How we rate providers and check prices

A rating is not an opinion we assign. It is computed from the data a provider publishes, using the weights below. The two price criteria are scored against fixed anchors, not against each other. The anchors are stated with the weights below, because they are part of the formula and are read from it rather than retyped here. Every provider is measured on its own published numbers, so a rating moves only when that provider's prices or terms move. It never moves because a competitor changed something, which is what an earlier version of this formula did and why it was replaced.

Two criteria cannot be guessed: a dated monthly price and a first-year total. A provider missing either is not scored at all. State coverage is weighted but optional. Several providers publish real, dated prices and no coverage list, and withholding their score over a list they never published told a reader nothing. We could produce a number by estimating the price, and that number would look exactly like the others while meaning something entirely different. The same rule governs prices themselves: a price we cannot date is shown as not published, never as an approximation.

A criterion we could not establish is left out of the arithmetic rather than scored as a zero. The weights that remain are shared out again over what is left. So the table above is the full set of criteria, not the set applied to any one provider.

A company whose cancellation terms we never established is scored on the criteria we did establish. It loses nothing for a gap in our research, and it gains nothing either.

That makes a weight in the table a ceiling rather than a promise. State availability is the clearest case: no provider on this site carries a sourced, dated list of the states it prescribes in, so that criterion is applied to none of them. Speed to prescription is in the same position. Every rating you see rests on the criteria we could establish for that company, and the panel beside it names them.

Full marks on every criterion comes to 9.5, not 10. The headroom above it is not awarded by any formula. When we first ran the rubric the cheapest provider scored a flat maximum on four data points, and a perfect score on thin evidence is the clearest sign of a review site that is not doing the work.

Prices come from the provider's own published plan pages, by one of three methods, recorded per provider and published here. Where a provider's site blocks automated access the method is a person checking it by hand. We do not work around anti-bot protection: the cost of that is a slower check, and the alternative cost is a number we could not stand behind.

A provider that has shut down or is degraded never appears in a ranking and never carries a call to action, whatever its price. Its page stays up, with what changed and when. A provider that publishes no price is excluded from a cheapest-first ranking rather than placed at the bottom of it, because absence of a price is not the same as a high one.

Rankings are ordered by the criterion named in their own heading. Nothing is ordered by what a provider pays us, and no row is promoted. Where we have a conclusion we state it as a numbered finding above the table with its criterion named, so you can disagree with the criterion rather than guess at it.