How the index is built

What the index measures

The advertised monthly price of a GLP-1 weight-loss medicine at UK online pharmacies. The measure is deliberately narrow: it is the price shown to a new, logged-out visitor browsing from the UK, at a named dose, for roughly one month of supply. It is not what any individual ends up paying, and it does not attempt to be. Consultation outcomes, subscription tiers and personal discounts all move the real figure, in ways a public page cannot see.

The basket

The index tracks a fixed set of medicine and dose combinations. Fixing the basket is what makes month-to-month comparison meaningful: if the contents changed with availability, movement in the headline would just be movement in the basket.

• Mounjaro (tirzepatide) at 2.5mg, 5mg, 7.5mg, 10mg, 12.5mg, 15mg
• Wegovy (semaglutide injection) at 0.25mg, 0.5mg, 1mg, 1.7mg, 2.4mg, 7.2mg
• Wegovy pill (oral semaglutide) at 1.5mg, 4mg, 9mg, 25mg

Liraglutide brands, Orlistat and Mysimba are outside the basket: each appears at only one or two pharmacies in our dataset, which is too thin to median honestly, and they sit in a different part of the market. Needles and other add-ons are not medicines and never count. Basket changes happen only at a release boundary, with a note in the change log below, never silently.

Foundayo (orforglipron), a daily GLP-1 tablet, was licensed by the MHRA on 10 August 2026. The September and October 2026 releases collected its prices under exactly the same rules and published them in their own table. It was not in their basket, so it did not affect their index value or basket cost. Thin coverage was not the reason: 12 of the 23 pharmacies we track had a Foundayo price that counted in September 2026. The reason was the fixed basket. September was the first release since Foundayo was licensed, and adding it then would have made September measure something different from July.

Foundayo joins the basket from the November 2026 release, at 0.8mg, 2.5mg, 5.5mg, 9mg, 14.5mg and 17.2mg. It had no UK prices in July 2026, so its prices are compared with September 2026 instead, its first month in the index's data. The July, September and October 2026 releases are unchanged.

Which pharmacies

UK-registered online pharmacies and prescribing services that publish a price for at least one basket dose on a page a logged-out visitor can reach. The current set is 23 providers, listed with their GPhC registration on the price explorer. We include a pharmacy because it publishes prices, never because of any commercial relationship, and no pharmacy can pay to be added, removed or ranked.

What counts as a price

• Advertised, logged out, UK. The figure shown to a new visitor without an account, browsing from the UK.
• Renewal, not introductory. Where a pharmacy shows both a first-order price and an ongoing one, the index uses the ongoing price. First-order discounts are recorded separately and are never blended into the median, because an intro price is a marketing offer rather than the cost of treatment.
• Pay as you go, where there is a choice. From the September 2026 release, where a pharmacy sells the same dose both as a one-off order and on a subscription or longer plan, we record the one-off price and note the plan prices. Where it only sells plans, we record the shortest one.
• One month at one dose. A row must represent roughly four weeks of supply at a single named dose. Rows priced per two months, per three pens, per pack of capsules, as a "from" figure with no stated quantity, or per pack where the pharmacy does not say how many tablets a pack holds, are excluded and listed on the release rather than converted by guesswork.
• Starter and all-dose bundles are excluded. A single price covering a range of doses cannot join a per-dose median.
• Total, as advertised. Where a pharmacy quotes an all-inclusive price, that is what we record. Where consultation or delivery is charged separately and disclosed, the headline product price is what enters the index, and the extras are noted on the price explorer.

Which prices count

A price counts towards a median, and can be named the cheapest, only if all three of these are true:

• It is not a pre-order price.
• We have not rated it low confidence. Low confidence means the figure was inferred rather than read, or that the pharmacy's own pages give two different prices for the same dose.
• It was read in the release month or the month before. After 2 months without a reading, it stops counting.

A price that does not count stays on the release, with the month it was last read and the reason, until it can be read again. It is never carried forward as though it were fresh. The rule applies price by price, so a pharmacy can count at some doses and not at others. The price explorer applies the same rule against the current month: an older, pre-order or low-confidence price is listed separately there and is never marked cheapest.

How the number is calculated

1. For each basket dose, take every price that counts for the month and compute the median. Medians rather than averages, so one unusually cheap or expensive pharmacy cannot move the headline.
2. Add the dose medians together to get the basket cost for the month.
3. The index value is the basket cost expressed against the baseline month, which is July 2026 and is set to 100. A later month at 103.2 means the basket costs 3.2% more than it did at baseline. Since September 2026 the comparison is made against July recalculated on the rules above: the prices the July release published, minus its pre-order and low-confidence prices. The July release itself, and every figure in it, is unchanged. If a basket dose has no price that counts in a given month, that dose is left out of both sides of that month's comparison, and the release says which.
4. A medicine that joins the basket later has no price in July 2026 to compare with. Foundayo joins from the November 2026 release. Its prices are compared with September 2026, its first month in the index's data, so its own index is 100 at September 2026. Its September medians are added to the July base. The headline still compares what the whole basket costs with what the same doses cost at the base.
5. Alongside the headline we publish the spread for each dose, the gap between the cheapest and dearest pharmacy in the same month, because the variation is often a bigger story than the average.
6. Because the pharmacies behind a median change from month to month, each release also counts how the same pharmacy's price for the same dose moved since the baseline release, and what those same prices add up to in each. Foundayo has no prices in the baseline release, so it is not part of this count.

Every published figure is calculated from the prices listed with it, and nothing is rounded to convenient numbers. A median of an even number of prices can fall on a half penny; pages and CSVs show it to the nearest penny. We never print a forecast next to a measured series.

Collection

The July 2026 baseline was collected by hand in June and July 2026. Most prices came from each pharmacy's own pages. Where a page could not be read, some came from comparison websites or search results instead, and the July release now carries dated corrections saying which.

The September 2026 release was collected on 13 September 2026 in one supervised session. A script read each pharmacy's own public pages without logging in, no more than one request every two seconds to any site, identifying itself as Mozilla/5.0 (compatible; HealthcountPriceIndex/1.0; +https://www.healthcount.app/glp1-price-index/methodology). It read each site's robots.txt first and skipped any page that file disallows. Where a site refused to serve robots.txt itself, returning an error rather than the file, we followed the robots.txt standard (RFC 9309), which treats that as permission. Where a site refused our reads of its pages, we took no price from it. We do not disguise our reads or use tools to get past a block. Every figure was checked against the page it came from before it went into the dataset.

No price in the September release came from a comparison website, a search result or an archived copy. Search engines were used only to find pharmacies' own pages, and the Internet Archive only to check what Asda's page showed in July, for the July correction. From September 2026, the dataset's tests fail if a price read from September onwards names a source outside the pharmacy's own website.

The October 2026 release was collected on 9 October 2026, again in one supervised session, by a script that applies the same rules in code: the same user agent, robots.txt read first, at least two seconds between requests to any one site, and no price from a page that refused the read. Some pages answer without their prices, because the prices are built by JavaScript in the browser. Those pages were opened in a headless browser that sent the same user agent and followed the same robots.txt and two-second rules for every request the page made, and loaded nothing more from a page that refused it. Where a page shows the price of a dose only once that dose is chosen, the script chose each dose in turn, as a visitor would. Every price that counts in the October release was read that way on 9 October 2026 or, where a site refused us this time, read under the same rules on 13 September 2026; the release says which sites. No price in it came from a comparison website, a search result or an archived copy, and every figure was checked against the page it came from.

Version 1.0 of this page said collection would be automated from the next release, with an archived copy of every page. That has not been built. Until it is, each release says on its own page how it was collected. A pharmacy that would like to be read differently only has to get in touch.

Release schedule and corrections

Releases are planned for the first Tuesday of each month, but the index has not come out every month: there have been releases for July 2026, September 2026 and October 2026, and none for August 2026. The September 2026 release was late: it was due on Tuesday 1 September 2026 and was published on 13 September 2026. The October 2026 release was late too: it was due on Tuesday 6 October 2026 and was published on 9 October 2026. The next release date is printed on every page: currently Tuesday 3 November 2026.

A published release is never edited in place. When a release is published, its figures are stored as a fixed snapshot, and its page and CSV are built only from that snapshot, never recalculated. If a number turns out to be wrong, we publish a dated correction note on that release, add a line to the change log below, and leave the original figure visible, because someone may already have quoted it. This is the single most important promise on this page: if you cite a Healthcount figure, it will still say what it said when you cited it.

Independence and money

The index pages carry no advertising and no affiliate links, and no pharmacy pays for inclusion or position. Healthcount does earn commission from some pharmacies elsewhere on this site, which is set out in our affiliate disclosure. Those relationships have no bearing on what the index measures or reports, and the index deliberately keeps a wall between the two: a page that names a prescription medicine never carries a commercial link.

Licence and reuse

Every release ships a CSV of the observations behind it, licensed CC BY 4.0. Use the figures anywhere, including commercially, with attribution to the Healthcount GLP-1 Price Index and a link to this site. Journalists are welcome to the underlying evidence for any row on request.

Known limitations

• Advertised prices are not transaction prices. Nobody publishes what UK patients actually pay, so no index can measure it.
• The intro-versus-renewal split is currently inferred from how each pharmacy describes its offer, not from a structured field. Where a description is ambiguous, the row is treated as a standard price and flagged.
• The July 2026 baseline carries mixed verification dates and some prices that were not read from the pharmacy's own page. Both are disclosed on the release page.
• Coverage changes from month to month. A pharmacy whose site blocks our reads drops out of the medians after 2 months, and pharmacies add doses, so the pharmacies behind a median are not identical between releases. Each release says who could not be counted.
• Coverage is UK online pharmacies. It excludes the NHS, which does not work on advertised prices, and high-street dispensing that is not published online.

Change log

Questions about the method, or think we have something wrong? That is genuinely welcome: contact us. Corrections get published.