Reading comparable sales without fooling yourself

Comparable sales are the closest thing the domain market has to evidence. They're also the easiest thing to abuse, because the data is patchy, self-selected and often stale, and because almost any name can be made to look underpriced if you're allowed to choose the three comps yourself. Finding comps is trivial. Deciding which ones you're permitted to count is the actual work.

What follows is how to do that without quietly assembling a case for the figure you'd already settled on.

One sale is a story, not a market rate

A recorded sale proves exactly this: on one day, one buyer paid that amount for that exact string. Nothing more. It doesn't prove the name was worth it, that a second buyer would have matched it, or that the price would repeat next spring.

Domain demand is thin, and that's the crux. For most names the pool of people who'd pay a serious price is small. Sometimes a handful. Sometimes one. A sale at the top of that pool records what the most motivated buyer paid, not where the market clears. In a liquid market, price is a consensus. In a thin one, price is a coincidence between two people who happened to find each other.

So every comp is a single data point with a wide error bar. Three comps clustered tightly is weak evidence. Fifteen clustered tightly starts to mean something. One big number is anecdote wearing a suit.

Tighten what "comparable" is allowed to mean

Most inflated valuations trace back to a generous definition of the word. A comp earns its place only if it matches on most of the following:

• Extension. A .com sale does not price a .io, .co or .ai name, and the reverse holds. Cross-extension comparison is the single most common route to a number the market won't honour.
• Word count and structure. One-word dictionary names, two-word compounds, invented single words and hyphenated names all behave differently. A two-word compound is not evidence for a one-worder.
• Commercial category. A name reading as fintech, insurance or health sits in front of deeper pockets than one reading as a hobby. Category often outweighs the string itself.
• Recency. Anything more than three or four years old describes a different market. Context, not arithmetic.
• Sale type. An open marketplace listing, a broker-negotiated deal, an expired-domain auction and a corporate acquisition following a trademark dispute are four distinct markets. Auction results in particular reflect what resellers will pay, which is a discount to end-user pricing.

Fail two or more of those tests and the comp becomes background reading. Keep it out of the average.

Every data source leans in a known direction

Which is useful, because a known bias can be corrected for.

The public record is the top slice

Sales databases are built from what people choose to report, and large numbers get reported far more reliably than small ones. Failed listings, price cuts, quiet sales at a fraction of ask: mostly invisible. Average the public comps and you are averaging the winners.

The denominator you can't see

For every name that sold well, similar names sat unsold for years. They don't appear in any database, so you never see the hit rate. A comp gives you a price. It says nothing about the probability of achieving that price, and probability is usually the number that decides whether you should be pleased.

Search long enough and the market agrees with you

If you keep looking until the comps support your figure, you'll find them. The market is big enough that supporting evidence exists for nearly any valuation. The fix is procedural rather than moral: set your comp criteria before you look, collect everything that fits, and keep the low results in even when they're irritating.

Similar strings, unrelated values

Two names can look nearly identical and be worth wildly different amounts. Singular versus plural. A common misspelling. One extra syllable. A word that means something unfortunate in another language. Any of those can shift a name from "a business would build on this" to "nobody wants it". Read the comp the way a business would, not the way a pattern matcher would.

Produce a range and admit how wide it is

Single numbers imply a precision nobody has. A method that works:

• Gather every comp that passes your criteria, ideally ten or more. If you found three, say so and widen the range to match.
• Discard the top and bottom outliers, or quarantine them and record why they were extreme.
• Take the middle of what's left. That's your central estimate for a name of that class.
• Adjust the specific name against the class: length, pronounceability, whether it needs spelling out loud, whether an obvious buyer category exists at all.
• Adjust for sale type. Broker-led end-user comps and an open marketplace listing means you land lower. Auction comps and an end-user buyer means you have headroom above them.

Then apply the adjustment almost every valuation skips: time. A price you might get from the right buyer within five years is not the same asset as a price you can bank this quarter. Need a fast sale and the honest figure sits meaningfully below the comp range. Stacking more comps doesn't move it.

Buyers make the mirror-image mistake

They use comps to argue a seller down on a name they've already decided they want. It rarely lands, because the seller's cost of waiting is usually lower than the buyer's. If the domain isn't their business, they can sit on it more or less forever.

Better question: not "what did similar names sell for" but "what is this name worth to me, and what's my next-best option". Price the alternative in concrete terms. If a strong available .com surfaced by something like The Name Studio would serve the business nearly as well at registration cost, that's your genuine ceiling regardless of what the comps say. If nothing else fits, because the name is the category term or the brand is already in market, comps become close to irrelevant and you're simply negotiating.

Use them to check you're not paying an order of magnitude above the class. Don't use them to argue about twenty per cent. Sellers read the same databases, and the gap between two motivated parties tends to close on terms rather than on data: instalments, a lease with a purchase option, how escrow is structured.

What to carry away

Comps are a floor for your reasoning, not a replacement for it. Define comparability before you search. Assume the public record over-represents good outcomes, because it does. Give a range rather than a number, and widen it below ten data points. Above all, keep price and probability separate. The expensive error in domain valuation isn't misreading a comp. It's forgetting that most names never meet the buyer who would have paid it.