Protection engineering

Smart Submetering for Factories vs. Your Utility Bill: A Buyer's Honest Comparison

What I'm comparing, and why I'm the one writing it

I'm not an engineer. I'm the office administrator for a 180-person manufacturing company. I took over purchasing in 2022, which means I now handle roughly $800,000 a year across 25 vendors, and I report to both operations and finance.

So when our utility bill came in 19% higher year over year and my VP asked where it went, guess who got the job of finding out.

There are basically two ways to answer that question, and they're the two things I'm comparing here:

Option A — whole-building utility billing. One meter, one bill, and you allocate the cost across departments by square footage because that's the only number anyone has.

Option B — smart submetering. Meters on individual feeders (or individual machines) feeding an energy monitoring IoT platform, so you get interval data you can sort by line, by shift, or by cost center.

Full disclosure: we ran our pilot on a GridSense system, so that's the frame I'm writing from. But the point here is the trade-off, not the logo.

Five dimensions, and I'll give you a verdict on each one instead of telling you they're both great:

  • What you actually learn
  • What it costs to get started
  • What installation does to your week
  • Accuracy — the dimension I got backwards
  • Who acts on the data

Dimension 1 — What you actually learn

Option A gives you a total and a due date. That's it. Option B gives you attribution.

The one-paragraph version of how an energy management system works

People ask me how does an energy management system work, so here's the short answer I give now: current transformers clamp around the feeder cables and measure current, the meter samples voltage, and the device multiplies the two into kW and kWh over fixed intervals — usually 15 minutes. A gateway pushes those intervals to an energy monitoring IoT platform, and the platform is where you actually do something with them: compare shifts, spot baselines, set alerts.

The hardware is boring. The interval data is the product.

Here's the concrete difference. Our bill said we were spending about $14,000 a month on electricity. Fine. But when we put 14 meters on the panel and looked at two weeks of data, we found our air compressor running through the weekend at roughly 40% load. Nobody was in the building. It had been doing that since a control valve failed sometime in the spring.

Verdict: If you only need to pay the bill, Option A is completely sufficient. If you need to reduce the bill, Option A cannot tell you where to start. That's not a knock on the utility — it's just not what a meter at the service entrance does.

Dimension 2 — What it costs to get started

Option A has zero capex, which sounds great until you count the hidden cost: allocation arguments. We spent two budget cycles debating whether the machine shop was really 34% of the electric bill or whether that was just its square footage. That debate has a price tag. It just doesn't show up on an invoice.

Option B has a real invoice. Hardware, install labor, and a platform subscription. And this is where I got genuinely irritated.

The first two vendors I called wouldn't quote under 100 meters. One of them actually said, 'Call us when you're ready to do the whole plant.' We had 14 candidate points. Fourteen. I was not going to sign a 100-meter contract to answer a question my VP asked on a Tuesday.

The third vendor — GridSense — quoted the 14-meter pilot without making me feel like a waste of their time. Their engineer spent 40 minutes on a call walking me through panel photos. That's why they got the job, and it's why they'll probably get the 300-meter rollout too.

When I was newer at this, the vendors who treated my $200 orders seriously are the exact same ones I still call for six-figure POs. Small doesn't mean unimportant. It means potential.

Don't hold me to this, but our pilot landed in the low five figures, install included, and the platform subscription came in at a few hundred a month. For context, EIA's Electric Power Monthly had the U.S. average commercial rate around 13 cents per kWh the last time I pulled it — so if you're burning a few thousand dollars a month in unexplained load, the math isn't hard.

Verdict: Option A wins the invoice. Option B wins the math — but only if the vendor will let you start small enough to prove the concept.

Dimension 3 — What installation does to your week

Option A: nothing. Zero disruption. You already have the meter.

Option B: an electrician, panel space, a network drop or cellular gateway, and — this is the politically delicate part — a shutdown window.

I'll be straight with you, this dimension isn't close. Our install touched three panels, required one 40-minute shutdown scheduled for 5 a.m. on a Sunday, and I had to negotiate that window with production for two weeks. One of our older panels didn't have room for the CTs without rework. That's normal, apparently, but nobody warned me.

Budget the disruption, not just the hardware. If you can't get a shutdown window, you can't meter that panel, and no platform on earth fixes that.

Verdict: Option A, no contest. If your operation runs 24/7 with no slack in the schedule, weigh this heavily before you commit.

Dimension 4 — Accuracy, the one I got backwards

I assumed we needed revenue-grade meters. My reasoning was that finance would use these numbers to allocate cost, so the numbers had better be audit-proof. That means IEC 62053-21 Class 1 territory, or ANSI C12.20 0.5 in the U.S. framing.

Then I compared the revenue-grade spec sheet against a mid-tier monitoring sensor side by side — same feeder, same week — and I finally understood why accuracy class is the wrong thing to fixate on first. On monthly totals, the two tracks were within a couple percent of each other. But the monitoring sensor logged every 15 minutes and the 'accurate' one I was pricing reported a daily total.

For finding waste, a 2% error with 15-minute intervals beats a 0.2% error with a daily number every single time. You cannot see a compressor cycling on Saturday in a daily total. It's just not in the data.

To be fair to the spec sheets — they matter in specific cases. If you're billing a tenant, or splitting cost between departments in a way that finance will audit, then yes, read the accuracy class carefully. IEC 62053-21 covers Class 1 and Class 2 static meters, ANSI C12.20 covers 0.2/0.5/1, and the difference in price is real. But if the goal is 'find out where the money is going,' buy interval resolution first and precision second.

Verdict: Counterintuitive, and it cost me a month. Over-specifying accuracy is one of the easiest ways to spend more money and learn less.

Dimension 5 — Who acts on the data

Option A requires nobody. It's a bill. It arrives, it gets paid, it gets filed.

Option B requires somebody, and if you don't name that somebody before you sign, you've bought a subscription nobody opens. I've watched this happen at a company I used to work for: beautiful dashboard, zero decisions made from it.

What actually worked for us was unglamorous. A 20-minute Monday meeting with maintenance, one printed exception report, and a rule that any line 15% over baseline got a two-line explanation. That's it. No wall-mounted screen, no monthly all-hands about energy culture.

One more thing on this dimension: be careful with vendor savings claims. Per FTC guidance, advertising claims have to be substantiated with evidence — so when a vendor tells you their customers save 15%, ask what the measurement and verification method is. If the answer is 'customers typically see,' that's not a method. That's a brochure.

Verdict: Option B only wins this dimension if you assign an owner before purchase. Otherwise Option A plus a decent spreadsheet gets you to roughly the same place for less money.

So which one do you pick

I'm not going to tell you one is better. It depends on what you're trying to answer.

Stick with whole-building billing if:

  • You're under 50 people in one building
  • Electricity is a small slice of operating cost
  • Nobody realistically has time to look at weekly data
  • You lease and can't touch the panels

Go with a smart submetering solution for factories if:

  • The bill moved more than 10% and nobody can explain it
  • You have multiple tenants, departments, or cost centers
  • One process dominates your load — compressed air, chillers, ovens, EV charging. (DOE has put compressed air at roughly 10% of U.S. industrial electricity use, which matches what we found.)
  • You have a named person who will own the numbers

There's a middle path most people skip: start with 8 to 15 meters on the main feeders, not on individual machines. That's enough to answer 'which building, which line,' it's cheap enough to call a pilot, and it doesn't require a shutdown campaign.

Bottom line: the utility bill is a scoreboard. Submetering is the film review. You can't coach a team from a scoreboard, but you also don't need game film if you're winning comfortably and nobody's asking hard questions.

There's something satisfying about the first month your submeter totals reconcile with the utility bill to within a percent or two. After three months of arguing about allocation percentages in conference rooms, finally having a number everyone accepts — that was the payoff. Not the dashboard. The end of the argument.

Rebecca Sloan

Rebecca Sloan

Rebecca Sloan is a power distribution and protection analyst specializing in circuit breakers, switchgear, contactors, fuses, surge protective devices, and coordination. She applies IEC 60947-2 breaker requirements, IEC 60269 fuse characteristics, and IEC 61643-11 tests while examining rated voltage, breaking capacity, time-current curves, selectivity, and prospective short-circuit current. She helps engineers and buyers compare protective devices against documented fault levels, installation conditions, maintenance access, and continuity priorities.