One model for 18 sites

Replacing a pile of separate reports with a single cost model, plus a sample built from scratch that you can download.

MARA · Lead Financial Analyst, Operations & Energy Transformation · 2025 to 2026

power-cost-model-sample.xlsx

K10fx=J10-I10

	B	C	D	E	F	G	H	I	J	K	L	M	N
1	Summary
2	Full year 2025, six sites. Made-up numbers.
3	Model checks	All checks pass
4
5	By site, full year 2025
6	Site	Contracted capacity (MW)	Metered (MWh)	Average load (MW)	Load factor (%)	Budgeted (MWh)	Metered vs budget (%)	Expected cost ($)	Invoiced cost ($)	Variance ($)	Variance (%)	All-in cost ($/MWh)	Flagged months (count)
7	Site A	120.0	989,816	113.0	94.2%	998,640	(0.9%)	$44,237,358	$44,284,111	$46,753	0.1%	$44.74	-
8	Site B	75.0	596,666	68.1	90.8%	611,010	(2.3%)	$27,019,725	$27,020,782	$1,057	0.0%	$45.29	-
9	Site C	48.0	400,406	45.7	95.2%	403,661	(0.8%)	$21,132,254	$21,130,981	($1,273)	(0.0%)	$52.77	-
10	Site D	90.0	723,011	82.5	91.7%	741,096	(2.4%)	$35,885,214	$37,372,784	$1,487,569	4.1%	$51.69	8
11	Site E	22.5	189,709	21.7	96.3%	191,187	(0.8%)	$10,597,708	$10,599,363	$1,656	0.0%	$55.87	-
12	Site F	97.0	759,118	86.7	89.3%	781,742	(2.9%)	$34,449,671	$34,420,250	($29,421)	(0.1%)	$45.34	-
13	Fleet	452.5	3,658,726	417.7	92.3%	3,727,336	(1.8%)	$173,321,930	$174,828,271	$1,506,342	0.9%	$47.78	8
14
15	Average load is metered energy over the hours in the year. Load factor is average load over contracted capacity.
16	Variance is invoiced minus expected. A positive number means the bill was higher than the contract terms and meters support.
17	A flagged month is one where the bill is further from the meters or the expected cost than the tolerance allows. Send the utility the meter data for that month and ask for a corrected bill.
18	The Fleet row comes from the Fleet column on Calc, not from adding up the rows above, so Checks can test one against the other.
19
20	By month, all sites
21	Month	Metered (MWh)	Expected cost ($)	Invoiced cost ($)	Variance ($)	Variance (%)
22	Jan 2025	314,808	$13,888,826	$13,897,101	$8,275	0.1%
23	Feb 2025	270,690	$15,529,250	$15,532,124	$2,874	0.0%
24	Mar 2025	320,482	$13,369,762	$13,363,639	($6,123)	(0.0%)
25	Apr 2025	310,360	$12,561,388	$12,579,547	$18,159	0.1%
26	May 2025	319,341	$13,831,248	$13,996,280	$165,032	1.2%
27	Jun 2025	300,516	$14,336,307	$14,523,400	$187,093	1.3%
28	Jul 2025	292,886	$16,889,712	$17,110,161	$220,450	1.3%

Read meSitesMarket pricesMeter dataInvoicesCalcSummaryChecks+

At MARA, each site’s power costs lived in its own reports. A fleet-wide question, like what we paid per megawatt-hour last month or which site ran over budget, meant stitching those reports together before anyone could answer it.

I built one model in Excel that became the single source for power costs across all 18 sites. It tracked each site’s usage, capacity and costs, and the gap between what we expected to pay and what we were billed. It connected to DOMO for reporting and to Zip for approvals, so leadership could see the whole fleet in one place, kept current.

The sample model

The real model belongs to MARA and stays there. This sample is built from scratch with made-up numbers to show how I structure a model like it: six sites totaling 452.5 MW and a year of monthly data. It follows the conventions I’d expect from any model handed to someone else:

Inputs are in blue, formulas in black, and numbers pulled from another tab in green.
Inputs live on their own tabs: sites and contracts, market prices, meter data and invoices.
The Calc tab builds each site’s expected bill from its contract and compares it with the invoice, month by month.
Summary is the one-page view, and Checks confirms that everything ties out.

Site D is there to show what a billing problem looks like. Starting in May, its invoices bill 5.6% to 6.8% more power than its meters recorded, and the model flags all eight of those months. Over the year, that comes to about $1.5 million billed above what Site D’s contract says it should have paid.

Download the sample model Excel, 38 KB

Exhibit 1 The Summary tab, from the sample: full year

Site	Metered (MWh)	Load factor	Expected ($000)	Invoiced ($000)	Variance ($000)	All-in ($/MWh)	Flagged months
Site A	989,816	94.2%	44,237	44,284	47	$44.74	0
Site B	596,666	90.8%	27,020	27,021	1	$45.29	0
Site C	400,406	95.2%	21,132	21,131	(1)	$52.77	0
Site D	723,011	91.7%	35,885	37,373	1,488	$51.69	8
Site E	189,709	96.3%	10,598	10,599	2	$55.87	0
Site F	759,118	89.3%	34,450	34,420	(29)	$45.34	0
Fleet	3,658,726	92.3%	173,322	174,828	1,506	$47.78	8

Made-up numbers. Expected cost is built from each site’s contract and its metered usage; variance is what the invoices billed above that. Negative numbers are in parentheses.

Exhibit 2 Site D: how much more power the invoice billed than the meter recorded, by month

0% 2% 4% 6% 8%

Made-up numbers from the sample. The model flags any month where billed MWh or billed dollars differ from what the meters and the contract say by more than the tolerance set on the Read me tab.

Next Archibald, the system that runs my admin →

Questions about the model? Email me at jlraney99@gmail.com.