How this site works

The source. The archive is built from a public menu archive on the web. Every shop has a page listing its menus, and every menu entry carries a date. A read-only crawler visits those pages and keeps the prices and the date written on the page. The date is always read from the page text, never from a file name.

From menu to price record. The menu image is read by OCR into a small table: product name, price, format, and the date of the menu. The name is then normalised (lower case, trimmed, plus a short alias list for spelling variants). A price per gram is computed only when the menu prints the weight: price ÷ grams. When the weight is not on the menu, the line stays a unit price and never enters a per-gram comparison.

What we do not publish. We do not republish the menu images, and we do not link to a shop's menu from a page that sells something. The archive is a list of dated prices, nothing more.

The date is the menu date. It is not the date of the crawl, not the date the archive was rebuilt, and not the date the product is on the shelf.

No human check. Nothing is confirmed by a visit or a phone call. A doubtful line stays in the archive with a flag instead of being silently deleted.

One price record, field by field

Two real records from the archive. Every field below is copied from the menu, and nothing is filled in when the menu does not print it.

Per gram against per gram. For weed, hash and extracts the comparison unit is the gram, and it is used only when the menu prints the weight. Joints and edibles are compared per joint and per piece. A unit price never enters a per-gram median.

One price per shop per product. Inside a group the archive keeps the cheapest price each shop printed, so a shop that lists a 1 g and a 5 g line counts once, not twice.

Variants are not mixed. A cali product and a dutch product with the same name are two separate groups. Hash is split by method (sift, ice, import) and by origin. A product is never averaged with its variant.

Packs stay separate. A 5 g pack keeps its own line and is flagged as a pack. The price pages list single grams and packs in two blocks, so a pack price is never read as a per-gram price.

Median, not average. The median published on the price pages is the middle value of the sorted prices; with an even number of prices it is the average of the two middle values. One expensive menu cannot move it, and one cheap menu cannot either.

The median is computed on fresh menus only. Only prices written on menus from the last 30 days enter the median. Older menus keep their own entry, marked as last known, and their difference to the median is shown as indicative rather than as a current discount.

The range is printed next to the median. Lowest and highest price of the same group, so a median is never shown without its context.

A worked example

Fresh entries, per gram, with the median recomputed from the same rows. For the 379 fresh groups that publish a median, recomputing the median from the raw records reproduces the published value in 379 of 379 cases.

The shop behind each best price and the menu date behind it are on the price pages: all prices, cheapest entries, products.

• A price is not today's price. A price is dated to the menu it was written on. The newest menu in the archive is dated 2026-10-28; the snapshot of this page is 2026-11-03. A shop may have changed its prices the day after the menu was printed.
• Availability is not checked. Nobody calls a shop to ask whether the product is on the shelf. A menu line proves that the product was on a menu on that date, nothing more.
• 39 of the 251 tracked shops have no menu in the archive. That is missing data, not a statement about their stock. A further 27 shops have a menu in the directory but no price record in the archive. In total, 185 shops carry price records.
• Genetics and THC exist only where the menu prints them. 710 of 9,542 records (7.4%) carry a THC percentage and 2,133 (22.4%) carry a genetics label. The rest are unknown, not zero. A THC figure here means "printed on that menu", nothing more.
• OCR can be wrong. 1,171 records carry at least one flag. Among them 50 are marked hard to read, 53 have an unusual format and 105 are outside the normal price range. Lines with an unusual format are kept out of the rankings; the others stay visible with the flag attached.
• Names are not canonical. The same product can appear under more than one spelling, and 313 records are packs rather than single grams. A short alias list merges the obvious spelling variants and nothing else.
• The site sells nothing. It is information only, 18+, and it does not arrange purchases, take orders or take payment. See the legal pages and the how to read the data note.

Every price carries the date of the menu it was written on, and the archive runs from 2007-04-17 to 2026-10-28. Nothing in it is extrapolated forward: a price from 2007-04-17 is still a price from 2007-04-17.

The archive keeps the most recent menu it found for each shop, so a year page is not a monthly index of one shop. A year page is the median of the per-gram prices printed on the menus dated in that year, across all shops, and it uses every record of that year, not only the last 30 days.

A shop that has not published a menu for years still shows its old prices, labelled as last known and never mixed into the current median. The older years are therefore built from fewer menus, and the number of records is printed next to each median.

Full year list for weed: /prices/weed/2026/ · all categories: /prices/. The medians above are the ones published on the year pages: the middle value of the sorted per-gram prices of that year, taking the upper middle value when the count is even.

• Prices — median per gram by category, with ranges and menu dates.
• Products — every product name in the archive, with its price history.
• Cheapest — entries below the Amsterdam median, on fresh menus.
• Shops — the 251 tracked shops and the date of their last menu.
• Brands — makers as printed on the menus, never estimated.
• API docs — the read-only JSON endpoints over the same records.
• Legal — terms, privacy, and how to read the data.