What is your domain actually worth?

How the appraisal works

Anchors retrieved July 2026 · Last reviewed: July 2026

The shape of the model

The engine is deterministic and runs entirely in your browser: the domain is parsed, segmented into dictionary words against Peter Norvig’s word-frequency corpus (top 30,000 words of the Google Web Trillion Word Corpus), assigned to a pattern class — single word, two words, brandable coinage, acronym, numeric, or long-tail — and then a fixed, ordered chain of factors adjusts that class’s base range. The running subtotal is shown after every step, so the line items sum to the final range exactly. Same input, same output, every time. Nothing you type is sent or stored.

Classification happens first because honest base ranges differ by orders of magnitude between classes — a length multiplier means nothing across classes. The class boundaries themselves are labeled judgment calls.

Where the dollars come from

Every dollar figure in the engine traces to one of two measured cells from NameBio’s public sales database, read July 2026 over a past-three-years window, sorted by price with quartiles read at their rank positions. The base range is the interquartile spread — the middle half of reported sales, outliers excluded on both ends:

Classes without a directly measurable cell — brandable, acronym, numeric, long-tail — get a base range that is a stated fraction of the measured two-word cell, labeled as a judgment transform on the record itself. No dollar constant in the engine was invented in a spreadsheet or a chat session.

Known bias, worth stating plainly: NameBio reports sales above venue minimums (roughly $100+), so these anchors describe successful, reported sales — not the far larger population of domains that never sell at all. That is why every record carries the sell-through note, and why a range topping out under $100 becomes a floor verdict rather than a number.

TLD multipliers

Measured the same way, holding name quality fixed: the median one-word dictionary sale on each TLD divided by the same cell’s .com median. .com carries roughly 72–74% of reported aftermarket dollar volume (NamePros analysis of NameBio data).

Any TLD outside this set gets a wide judgment band (×0.1–×0.5) and a confidence grade of C — an honest “we didn’t measure this” instead of a fake precise multiplier.

The factor chain

Factors apply in a fixed order, each tagged on the record: DATA means the value is anchored to a cited dataset; JUDGMENT means we chose the magnitude and say so.

1. Base — the class’s anchor range, shown as its own line, never hidden.
2. Length — shorter is worth more; GoDaddy’s published valuation research confirms the direction. The step sizes are judgment calls, because NameBio’s free tier can’t slice sales by length.
3. Word frequency — where the domain’s words rank in the Norvig corpus (top 1k / 10k / 30k). The band is data; the multiplier magnitude is a labeled judgment, and we state the gap: frequency is not commercial intent — “insurance” outsells “the.” Pure-stopword domains are floored.
4. TLD — the measured table above.
5. Pattern penalties — hyphens and digits mixed into word domains take steep, stated discounts. Direction is market consensus; magnitudes are judgment calls.
6. Pronounceability — brandables and acronyms only: a deterministic consonant–vowel heuristic mapped to ×0.9–×1.2. There is no citable formula for “brandable,” so this one is all judgment, and says so.

Wholesale, retail, and the two honest non-answers

The primary output is a wholesale range. The retail range applies the widely published ×2–3 dealer-to-end-user guideline (DNAcademy, DomainDetails); documented end-user outcomes of 5–20× are real, which is one more reason the output is a range.

Two verdicts are deliberately part of the model. The floor verdict — likely no meaningful resale value — fires whenever the appraised range tops out under NameBio’s own ~$100 reporting floor, and it is the true answer for most registered domains: typical portfolios sell 1–2% of names per year (NamePros, Namecheap). The refusal covers ultra-short domains (1–3 character .coms), which trade as liquid assets at market prices a factor model cannot capture — we point at market data instead of inventing a range.

Comparable sales come from DNJournal’s weekly sales reports and are displayed evidence with source links — never inputs to the computed range. Folding them in silently would rebuild the black box this site exists to oppose. DNJournal charts the top of the market (venue floors around $2k), so comps show what strong sales of a shape look like; the NameBio anchors carry the central tendency.

What we deliberately don’t model

• Search volume and CPC — needs paid APIs; a v1.1 candidate.
• Traffic and backlinks — needs crawler data we don’t have.
• Domain age — checkable via free RDAP later; GoDaddy’s research finds it weak anyway.
• Trademark risk — no reliable free dataset exists, so every record carries the blanket warning instead: a domain containing someone else’s mark can be worth less than zero.
• The buyer — the single biggest real driver of price is the specific buyer who needs this specific name, and no model can see them. That is why the output is a range.

The longer argument — why automated appraisals disagree by multiples, what reported sales actually show, and how to sanity-check any appraisal including this one — is in our guide, what a domain is actually worth.