks the first time it came up. The battery kit was clearly the smaller line item; a full chassis replacement was the bigger check. The battery failure in March 2023 changed how I think about this, actually—one critical deadline missed because a six-year-old unit died mid-week and the replacement took four days to arrive. Suddenly the bigger invoice looked cheap. I chose full replacement that time, not because the batteries were dead, but because the chassis had already thrown two fault codes in eighteen months and I wasn't going to defend a third outage to my operations manager.
Verdict: Schedule the decision, then judge on chassis age and fault history—not on the battery invoice alone. The cheaper line item is often the more expensive path once you count labor, shipping, and downtime. (Note to self: put that sentence in the procurement wiki so I stop re-learning it.)
Dimension 3: How the Transfer Switching Gets Specified
The old way: the electrician shows up with whatever contactor is on the truck, and the admin signs the invoice. Nobody checks the coil voltage or the pole count against the drawing, because nobody has the drawing.
The updated way: the specification lives on the purchase order. If the transfer scheme calls for a 3 pole contactor 120V coil, that's what's printed on the PO, and receiving checks it against the packing slip before the electrician is scheduled. I don't wire it—that's for the licensed electrician, and it falls under NFPA 70 (the NEC) plus OSHA's lockout/tagout requirements at 29 CFR 1910.147. But I do own the spec, and I own the mismatch when the wrong part shows up.
Early on, I treated electrical components as "the electrician's problem." That worked right up until a two-pole contactor arrived for a three-phase transfer and the install stalled for a week. The part was maybe $40. The week was not.
Verdict: The admin's job isn't to design the electrical system; it's to make sure the PO matches the design the electrician signed off on. Small distinction, big difference in who gets blamed.
Dimension 4: How You Verify a Circuit Is Actually Dead
Old approach
Flip the breaker, poke it with a multimeter, call it dead. Or trust the label. Or trust the last guy who said he trusted the label.
What's actually required
You use a voltage tester the way it's meant to be used. Test the tester on a known live source first—every time. Then test the circuit. Then test the tester again on the way out. If you don't know how to use a voltage tester beyond "touch the probes to the wires and see if it lights up," you shouldn't be the person verifying a disconnect.
OSHA's electrical safety rules (29 CFR 1910.333) put verification of de-energization on the person doing the work, not the person who flipped the breaker. The tester's category rating matters too—a CAT III meter on a CAT IV service is a shortcut to a very bad afternoon. In my opinion this is the single most under-trained step in small-facility electrical work, and it's the one I now write into the vendor scope explicitly.
Verdict: Process beats tool. The meter in your drawer is probably fine; the habit of testing the tester is what actually keeps people alive.
Dimension 5: What You Do About Power That Moves
The old way: either haul a fixed UPS to a job site (and wonder why it doesn't appreciate the vibration) or accept that off-site work simply runs unprotected.
The updated way: mobile loads get mobile power. An RV inverter generator is a genuinely different product category from a UPS—it produces power rather than buffering it—and it's the right tool for field teams, remote trailers, and events where there's no grid to lean on. The mistake is treating it as a substitute for a UPS on sensitive electronics. The inverter output is clean, but it's still a generator, and it still needs proper grounding and a proper transfer arrangement.
Verdict: Match the tool to the mobility requirement. Fixed UPS for fixed loads; inverter generator for loads that travel. The two aren't interchangeable, and pretending they are is how you end up with a dead laptop in a parking lot.
Which Playbook Should You Actually Run?
Here's where I've landed after about five years of managing this:
- Keep the old playbook for low-criticality loads—a single desktop, a shared printer, a phone. A Back-UPS that runs until it dies is a perfectly rational choice when the downside is a reboot and a grumble.
- Run the current playbook for anything that carries revenue, safety, or a service-level commitment. Right-sized units, scheduled battery decisions, PO-level spec control, documented verification, and a separate plan for mobile power.
The part people underestimate is the middle. Most facilities don't have one critical closet—they have three or four semi-critical closets, and the old playbook assumes all of them can tolerate the same level of neglect. They can't. That's the whole argument for updating the playbook, and it's also the argument for not overcorrecting: the desk that only needs a reboot doesn't need a 3000VA unit and a service contract.
It took me roughly three years and about 40 battery replacement cycles to understand that the equipment is the easy part. The hard part is the decision framework—knowing when to replace batteries, when to replace the unit, when to call the electrician, and when to leave the whole thing alone. That framework is what the current playbook is actually buying you, and it's the piece you can't put on a purchase order.
What was best practice in 2019 wasn't wrong—it was optimized for a different risk tolerance and a cheaper cost of downtime. With tighter SLAs and much higher labor rates, the math has moved. The fundamentals haven't changed. The execution has.
Leave a Reply