Domain terms, with the numbers behind them
Definitions are easy to find and worth little on their own. Each term here carries what this site actually measures for it, drawn from a recorded-sales ledger and a published back-test rather than from industry lore.
Why every figure names its window
The domain aftermarket is two markets wearing one name: a thin high-value trade and an enormous low-value bulk trade. Quote a median over five years and you describe the second; quote it over three and you describe the first. Both are true and they differ by more than an order of magnitude, so no figure below appears without the slice it came from. Cells with fewer than 5 recorded sales are omitted rather than shown as a rate a sample that small cannot carry.
Aftermarket
The resale market for domains that are already registered. A name in the aftermarket is bought from whoever holds it, not from a registry at the standard registration price.
What the record shows: The recorded-sales ledger on this site holds 461,247 sales with a usable date and price, spanning 1999 to 2026, totalling $1.80 billion. That is what has been recorded and screened, not what the market traded: private sales leave no public record, so treat every aftermarket total as a floor.
Comparable sales
Prior sales of similar names, used as evidence for what a name is worth now. Comparables are the closest thing domain valuation has to an appraisal of a house against its street.
What the record shows: The valuation model draws on a screened corpus of 2,994 price-verified sales. How well that pool prices a name it has never seen is measurable: across 218 recorded sales that were never in it (each sale withheld from the engine's evidence), the estimate lands within 2x of the real price 29.4% of the time. Every one of those rows is published.
One-word domain
A name that is a single existing word, with no compounding, prefix or invented spelling. The scarcest class in any extension, because the supply is fixed by the language.
What the record shows: 583 dictionary-word sales were recorded in the three years to 2026-08-05, at a median of $26,550 - the highest median of any well-sampled name class in that window.
Brandable
An invented or evocative name with no dictionary meaning, chosen to be ownable as a mark rather than descriptive of a product. Most venture-funded company names are brandables.
What the record shows: The largest recorded class by volume: 1612 sales in the three years to 2026-08-05, median $4,800. More sales than dictionary words, at a lower median - brandables are where the market is busiest, not where it pays most.
Acronym domain
A short initialism, typically two to four letters. Value tracks how many established organisations share those initials, not what the letters spell.
What the record shows: 273 acronym sales recorded in the three years to 2026-08-05, median $11,582.
gTLD (generic top-level domain)
An extension not tied to a country: .com, .org, .net and the newer set including .ai, .io, .app and .xyz. A ccTLD is the country-coded counterpart, such as .de or .co, though several are marketed far outside their country.
What the record shows: Extension is not a multiplier you can apply to a name. In the three years to 2026-08-05, .com recorded 1616 sales at a median of $5,000, against .ai 286 at $59,694, .io 53 at $37,052, .co 50 at $22,406. The extensions differ in what gets sold through them as much as in what a name is worth in each.
Liquidity
How readily a name can be turned into cash at something near its estimated value. A domain can be valuable and illiquid at the same time, which is why an appraisal is not a price you can obtain on demand.
What the record shows: Recorded sales cluster into few venues: unknown accounts for 45% of the 3,608 sales in the three years to 2026-08-05. Concentration like that is why a name with no bidder at one venue can look unsellable while being perfectly ordinary.
Appraisal
An opinion of value, not an offer and not a price. An appraisal is only as good as the evidence behind it and the error it is willing to publish.
What the record shows: Appraisals here are deterministic - the same name always returns the same figure - and each carries a low-to-high interval rather than a point alone. Measured against 218 sales with each sale withheld from the engine's evidence, the median error is 85.1% and 29.4% of estimates land within 2x of the recorded price. The full error table is published, including the revision that did worse.
Where these numbers come from
Market figures are aggregates over the recorded-sales ledger, whose coverage runs to 2026-08-05; accuracy figures come from the published back-test. Both are committed artifacts regenerated by tooling, never typed by hand, so what this page says cannot drift from what the accuracy page and the Sales Research workspace report. Recorded sales are a floor on market activity, not a census: private transactions leave no public record.