NerdyWatts prices the whole filing, not the headline rate.

Utilities publish filings, not prices

Every US residential electricity tariff is public, and almost none of them are published as a price. One rate can carry a tier ladder whose thresholds reset daily rather than monthly, a time-of-use schedule that changes by month and by weekday, a fixed charge billed per day instead of per month, and a rider added to the headline rate rather than included in it.

Get any one of those wrong and the answer isn't slightly off, it's wrong by a multiple. NerdyWatts decodes the filings and prices them against what a household actually uses, so what comes back is an annual total with the fixed charges already in it, rather than a cents-per-kWh figure that leaves part of the bill out.

The filing date matters as much as the rate

A tariff sitting in the national database with no end date is not necessarily in force; it means nobody recorded an end. Filings more than a decade old run about 17% below what their utility now collects, so ranking on price alone puts the stalest data at the top of every cheapest-rate table, because being out of date and being cheap look identical in the arithmetic.

So only tariffs filed since 2020 are ranked, and every figure carries its filing year. In Georgia that is the difference between naming a cheapest rate of $614 a year, off a 2018 filing, and $1,224. It also means publishing 911 tariffs out of the 6,357 on file: a much smaller site than the data would support, and the right size for one that quotes prices.

What it won't estimate

Some rates can't be priced from consumption alone. They need the customer's peak demand in kilowatts, or their tier ceilings are indexed to a motor rating, or their time-of-use matrix is missing from the filing. Those get a reference page showing the tariff and no estimated bill at all, which makes for a thinner page and an honest one.

The upstream data is crowd-maintained and has transcription errors in it, and none of them are quietly corrected, because a published figure has to be traceable back to its source. A fixed charge filed in the wrong units, or a demand charge typed into an energy tier, is flagged instead: kept out of the rankings, shown with a caveat, and reported back upstream. The parts that are modelled rather than measured get the same treatment, so the usage pattern behind every time-of-use figure is written out in full on the methodology page, because it is an assumption and it changes the answer.