$31,416.26
Charges Xcel canceled
Case study · Brasa Rotisserie & ALMA Group · April–July 2026
One Minneapolis restaurant’s bill recorded a peak ten times higher than reality. Eleven bills quietly inflated. Here is how we found it, proved it with Xcel’s own data, and got it corrected — to the dollar.
In April 2026 we ran a portfolio audit for Brasa Rotisserie and Alma — five sites, seven electric meters, all on Xcel’s General Service schedule. Our first pass reconstructs every line of every bill against the posted tariff and active riders. Every meter reconciled to within two cents — the billing math was clean.
But one meter’s inputs weren’t. It was billed perfectly — and it was paying demand charges on 102 kW of billable demand in a month when the building’s actual peak was 35.9 kW. That gap has a name: a demand ratchet. On A14, Xcel bills each month against the greater of the current peak or 50% of the highest adjusted peak from the preceding 11 months. Something in the past year had set an anchor almost three times the size of the building’s real load.
Fourteen months of statements traced the anchor to a single read. The August 2025 bill recorded 377.36 kW of actual demand — on a meter whose actual demand never registered above 38.66 kW in any other month. Xcel’s own systems flagged it enough to cap that month’s billable demand at 139 kW, but the 377.36 stayed on the books as the demand of record, and the ratchet did the rest.
| Billing month | Actual peak (kW) | Billed demand (kW) |
|---|---|---|
| Aug 2025 | 377.36* | 139 |
| Sep 2025 | 35.10 | 119 |
| Oct 2025 | 37.37 | 116 |
| Nov 2025 | 26.73 | 89 |
| Dec 2025 | 26.86 | 85 |
| Jan 2026 | 27.58 | 90 |
| Feb 2026 | 27.86 | 94 |
| Mar 2026 | 35.94 | 102 |
*The recorded August value. Interval data showed the true peak that month was 37.74 kW.
The customer authorized Green Button Connect — Xcel’s own interval-data export — and we pulled 121,007 five-minute readings covering fourteen months. For every month where both sources existed, the bill’s recorded peak matched the interval data to two decimal places. Except August 2025.
The interval data put the true August peak at 37.74 kW, on a Sunday afternoon at 3:15 PM. The bill said 377.36 kW — ten times higher, consistent with a single decimal-point shift somewhere between the meter register and the billing system.
The physics sealed it. The August bill’s own energy total — 13,473 kWh — matched the interval data within 1%. At a sustained 377 kW, that entire total would be gone in 36 hours. The peak and the energy total could not both be true, and they came from the same meter.
The building’s team later recalled the likely trigger: electrical work on the building’s exterior boxes that summer, including a corroded box replaced by the landlord, with power lost during the work. Meter events follow panel work — one of the patterns we now screen every audit for.
The dispute letter — dated April 30, 2026, with our exhibits attached — made no accusations and asked for no benefit of the doubt. It asked Xcel to reconcile two of its own records that disagreed: the billing register and the interval data. It requested the August reading be corrected, the ratchet recomputed from the corrected value, and the difference credited.
Our exhibits estimated $9,822 of overcharge through the March 2026 statement. Xcel’s own recomputation of those same periods came to $10,152.62 — our estimate ran 3.3% conservative.
Xcel’s corrected statement is dated July 1, 2026. It cancels all eleven invoices — read periods spanning July 21, 2025 to June 21, 2026, $31,416.26 of charges — and rebills every period at the building’s actual demand, ratchet removed. The corrected August demand on the rebill: 37.736 kW. Exactly the disputed reading, divided by ten.
“We found a billing error on your account. To correct our error, we canceled previous charges and rebilled your account to reflect the accurate amount owed… We apologize for the inconvenience.” — Xcel Energy, July 1, 2026 statement
Net credit: $13,590.07. Before the outcome, our audit had put the error’s end-to-end cost at roughly $13,500 if left uncorrected — the forecast landed within 0.7% of what Xcel ultimately credited.
“This is quite impressive. Quality work, delivered seamlessly.”
Lilia Spitzack · Accountant · ALMA GroupNothing about this looked like a problem. The bills arrived on time, the arithmetic was correct, the amounts drifted the way energy bills drift. An attentive accountant paid every one of them, correctly, for eleven months — because each bill was internally consistent. The error lived in the one place nobody looks: the relationship between months, and between the bill and the meter’s own interval record.
That is the audit. Twelve months of PDFs, about an hour of your time, and every meter you run checked against the tariff and the interval data — for ratchets, wrong schedules, power-factor penalties, and readings that don’t match reality.
Free · No win, no fee on disputes · Xcel Energy MN