How we score, published at the point value

Why there are two numbers and not one

A single score collapses two questions that behave differently. "How good is this offer" and "how honest is this company before you pay" can move in opposite directions, and a provider can be strong on one while failing the other. Folding them together hides exactly the case a shopper needs to see. So the 0-10 score answers the first, the A to F Transparency Grade answers the second, and both are printed side by side on every ranking surface on this site.

The score has a transparency dimension worth 15%, which overlaps in subject matter with the letter grade and is computed independently of it. The grade is never an input to the score.

The five weighted dimensions, with every sub-criterion

Weighted score = cost x 0.25 + formulary x 0.20 + clinical depth x 0.20 + patient experience x 0.20 + transparency x 0.15. Each dimension is capped at 10 points and floored at 0, so a stack of penalties cannot drive a dimension negative and drag the weighted total below what the evidence supports.

The reputation modifier, applied after weighting

Derived from the evidence file's reputation block and never hand-entered. It exists because a good offer from a company that stops answering its complaints is still a bad purchase.

The BBB F cap is live on one provider right now: Maximus at 6.5. A capped provider can still fall below 7.0 on the weights alone, and the cap only ever pushes a score down.

Relevance gate. Platform-wide ratings that reflect other product lines set trustpilotRelevant: false. They earn no rating credit and no band penalty, and they display with a platform-wide-rating disclaimer instead. A telehealth company whose 4.6 comes from its hair-loss line does not get to spend that rating on its hormone line.

The Provisional flag and the [Verified] / [Reported] split

A provider is flagged Provisional in the evidence record, with no score change, when the relevant product line is under 6 months old, or when fewer than 25 relevant-line independent reviews exist. It flags a provider we do not yet have enough independent history on, and it changes no number. Currently on 8 of 16: Sesame Care, Hers Menopause, Wisp, Gala Health, Gennev, Inner Balance, MangoRx, Telos RX.

Every scoring-relevant fact carries a status. [Verified, with a month and year]means it was read directly from the provider's own site or official page during that pass. [Reported] means it came from a third party and counts as evidence only when several independent sources corroborate it. A [Reported] fact is never presented as verified.

Our verification crawlers cannot read 7 of the 16 providers, so their facts sit at [Reported] until a manual browser pass upgrades them: Winona, Hers Menopause, Gala Health, Inner Balance, PeterMD, MangoRx, Maximus. That list is read from the evidence file's own crawler-access flag at build time, so it cannot go stale in prose while the data says otherwise. Two different things put a provider on it, and they are not the same accusation: some sites answer an automated request with an error and some serve a page whose prices only appear once a browser runs the scripts. Either way we could not read the number, and saying which providers we could not read is part of the method. A comparison that quietly upgraded unreadable claims to verified would be selling confidence it did not have.

Recompute any score yourself, including where we disagree with our own rubric

The rubric above is the whole rubric. The evidence behind it lives in src/data/scoring-evidence.json in this site's repository, one block per provider recording which sub-criteria the verification pass found met, with a note explaining each judgement. The arithmetic runs in scripts/recompute-scores.ts. 15 of the 16 providers carry a full evidence record. The one that does not yet: Elektra Health, added to the roster after the last full evidence pass.

Here is the part a competitor would leave out. The strict score the rubric produces is currently 1.6 to 5.6 points below the score we display. The reason is what the strict calculation counts. It credits only sub-criteria the evidence file records as directly confirmed, so anything we could not read counts as unmet. For the 7 providers our crawlers cannot read, that is most of the clinical and experience columns. The strict number is the floor of what we can prove about a provider rather than a verdict on it, and the displayed score is calibrated across the roster on the same evidence.

That gap is a debt and it is treated as one. Each provider's current gap is frozen in scripts/score-drift-baseline.json, and the enforcement script fails the build if any gap grows or if a provider outside that list drifts at all. Gaps may shrink. When a provider's evidence reaches verified status, its displayed score must equal its computed score exactly and any difference fails the build. Publishing the gap is how it gets paid down.

Dimension points are out of 10 each, before weighting. Modifier is the reputation adjustment applied after weighting. Every figure in this table is computed at build time from the evidence file, so it cannot disagree with the data it describes.

Worked example: the score leader, line by line

Alloy leads the roster at 8.9 and holds a Transparency Grade of A. It is not a commercial partner, so nothing about the placement below earns this site money. Here is the arithmetic.

The displayed score is 8.9, which sits 2.2 above the evidence-only figure. Where the gap comes from is readable in the table: Alloy takes 9 of 10 on transparency, the roster ceiling, shared with 1 other provider, and loses most of its clinical points to criteria it does not offer at all, such as video visits and lab work. The calibration reflects that transparency standing across the peer set. The floor is what we can evidence, the displayed number is where it sits against the field, and both are printed here so you can judge the distance for yourself.

What the rubric currently produces

Derived at build time, so this paragraph updates when the data does.

The rankings these numbers produce are at best HRT providers for the women's wing and best online TRT clinics for the men's.