Where sold-price data comes from — and what never makes it in

There is no filing requirement for selling a website. A private buyer and a private seller sign an asset purchase agreement, money moves, and the price is known to four people. Nothing about that transaction is wrong, and nothing about it will ever reach a database.

Three habits of the market then decide what does get out:

• Auctions publish because the mechanism is public. A closing bid is visible by design, so auction marketplaces are the one venue where sale prices are routinely stated on the record.
• Brokers publish asks, not sales. A brokerage's public pages exist to market current inventory. Once a deal closes the interesting number — what was actually paid — is the number with the least commercial reason to be posted.
• Announcements are voluntary and self-selecting. Founders post about exits they are pleased with. A quiet sale at a disappointing price does not get a thread.

So the published record over-represents small auctioned assets and happy announcements, and under-represents brokered mid-market deals and every private sale. That is the bias you inherit the moment you price anything off comps — ours included.

The rows above marked yes or partly are the lanes that actually produce citable comps; the live counter in section 04 reports how many of them are cited in the database as of right now, which is the figure to trust if it disagrees with this table. Two further lanes are not outside sources: deals sellers and buyers report to us directly, which are marked self-reported and unverified, and an archive lane that reads old snapshots of marketplace pages. Neither changes the picture below.

It matters which kind of "no" a source gives, because only one of the three might change.

Refusal

A site's robots.txt or terms can forbid collection of the pages that hold the data. Motion Invest disallows the marketplace paths its sold table lives on, so we do not read it — not because the data is unusable but because we were told not to. That is recorded and honored, and it stays honored if it makes our coverage worse.

Absence

Some sources are wide open and simply do not state the number. Empire Flippers publishes a sold-listings feed — 2,665 records when we last counted it, on 2026-08-20 — in which the money field is the listing price, i.e. the ask. Ingesting that as a sale price would be recording an asking price as market data, which is the one thing this dataset exists not to do. Their own aggregate reporting of how close sold prices land to list prices is itself an acknowledgement that the two differ. So a brokerage with thousands of completed sales contributes exactly zero comps, and there is nothing to fix at our end. Acquire.com is the same story with a login in place of a missing field.

Ambiguity

The hardest case is a page that states a number without committing to what it is. Microns' homepage cards label a figure "Exit Price"; the listing page each card links to labels the same figure "Asking Price" beside an already-sold marker. Two labels, one number, one of them an ask — so we take the price from the surfaces that state it unambiguously and exclude the cards. Acquisition journalism fails the same test far more often than it passes it: on 2026-08-21, 6 of 23 recent deal articles on TheyGotAcquired stated a definite attributed price, while the rest gave "not disclosed", a magnitude ("high 7 figures"), a range, or a hedged report of someone else's figure. None of those four is a comp.

Loading live source and category counts from /api/stats…

A source URL is a claim that a page states a number. Pages change. Two things have already happened to ours:

• A citation can stop resolving. When Flippa's API closed on 2026-08-28, 48 of our rows were left citing an endpoint that now answers 401 to an anonymous reader. Their prices were verified against the listing page at ingest and each row carries that page as secondary evidence, so the fact stays checkable — but the primary link no longer opens, and we say so rather than letting it look fine.
• A page can retract its own number. One Flippa listing that stated USD $101 when we read it on 2026-08-20 now reads "Contact Flippa for Price". The sale happened; the evidence for it was withdrawn afterwards.

This is the strongest practical argument for keeping the source link on every single row instead of publishing only aggregates. An aggregate cannot rot visibly. A link can, and when it does you are entitled to know which number is now resting on a page that no longer says it.