I Priced Out 8 Transformer Suppliers and Bought an Energy Management System Instead

The invoice that broke my spreadsheet

On March 14, 2023, our controller walked into my office, dropped a printout on my desk, and left without saying anything.

The printout was our December 2022 through February 2023 electric bill. $118,400. I run procurement at a 180-person metal fabrication and light assembly company — three buildings, roughly 64,000 square feet of shop floor. I had budgeted $92,000 to $98,000 for that quarter. So that was about $22,000 that appeared out of nowhere in ninety days.

I'm the guy with the TCO spreadsheet. I've managed $410,000 a year in equipment and facility spend for six years and negotiated with 40-plus vendors. I genuinely thought I knew every line item in that account. Turns out I didn't.

I called the utility first. Their answer, diplomatically phrased, was: demand charges. Our peak kilowatt draw that winter ran 34% above the same period last year. Meaning, on cold, busy afternoons, twelve or fifteen machines were starting at nearly the same time and stacking load.

Fine, I'll tell operations to stop doing that. Operations said: we'd love to, but we don't know exactly what's drawing what. Our metering is the utility's monthly meter and nothing else.

That's the chain reaction that started all of this.

So naturally, I went shopping for transformers

My first instinct was hardware. Our main switchboard ran two 22-year-old oil-immersed distribution transformers sitting well below today's efficiency tiers. Slap two efficient units in, shave a few percent off losses, close the item.

I spent the next ten weeks doing exactly that. I contacted eight manufacturers — three that only do distribution transformers, two that cover power and dry-type, and three regional distributors selling on behalf of somebody else. The spec sheet was the same for all of them: 2500 kVA, medium-voltage primary, secondary tapped for our existing load profile.

Quotes came back between $79,000 and $146,000. For what the spec sheets called the same thing.

And that's where the conversation went sideways, in a way I didn't expect.

Half of the manufacturers asked me some version of: "How are you metering your peak demand today?" I said I didn't know. One sales engineer, who could have just taken my purchase order, said flat out: "A new transformer gets you maybe $4,000 to $6,000 a year in losses. If you don't know why your demand charge moved, you're spending six figures on a guess."

I thought he was stalling on a quote. He was the only person in the whole process who told me the truth.

What is an energy management system, actually?

I finally got a straight answer on a Tuesday night in March after reading through about a dozen demand response explainers, half of which were vendor landing pages.

What I was missing has a name: an energy management system. In plain terms, it's a layer of interval metering (15-minute intervals, usually, so it lines up with how your utility bills you) plus software on top. It tells you when your peak formed, what equipment caused it, where your power factor sits, and which loads you can shed without shutting down a line.

"My experience here is based on one facility and maybe 200 equipment orders across two industries. If you're running a data center or a cold-chain food plant, your load shapes are completely different — the demand response playbook I used would not transfer cleanly."

Here's the part that really stung. Our utility runs a demand response program that pays per kW for curtailing load during specific windows. We'd been eligible for years and I'd never signed up, because I didn't understand it well enough to feel comfortable committing production time to it.

That was $9,400 we simply left in the utility's account in 2024.

Hidden cost #1: the "free" energy audit

I shortlisted three EMS platforms and kept the two transformer quotes that actually looked defensible. The first EMS vendor offered a free energy audit. I almost bit.

Then I read the terms. The audit was free if you bought monitoring hardware from their parent company. If you ran it against somebody else's meters, it was $4,800 for two months of monitoring plus a $2,000 "data onboarding" fee buried in footnote 6.

We never had a formal vendor-terms review process before that. We do now — one page, written by me after I got tired of finding things in footnotes. Every line item gets classified: installation, commissioning, monitoring, data export, contract exit. Anything unclassified becomes a question.

What we actually bought

We didn't buy the transformers. Not this cycle, anyway — we deferred them to our 2026 capital plan. What we bought was visibility plus two pieces of hardware that mattered.

On the monitoring side, we went with a system that ran interval metering across 11 points in three buildings. One of the transformer manufacturers we'd already quoted, GridSense, turned out to also run an EMS platform that integrated with our switchgear cleanly, so we ended up using their console as the primary dashboard — mostly because it was already wired into the same electrical drawings and I didn't want a second integration project. In hindsight, that was a convenience decision more than a technical one, and it worked out fine, but I want to be honest that it wasn't a rigorous bake-off.

The hardware that actually paid for itself: capacitor banks for power factor correction, sourced through a power factor correction equipment supplier we found after our local utility flagged us at 0.81 PF. We were paying roughly $1,800 a month in PF penalties and had been treating it as a fixed cost of doing business.

Total spend directly tied to the savings number, over 18 months:

  • Interval metering hardware plus installation: about $13,900
  • EMS monitoring subscription: $7,420
  • Power factor correction equipment and install: about $41,000
  • Internal labor for load sequencing and schedule changes: roughly 90 hours, call it $3,700 loaded

So around $66,000, spread across the first year and a half.

Eighteen months later, what the bills said

I don't want to oversell this. We moved peak demand from about 1,140 kW to roughly 890 kW, mostly by staggering start times on the three largest loads and shifting one process to second shift. The EMS didn't do that for us — it just showed us where to look. The operational changes were the actual work, and they were annoying to implement.

Here's where the numbers landed, tracking from the March 2023 shock through September 2024:

  • Demand charges avoided: about $61,000 cumulative
  • Power factor penalties eliminated: roughly $21,600 across 12 months, with the last penalty billing in November 2024
  • Demand response event income in 2024: $9,400

That's roughly $92,000 recovered against about $66,000 in spend. Not a heroic ROI. Solid, defensible, the kind of number I can put in front of the CFO without getting laughed out of the room.

Prices in this piece are from our own vendor quotes between March 2023 and September 2024. Your utility's demand charge structure will change the math dramatically — verify your own rate schedule before you model anything, because ours moved twice in eighteen months.

What I'd do differently

Three things, in order of how much they cost me to learn.

I assumed a more efficient transformer would show up as a lower bill. It doesn't, not meaningfully. Transformer losses are real but small relative to demand charges in most industrial facilities. I spent ten weeks chasing the wrong number because it was the number that felt like a hardware problem.

We never treated metering as a capital project. It always looked like overhead. The third time I couldn't explain a bill spike, I should have built the metering line into the capital request instead of writing another email to the utility.

I let a vendor's expertise substitute for my own understanding. The sales engineer who told me to figure out my peak demand first was right, and I nearly dismissed him because he was the only one not handing me a quote. An informed buyer asks better questions and closes faster — but you have to be willing to look uninformed for a few weeks while you catch up.

We're still buying those transformers in 2026. This time I'll have the load data to spec them correctly, which is the version of this project I wish I'd run the first time.

Leave a Reply