What makes a sold comp actually comparable

Most comp sets are assembled on category alone: it is a SaaS business, these are SaaS businesses, therefore these price it. Category is necessary and nowhere near sufficient. A comp earns the name when it matches on four things at once:

• Asset type. The easy one, and the only one most tools check. A content site and a Shopify store are priced by different buyers for different reasons.
• Size. A set of $400 sales does not price a $40,000 business, however many of them you stack up. This is the match that gets skipped, and section 02 is about it.
• Basis. A multiple is a fraction, and both halves have to agree: profit or revenue, and annual or monthly. Two of our guides exist because these get crossed — revenue multiple vs profit multiple and monthly vs annual multiples.
• What the price was. An asking price is not a sale (why the gap exists), and a headline sale price is often not a payment either — cash at close, deferred instalments and contingent earn-outs all hide inside one reported number (taking a headline apart).

Fail any one of the four and the arithmetic still runs. That is the problem: a mismatched comp set does not error, it just produces a confident number about a different business.

Our own estimator draws the line at a factor of four. When the business being valued states an annual figure four times larger than the median comp in the sample — or a quarter of it — the subject and the sample are not the same kind of business, and the comparison is flagged rather than quietly used. Four is a wide tolerance chosen deliberately: it is loose enough that ordinary variation inside a category does not trip it, so when it does trip, something real is wrong.

What happens then is a downgrade, not a refusal. The calculator tells you the comps are much smaller than your business and says to read the range as a floor. The paid report says the same and shifts weight onto public-web comparables of the right size. Neither pretends the number is not there; both stop presenting it as an answer.

Be aware of an asymmetry in what we currently surface. The flag is computed in both directions, but the warning you will see is the one for comps smaller than your business — the direction that makes our range too low. The opposite case, where your business is a quarter of the size of the sample, is the direction that would make a range too high, and today you have to catch it yourself by reading the median stated figure in the table below against your own. We would rather say that here than have you assume silence means a match.

One row per asset type we can price at all. Where a type supports both bases we show profit, because that is the basis the estimator prefers when your business states a profit. The "median stated annual figure" is the middle comp's own profit (or revenue) as its listing stated it — not audited, and section 04 is about what that does to this table. These figures are measured over the paid report's window of the most recent 200 priced sales of that type, which is a narrower population than the all-time medians on what small online businesses actually sell for; compare within a table, not across the two. A type with fewer than five usable comps on a basis publishes nothing here.

The size test compares stated figure against stated figure, and both sides of that comparison are claims made by sellers. When a whole sample's claims are inflated in the same direction, the test passes on a sample it should have failed.

The clearest instance is live in the table above. Look at the row whose median comp states five figures of annual profit and sold for a few hundred dollars. Measured on stated profit alone, that sample looks like a peer of a genuine five-figure-profit business, and the four-times rule says "similar" and stays silent. It is wrong: those are not five-figure businesses, they are listings whose profit claim the sale price refused to bear out.

What catches it is the other column. A sample whose median sale clears less than one year of its own stated profit is telling you the claims are not corroborated, whatever the size test says — which is why every surface that shows you a range shows that flag too, and why what a profit multiple actually means spends most of its length on medians below 1×. Two checks, because the first one has a blind spot the second one covers.