How every price here is made

A sale counts only when all of these are true:

• It sold. Lots that went unsold are never priced. On eBay only listings that sold count — never live listings or asking prices.
• We know what it is. Each sale is matched to one entry in the catalogue (and to a version where the listing says which). Every match passes an automatic check, and people check samples by hand. A sale we can’t place confidently is left out rather than guessed.
• It wasn’t rejected at review. A sale marked as an error at review is taken out of every figure, list and data copy on the site. About 12% of the eBay listings we collect are rejected this way.
• It belongs to a live catalogue entry. When two entries are merged, their sales move with them to the one that stays.

That is what “published” means everywhere on the site: on a page, in its data copy, in search and in the monthly report.

Auction houses don’t all publish the same number. Most publish the hammer price — what the lot sold for in the room. Some publish the price with the buyer’s premium added (about 29% of our records). For those we take out that house’s premium at the rate it charged on the date of the sale, so every auction figure is on the same hammer basis.

eBay prices are what the buyer paid for the item, as sold. Postage is never included.

If a house doesn’t say which number it publishes, its sales are listed as published but kept out of every figure. Each sale row on a page shows its own basis: “hammer”, “premium removed” or “as sold”.

Each sale keeps its original price and currency — a sale in Vienna in euros, one in Chicago in dollars — and that is what its row shows. 39% of our records are not in pounds.

For the figures, each sale is converted to pounds at the European Central Bank’s reference rate for the month of the sale (if a month has no rate, the most recent rate up to three months earlier; sales before 1999, when the series starts, use its first month). Dollar and euro figures are worked out the same way, sale by sale, and each is the median of its own conversions — never the pound figure multiplied by today’s rate. A sale with no date has no dollar or euro figure.

A sale we can’t convert is never priced in its own currency as if it were pounds: it is listed apart as “no convertible price”.

The typical price is made only from sales of one item. A camera with its own lens, or in its box, is still one item; its row is marked “+ lens” or “boxed”.

Everything else is listed under Not in the typical price, each group with its count and range (and its own median once it has three sales), and never blended in:

• Outfit (body with several lenses)
• Display, dummy or cut-away piece
• Lot or parts
• Awaiting classification
• No convertible price
• Other (not a sale of one item)

New sales wait a short while to be classified; until then they show as “awaiting classification”, not as a mistake. Display, dummy and cut-away pieces are priced on their own. A working prototype is a camera and stays in.

Many cameras and lenses came in versions — a finish, a marking, a mount. Where the listing says which version was sold, the sale is filed to it. Every sale on a product with versions is in one of three states, shown as a bar under the price:

Matched to a version

The listing names it, and the sale is filed to that version.

Checked — the listing doesn’t say

We looked; the title and description don’t tell us.

Not yet checked

Waiting for the matching job. This part shrinks as it runs.

Price by version lists each version with at least three sales: its typical price and middle half on one shared scale. Versions with one or two sales are listed as “rarer versions” with their sales. A version’s own page uses only the sales filed to it.

The typical price is the median: half the sales were for more, half for less. With an even number of sales it is halfway between the middle two. Figures are rounded to the whole pound (or dollar, or euro).

The middle half is the range from the quarter mark to the three-quarter mark: half of all sales fall inside it. We show it instead of the lowest and highest sale, because one odd sale can stretch those without telling you anything.

Nothing is trimmed. Unusually low or high sales stay in and are flagged (M9). The median barely moves for one odd sale, so it doesn’t need them removed.

Auction and eBay together. The figure pools salerooms and eBay. When the auction median and the eBay median differ by more than 30% (each with at least three sales), both are shown under the figure.

Time. The figure covers every recorded sale, and the dates it covers are printed beside it. On Fair and Solid prices, a “last 3 years” figure is added when those three years hold at least ten sales. We don’t draw trend lines or print “up 12%”; the chart shows every sale over time instead.

One record. Each figure is worked out once into a single stored record. The page, its data copy, search results, brand lists and the citation line all read that record, so they can’t disagree.

Condition is not adjusted for. A mint boxed example and a worn one sit in the same set. That is the biggest single reason for spread, and why the middle half matters.

Every price carries one plain word saying how much evidence is behind it:

Small tags after the grade add context:

Wide spread

At least ten sales, and the top of the middle half is more than three times the bottom. A sentence then says why (M8).

One source

Every sale (three or more) came from one venue.

Old

The newest sale is more than three years old.

Version mix

The product has versions, and fewer than half its sales are matched to one.

On an Indicative price we also show how much one sale moves it: the median without the highest sale, without the lowest, and without any flagged sales.

When a price has a wide spread, the card says why in one sentence. The sentence comes from fixed rules, not from a writer or a language model. The rules are tried in this order, and the first that fits is used:

1. Period. Sales before and after some year differ by more than double, and each side holds at least a fifth of the dated sales. The year with the biggest difference wins. The sentence names both periods and their typical prices.
2. Version. A version with at least five sales has a median more than three times that of the sales where the version isn’t stated (or, when those are fewer than five, the standard version). The sentence names it.
3. Saleroom. Within the last ten years, one venue holds at least a quarter of the sales, both it and the rest have at least five, and their medians differ by more than double. The sentence names the venue.

Period comes first so that an old era is never mistaken for a cheap saleroom. If no rule fits, the sentence is “Prices vary widely; see the versions below.”

Flags help you read single sales. They never change a figure.

◆ far below the rest

More than four times below the median of the other sales of the same version (or of the “version not stated” group) when that group has at least five other sales; otherwise of the product’s other typical sales, when there are at least five.

◆ far above the rest

The same test upwards — but not on prices with a wide spread, where high sales are usually real, rarer versions that the sentence in M8 already explains.

◆ title doesn’t name the model

A text check: the listing title lacks the word that tells this model from its relatives (a Contessa listing that doesn’t say “LK”). It is a prompt to look, not a verdict.

+ lens · boxed · checked by a reader

What the listing says came with it, and sales a reader has checked in the past.

If a sale is wrong, we want to know: a misplaced lot distorts a price for as long as it stays. Every product page has a Report a wrong sale form at its foot — pick the sale and say what is wrong. Or email info@cameraworth.com with the page and, if you can, the lot or listing.

We check each report against the original listing. A wrong sale is removed or moved to the right entry — not left in with a note — and the figure is recomputed.

Reusing the figures

Each product and version page has a data copy (data.json, linked at the foot of the page) holding exactly what the page shows. You are welcome to quote and reuse the figures with attribution. The data copy carries this line:

Data from cameraworth.com — a public register of what cameras and lenses actually sold for. Cite the page URL and the retrieval date; see /methodology/ for how each figure is made.

We don’t publish the whole sales ledger as a download. If you need a cut of the data we don’t publish, ask: info@cameraworth.com.

To cite a product, use the “Cite this page” sentence on it: it carries the figure, middle half, number of sales, dates, grade, method v1, the date you read it and the address. To cite this page: cameraworth.com, “How every price here is made” (method v1), retrieved 8 October 2026. https://www.cameraworth.com/methodology/

Accuracy figure: not yet published. Each month we will check a random sample of published sales by hand against the original listings and print the share that was right, here and only here, with the sample size and date.

Matching a listing to the right camera is the hard part, and it is never perfect. That is why every sale is shown with a link to its original listing, why flags point out the odd ones, and why the median is used: one wrong sale moves it very little.

The trust test. Before a change to these rules goes live, a person who knows the market judges the new figures for a fixed list of thirty well-known cameras against the old ones. A change that makes those figures worse doesn’t ship. The list stays the same from one test to the next, so the results can be compared.