A comparables dataset is only as good as the kind of price it holds. Two marketplaces in this cluster publish two different kinds, and both get called “the price” in a spreadsheet column.
Three different numbers called price
An asking price is what a seller requested. A sale price is what a buyer and seller agreed after negotiation. A price realized is what an auction achieved when the hammer fell. The first is an expectation, the second and third are outcomes, and the gap between expectation and outcome is usually the thing a valuation is trying to find.
Most public deal data holds exactly one of the three. Knowing which one is the first question to ask of any dataset, and the answer is rarely in the column name.
What Empire Flippers records
Empire Flippers is a brokered marketplace for online businesses. Its public listing data covers three statuses, and on 23 September 2026 the catalogue held 2,877 listings: 180 For Sale, 23 Pending Sold and 2,674 Sold. The Sold listings are what makes it useful for valuation, because each carries its price, its multiple and twelve months of P&L.
But the price on a Sold listing is the final asking price, as the Empire Flippers Scraper documents it. Nothing in the output is labelled as the price the business changed hands at. So a set of Empire Flippers comparables answers “what multiple did sellers list at, for businesses that did sell?” — a useful question, and not the same as “what multiple did buyers pay?”
The multiple itself is internally consistent: on 2,877 of 2,877 listings the monthly multiple equalled the asking price divided by average monthly net profit, to within one. That makes the multiple a derived figure you can recompute rather than an independent one. The walkthrough is in how to scrape Empire Flippers.
What the Sold history does well is breadth. 2,674 closed listings span 21 monetization models and 59 niches, from Display Advertising and Amazon FBA to SaaS and newsletters, each with the profit, revenue and expense series the listing was priced on. Grouped by model, those rows show how the market priced a content site against an Amazon business at the moment each was listed — which is the right question for a seller setting an asking price, and only a proxy for a buyer estimating what to offer.
What Invaluable records
Invaluable aggregates auction houses. Each sold lot carries a price realized, and next to it the auctioneer's own low and high estimate. Over a stratified sample of 1,020 sold lots measured on 1 October 2026, the price realized was filled on 100% and the estimate on 91.7%. That pairing is what makes estimate-versus-result analysis possible: you can see, per category or per house, how often lots beat the high estimate.
The Invaluable Auctions Scraper describes price_realized as the hammer price. Whether a given house's published result includes buyer's premium was not verified for this page; the optional lot page delivers the premium tiers so the all-in figure can be computed explicitly rather than assumed. More on the data and its terms in Invaluable auction results.
An auction result is also a narrower kind of outcome than it looks. A lot that failed to sell has no price realized at all, and by default only lots with a published result are returned, so a set of sold comparables leaves out the objects the market declined at their reserve. That bias runs in the opposite direction to the asking-price bias: one dataset overstates what sellers got, the other is silent about what did not sell.
Side by side
| Question | Empire Flippers | Invaluable |
|---|---|---|
| Kind of price | Asking price, final for Sold | Price realized, with estimate |
| Price filled | 99.9% of 2,877 listings | 100% of 1,020 sold lots |
| Currency | USD | The lot's own, never converted |
| Missing value stored as | null (2 unpriced listings) | 0 at the source, null in the output |
| Sold history size | 2,674 listings | Millions; the counter is approximate |
The two are complementary rather than competing. For a business, an asking multiple from a comparable Sold listing is the best public anchor available; for a painting, a watch or a coin, the price realized against its estimate is a measured outcome. What neither supports is a statement about the other's market.
Four traps in a comparables dataset
Reading an asking price as an outcome. An Empire Flippers Sold multiple is where the negotiation started. If you need what buyers paid, this data does not contain it, and no amount of cleaning will put it there.
Averaging zeros. Invaluable stores “no estimate” and “no result” as 0. Passed through, those zeros drag every mean down with no visible symptom. The Actor delivers them as null; a dataset from elsewhere may not.
Mixing currencies. Invaluable lots arrive in USD, EUR, GBP, AUD, CAD, JPY and more, and a price sort compares the raw numbers. The failure is the one described in Mercado Libre's mixed-currency pages: every row is plausible and every aggregate is wrong.
Trusting a coverage figure from a different sample. Invaluable estimate coverage was 91.7%, 85.2% or 70.8% depending on which lots were sampled. A fill rate describes a sample, which is the argument of measured fill rates, not promises.
The two sites' terms are not alike
Empire Flippers' Terms of Use, effective 8 November 2025, contain no clause on scraping, crawling, robots or automated access; its obligations are addressed to registered users, and the Actor registers no account. Its robots.txt disallows only /wp-admin/.
Invaluable is the opposite case. Its Terms of Use, version 3.11, prohibit robots, spiders and “any form of data extraction or data-mining whatsoever” without written permission, and its robots.txt opens with the same prohibition. Invaluable also sells subscriptions to its database of past auction results. Those facts are on the Actor page because a user deciding whether to run it needs them.