Data and research
Data and research
Public commercial electricity data, checked and republished with its sources, and the original research we intend to publish only once there is enough data behind it — with methodology, sample sizes and stated selection bias, or not at all.
Published now: commercial electricity prices by state
The average commercial electricity price for every state and DC, from the U.S. Energy Information Administration: final annual figures since 2010 and the trailing 12 months from EIA's monthly survey. It is public-domain government data. We publish it as a clean, dated, cited table because the original is spread across several workbooks, and because every number is checked before it appears.
- commercial-electricity-prices-by-state.csv — one row per market and the US: price, change on the previous 12 months, rank, and where EIA reports it the energy and delivery split.
- commercial-electricity-prices-by-state-annual.csv — every market, every year from 2010.
- commercial-electricity-average-bill-by-state.csv — the average commercial account's monthly use and bill in each state and the US, from EIA's final annual sales, customer counts and prices. Read it with the page that explains what an average over every account hides.
- commercial-electricity-price-by-utility.csv — every U.S. utility with at least 500 commercial customers: how many, the average price it billed them (bundled and delivery-only kept apart), the counties it serves and the state average beside it, from EIA's final annual utility data. The page says what an average over every customer hides.
- commercial-building-electricity-use-by-type.csv — kWh per square foot per year for 21 building types and their regional ranges, from EIA's 2018 building survey, with sample sizes. The same figures as cost per square foot are on the cost per square foot page.
- commercial-building-price-by-size.csv — how much more or less each size of building paid per kWh than the average building in the same survey.
Each figure is the average revenue per kWh from every commercial customer in the state, so it is context for reading a bill, not a quote. The same table, ranked and sortable, is on the rates by state page. Free to use with credit to the U.S. Energy Information Administration and this site; the methodology says how the figures are checked.
What we intend to publish
| Product | What it shows | Cannot publish until |
|---|---|---|
| Commercial Rate and Renewal Index | Effective all-in cost by territory and usage band, supply versus delivery share, demand exposure, contract term mix, renewal timing patterns | At least 30 reconciled accounts per published cell |
| Renewal calendar report | Distribution of commercial contract expiry dates by market and segment | Sufficient consented contract records |
| Supplier response benchmark | Quote completeness, response time and conforming bid rates by market | A meaningful number of completed bid events |
Publication rules
- Minimum cell size of 30 records. Below that we publish nothing and say "insufficient data". It is a floor, not a guideline.
- Suppression where any single account is more than 20 percent of a cell.
- Ranges and percentiles, never an individual account's figures.
- Sample size stated on every published statistic.
- Selection bias named. Bills sent to us will come mostly from businesses who suspect they are overpaying. That skews the sample and we will say so on the page rather than imply it is representative.
- No executable supplier quote is ever exposed, and nothing traceable to an identifiable customer is published.
Why we will name our own selection bias
Bills sent to us will come from businesses who already suspect they are overpaying. That is a convenience sample, not a representative one, and any index built from it will skew high relative to the true population.
We will say so on the page, every time. A published statistic that names its own limits is more useful to a careful reader than one that implies a rigour it does not have — and a source that is honest about its sample is the one worth citing.
No original research is published yet. There is no corpus of our own. Publishing an index now would mean fabricating one, which would destroy the only thing that makes an index worth citing.