Editorial policy · reviewed
What we publish, and what we will not
Every number on this site is derived from something a seller actually published, and every derivation is reproducible from what we show. Where we cannot derive one, the space stays empty.
Who writes this
An independent publisher with no ownership stake in any seller listed here, no board seat at one, and no manufacturer behind it. Nobody writing for this site is a licensed clinician, and no page claims a clinical review that has not happened — which is why the drug-facts pages report what the approved label and the published trials say instead of offering an opinion about them.
The site holds no certification, accreditation or seal, and it does not display one. What it offers instead is the source and the date beside every figure, so an assessment can be checked rather than trusted.
What counts as a source
In order of strength: the page where a seller actually takes money — a checkout, a plan selector, a live price table on its own site. Then the seller’s own published pricing page. Then a figure from somewhere else entirely, which we mark reported and never describe as listed.
A press release is not a source. A comparison table on another site is not a source. A seller emailing us its own price is not a source either, however cooperative the seller — if it is not on a page we can point at, it does not go in. A price inherited from another dataset is a starting point for a check, never the result of one, and it can never be recorded as verified.
Where a price sits behind an intake form, the funnel is walked with dummy answers until the price appears, and the page it appeared on is recorded. Where a seller shows a price only after a consultation, there is no figure to publish and the cell stays empty.
Prices are computed, never copied
We do not repeat a seller’s “per month”. We take the amount billed and the number of days it covers, and compute the monthly cost from those two values — so a wrong figure means a wrong input rather than a typo, and anyone can check the arithmetic.
Sellers bill on cycles that are not calendar months: every four weeks, every eleven weeks, per twelve-week course, per year. A charge billed every four weeks lands thirteen times a year, so calling it monthly understates the year by about eight percent. Restating it against a month of 30.4375 days is what makes two sellers comparable at all.
Where the seller’s own information is ambiguous — two different cycles, or an annual total quoted beside a four-week charge — we publish no computed figure at all, because a confident number built on a guess is worse than a blank. A mandatory membership or a periodic consult fee is part of what a month costs, so it goes into the figure and is shown separately beneath it. A price that covers only the starting dose is labeled as one, because the number climbs as the dose does.
The evidence tiers
Every row carries one, and the tier is as much a part of the fact as the figure:
- Verified — someone opened the seller’s own page on a stated date, read the price there, and recorded it with the URL. The strongest thing this site can say about a number.
- Listed — the seller publishes the figure, and it has been read from that publication rather than walked to a checkout. It becomes verified when someone walks it.
- Reported — the figure comes from somewhere other than the seller. It is shown because a gap is less useful than a sourced third-party number, and it is never described as the seller’s own listing.
A tier can be earned and it can lapse. It cannot be bought. How every price is checked →
How often we re-check
A walked price is evidence about the day it was walked. After 90 days it stops being described as verified and falls back to what the seller publishes — automatically, as a property of its own date, so the roster cannot quietly rot while the badges stay green.
Re-checks also run out of turn when there is a reason: a seller redesigning its pricing page, a reader writing in, or one price moving in a way that suggests the rest have too. Every check we have run, and what each one found, is published at what changed — including the checks that found nothing and the ones that found we were wrong.
How FDA letters are screened
A warning letter on a seller’s page here means FDA’s own letter names that seller’s own domain. Not a similar company name, not a brand that sounds the same — the letter, opened and read, with the domain in its body. Screening by name alone produces both kinds of error: it invents letters against innocent sellers, and it misses the letter filed against an operating company that trades under four different brands.
FDA’s search endpoint cannot be relied on for this, because it will silently ignore a search term and return everything, which reads exactly like a clean result. So the screen runs against the complete letter corpus held locally, and it has to re-find every letter already on file before any page is allowed to report that a seller has none. An absence is only publishable when the instrument has been shown to work.
A letter never removes a listing. A seller holding one stays in the index at its real price, in its real position, with the letter shown on its own page and in the list of every seller that holds one. Hiding the seller would hide the letter with it, and the reader most in need of that fact is the one comparing prices. What a letter changes is on our side: a commission from that seller is declined. The full list →
The clinical pages
Anything clinical is sourced to the FDA prescribing information for the approved product, cited to its section and its revision, or to a paper resolved by its PubMed ID and checked against the record before a word is written about it. A citation is never written from memory, and a negative claim — that a label says nothing about some topic — is only publishable when the whole document has been searched for the words that would settle it, and those words are named on the page.
These pages stay carefully separate from the price index, because the clinical evidence for tirzepatide was generated on the approved product and a compounded vial does not automatically inherit it.
What a paying seller cannot buy
Position, first of all. The table is sorted by computed monthly cost, ascending, and there is no other order — so there is no position to sell. A seller that pays us and a seller that has never heard of us appear wherever their prices put them.
Nor can a commercial relationship remove an unflattering number. Where an advertised price differs from what a seller bills, the difference is shown either way. Where a price is a starting dose that climbs, the page says so. Where a fee sits on top of the medication, it goes into the monthly figure. Where an FDA letter exists, it appears.
No seller sees a page about itself before it publishes, and no seller can approve, edit or veto a conclusion afterwards. A seller’s remedy is the same as anyone else’s: show us the page, and the record changes. How the site makes money →
What is automated, and what a person reads
Automated: the arithmetic. A charge is divided by the days it covers and multiplied by 30.4375, by one function used everywhere, so a figure in the table and a figure from the calculators cannot disagree. Sorting, ranking, the charts, and the expiry of a stale verification are computed too.
Not automated: deciding what a seller’s page actually says. A price becomes verified only when a person opens that page, reads it, and records what it said with the URL and the date. The same goes for a billing cycle, a service area, and whether a quoted price covers every dose or only the first one.
We use software tools, including AI tools, to gather and to draft. They do not decide what is true here: every published price traces to a record with a source and a date, and anything that cannot be traced that way is not published. No clinical claim, no price, no billing cycle and no citation reaches a page on a tool’s say-so.
How we correct a mistake
Corrections are made to the underlying record, not to the sentence that happened to show the error. One record feeds the index, the state pages, the dose pages, the seller’s own page and the exported data, so a fix reaches all of them at once — a correction that only repairs the page somebody complained about is not a correction.
When a price changes because the seller changed it, that is not a correction: it is a new check, and it appears in the log with its own date. When it changes because we were wrong, the log says so.
If you can show that a figure here is wrong, tell us what the seller’s page says and where. Evidence beats argument, and the seller’s own page beats both. The corrections policy in full →
What we will not publish
- A rating, a score, or an editor’s pick. We run no rating system, and a badge with no computable basis is inventory dressed as advice.
- An estimated price. Where a seller publishes nothing, the cell stays blank.
- A dose, a protocol, or anything that reads as instruction on how to use a medication.
- A claim about whether a seller is operating lawfully. We publish what a seller charges and what it says about itself; whether it meets a regulatory standard is not something a price index can determine.
- A citation nobody resolved, or a clinical claim written from memory.
- A credential, an audit or a clinical review that has not taken place.
- Sponsored copy of any kind, in any position, under any label.
Reaching the editor
Source requests, factual disputes and questions about anything on this page go to corrections@compoundedtirzepatide.org. If you want to know where a specific figure came from, ask — the record behind it has a URL and a date, and both can be sent to you.
Journalists, and anyone citing the dataset in their own work, are better served at press@compoundedtirzepatide.org — the exports are open, and a figure quoted with its date and its evidence tier is a figure a reader can check.