"Comps" is short for comparables, a term domainers borrowed from real estate. The idea is the same. You cannot know what your exact name will sell for until it sells, but you can look at what close equivalents actually fetched and reason from there. Our page on comparable sales covers the general method. This page is about doing it well for domain names.
What makes a domain comp good
A comp is useful when the differences between it and your name are small, or at least understood. In practice that means checking the features that move domain prices, in roughly this order.
| Feature | Why it matters |
|---|---|
| Extension (TLD) | Different extensions have different buyers and price levels. A .com word and the same word in .ai are separate markets. |
| Length | Shorter names are scarcer. One or two characters of difference can change the price band. |
| Style | Exact match keyword, brandable, two-word, acronym and numeric names are priced differently. |
| Term class | A developer term, an AI term, a dictionary word and an invented brandable attract different buyers. |
| Meaning | A name that serves the same kind of business is a better comp than one that only shares letters. |
| Singular or plural | Usually the singular is stronger, with exceptions. |
| Digits and hyphens | Both usually lower the price for word names, but numeric names are their own market. |
| Venue tier | An expired auction price and an end-user marketplace price are different kinds of evidence. |
| Date | Fast moving segments, such as .ai, make older sales weaker evidence. |
You will rarely find a comp that matches on every line. The skill is in knowing which mismatches are minor and which change everything.
Extension first, almost always
Mixing extensions is the quickest way to get a wrong number. Of our 128,867 recorded sales, 100,357 are .com, 4,925 are .ai and 1,431 are .io. A .com sale of the same word is context for a .ai name, not a comp. Start inside the same extension and only widen if you must.
Singular versus plural is not a rule
Plural names usually sell for less than their singular form, but the data has exceptions. rebate.com and rebates.com each sold for 500,000 USD (Sedo, 2022-12-01), and bagels.com sold for 500,000 USD (Sedo, 2024-03-18). When a plural is the natural way a business describes itself, the discount can disappear. Treat plural as a feature to weigh, not a fixed penalty.
Meaning beats spelling
Keyword search finds names that contain your word. It misses names that serve the same buyer with different words. If you own a name about scheduling jobs, sales of task, workflow and job names are relevant even though none contain your exact term. For example, workflow.ai sold for 35,000 USD (Park.io, 2023-01-07) and kernel.ai for 14,599 USD (Dynadot, 2024-10-08). Both are developer terms in .ai, and both are better evidence for a technical .ai name than a random .ai sale that happens to share four letters.
Wholesale comps and retail comps
Decide first what you are estimating. An investor asking what to bid at an expired auction needs wholesale comps. A seller setting a buy now price for an end user needs retail comps. Using one to answer the other produces a wrong number in a predictable direction.
| Venue tier | Recorded sales | Median | Average |
|---|---|---|---|
| Expired auction / investor venue | 30,021 | 1,725 USD | 3,312 USD |
| End-user marketplace | 23,816 | 2,288 USD | 6,437 USD |
The same name can sit on both sides. zip.ai sold for 1,610 USD (Park.io, 2021-08-10) and later for 200,000 USD (Afternic, 2025-03-27). Both are valid comps, for different questions. The domain sales history page has more repeat sales like this.
How many comps you need
One comp is an anecdote. Three is the minimum for a range you can defend. Five to ten close comps, with the closest ones weighted most, is a solid basis. More than that only helps if the extra comps are genuinely close. Padding a list with weak comps makes the range look tighter than the evidence supports.
If you cannot find three good comps, the honest conclusion is that the evidence is thin. Widen the range and say so. And remember that every recorded sale here is at least 1,000 USD. If no close comp exists above that floor, names like yours usually trade below it.
Common mistakes with domain comps
- Anchoring on the biggest sale. icon.com sold for 12,000,000 USD (Hilco Digital Assets, 2025-04-24). That is not a comp for any ordinary four letter word.
- Ignoring venue. Comparing your asking price to retail comps, then accepting a wholesale bid, or the reverse.
- Using stale data in a moving market. The .ai extension recorded 152 sales in 2021 and 2,070 in 2025.
- Counting letter overlap as similarity. Shared letters are not shared buyers.
- Hiding the comps. A number without its comps cannot be checked or defended.
How relevance scoring works in SoldComparables
SoldComparables turns the checklist above into a score. Every sold domain name in the corpus is split into words, and an embedding language model has precomputed which words are close in meaning, so "cron" sits near "scheduler", "task" and "job". When you search a name, each candidate comp is scored from 0 to 100 on extension, length, style, term class, singular or plural, digits, hyphens, venue tier and meaning.
Each comp shows up to three plain reasons for its score, for example "Same .ai extension", "Shares the word cron" or "Same length", along with its price, venue tier, date and a link to the public report. You see why a comp was chosen and can disagree with it.
The estimate built from those comps gives low, likely and high values, separate wholesale and retail figures, a confidence of High, Medium or Low, and a rationale that cites comps by name, price and date, plus what would raise or lower the number. With fewer than three good comps the report says so and widens the range. It is a data-backed estimate, not a certified appraisal.
To see your own domain comps, run a search on the domain appraisal tool, or compare plans and limits on the pricing page.