Model history
Domain Value Score model revision history
Every score is stamped with the model version that produced it — shown on each report as “Model dv-X”. This page records what each version changed. Scores keep the version they were computed under and are never re-scored, so older results stay comparable to when they were run. The full method behind the current version is on the Methodology page.
Model dv-5.8.0 — 2026-08-31
The English word list was rebuilt, and about fourteen thousand everyday words came back. The previous release fixed names like smoothie by hand, but the reason they were missing was still there: the word list behind the scan was filtered through a dictionary published in 1934, and that dictionary lists only base forms. It has “labor” but not “laborers”, “allocate” but not “allocating”, “whisper” but not “whispers”. Ordinary plurals and verb endings simply fell out, and a name built from one of them was read as fragments rather than as the word it plainly is. The list now also accepts a word whose ordinary ending traces back to a base form it already knows — plurals and the -ing, -ed, -er, -est and -ly endings. That is deliberately a rule about the shape of English rather than a lower bar for how common a word has to be, because the words being excluded are overwhelmingly names of people and places, foreign words, programmer jargon and website names, none of which trace back to an English base form and none of which are let in. The build now also folds in the hand-curated words from the previous release, so they can never be lost the next time it runs, and checks its own result against a list of words that must be present and junk that must stay out, refusing to write anything if either slips. Coined and meaningless names are unaffected: across eight thousand invented names the reading is statistically unchanged. Nine names already known to the scan now read correctly — “coined” had been read as coin + ed, and “curated” had been read as Spanish. Names built from the recovered words can score higher; nothing measured scores lower.
Model dv-5.7.0 — 2026-08-31
Ordinary modern words are read as words. The word list behind the scan was built from a web corpus filtered through a dictionary that predates the war, so a whole band of everyday English was simply missing from it — smoothie, ramen, matcha, gelato, granola, barista, hummus, latte, leggings, selfie, esports, fintech. That was worse than a blind spot, because a name has to be read somehow: with the real word unavailable, the reading was assembled out of whatever smaller pieces happened to fit, and the result was presented as if it were the name’s meaning. smoothie.com was read as “smooth”, ramen as “ram + en”, matcha as “match + a”, fintech as “fin + tech”. About a hundred curated words now close that gap, across food and drink, clothing and personal care, health, online work and media, and the settled industry blends such as fintech and proptech. Genuine compounds are deliberately left alone, because their parts are the right reading and carry the name’s commercial meaning: haircare is still hair + care, carwash still car + wash, blockchain still block + chain. Brand names are excluded on principle — a trademark is handled by the trademark checks, not valued as a word. The words that point to a clear market now also count toward the industry a name belongs to, which they could not before, since the word was never recognized in the first place. Names built from these words can score higher, and two names score lower because their old score rested on a reading that was never true — macaron was being valued as “car” and sorbet as “bet”. Every other name scores exactly as before.
Model dv-5.6.0 — 2026-08-20
Acronym coverage grows about five-fold. The first release of acronym recognition (dv-5.5.0) shipped a deliberately small set of about sixty; this release grows the curated table to about 290 — technology (URL, DNS, VPS, CPU, JSON, HTML), security (SSL, MFA, GDPR, HIPAA), business and finance (CPA, APR, REIT, HELOC, ESG, BTC, ETH), marketing (SEM, PPC, UGC), health and fitness (MRI, ADHD, BMI, IVF, MMA, HIIT), food and drink (BBQ, IPA), vehicles (ATV, VIN, OEM, DMV), real estate (HOA, FSBO), media and gaming (EDM, RPG, NPC), legal (DUI, NDA), places (UAE, NZ), and everyday speech (ASAP, RSVP, UFO). The same rules hold: no organization or brand acronyms — a trademark is handled by the trademark checks — and nothing whose ordinary-word reading would be mislabeled, so SAT, ACT, ARM, ABS and the like are deliberately left out; the easily-confused two-letter set grows by only five (UV, RV, NZ, MD, RN). Clearly commercial additions also count toward the industry the name points to, and BBQ and VPS join the premium-keyword list. One refinement comes with the growth: a two-letter acronym now counts only when the whole name reads as words — it can complete a name like uxhq, but two letters buried in an otherwise meaningless string no longer lend it meaning. Names carrying the newly recognized acronyms can score higher; every other name scores as before.
Model dv-5.5.0 — 2026-08-20
Widely recognized acronyms now count as meaning. A name built from common acronyms — UX for user experience, HQ for headquarters, HD, HR, AI, SEO, CRM, CEO, DIY, and a curated set of about sixty in all — used to read as a meaningless string, because only real dictionary words were recognized: a name like uxhq.com, which plainly says “UX headquarters” to anyone in the field, could be treated the same as random letters and capped as low-merit. The scan now recognizes these acronyms wherever they appear in a name, alone or joined with real words (seohub, cbdoil, evcharge), values them as the words they stand for, and shows each one expanded in the report so it is clear what was recognized. Clearly commercial acronyms also count toward the industry the name points to, and a handful with standing aftermarket demand — SEO, API, CRM, SaaS, NFT, CBD — join the premium-keyword list. Acronyms of organizations and brands are deliberately excluded: a trademark is handled by the trademark checks, not valued as a word. Names containing recognized acronyms can score meaningfully higher; every other name scores exactly as before.
Model dv-5.4.4 — 2026-08-15
Names offered for sale on a modern marketplace page are recognized again. A for-sale landing page used to be a thin placeholder; today the large marketplaces serve a full, working web page — navigation, written copy, payment options — and the scan, which had been taught to treat a page with real content as a genuine site, was reading those pages as ordinary active websites. Two things now identify a sale page regardless of how substantial it looks: the marketplace software the page is built from, and a page title offering that exact name for sale — which is how a sale page describes itself and how an ordinary business does not, since a business names its brand rather than its web address. A site that merely mentions “for sale”, including property, vehicle and marketplace sites whose titles say exactly that, is unaffected, because the name itself has to appear immediately before the phrase. The registration status reads “For sale” rather than “Active site”. Registration status is an overlay beside the score and is not part of the value math, but it does feed the checks that decide whether a name looks owned and in use, so a name newly recognized as for sale can lose that standing and score lower — which is why this carries a version of its own.
Model dv-5.4.3 — 2026-08-05
Scores are produced faster, and a lookup that cannot answer no longer holds up the rest. Each outside lookup now has its own time limit instead of all of them sharing a single ten-second allowance: the supporting checks — DNS, registration records, archive history, and domain authority — are given only as long as a healthy service needs, while the checks that decide whether the name has a live site keep the full allowance. A lookup that fails is also remembered for two minutes, so one unreachable service is not waited on again and again within the same scan or a run of scans; two minutes is short enough that a brief outage clears on its own. Almost every name is unaffected. A name is only scored differently when one of the shortened lookups was answering slower than a healthy service would: that lookup is now treated as unverified, which lowers the confidence in that part of the report and can therefore lower the score. Nothing about how the score is calculated has changed.
Model dv-5.4.2 — 2026-08-03
Parked and for-sale domains are recognized in four more situations. A parked name is one a registrar or parking service holds with no real site on it, and four common setups used to slip past the check and be reported as an ordinary registered name. First, when a name serves nothing at its plain address but its site actually lives at the “www” address, the scan now looks there too instead of concluding there is no site. Second, when a parked name forwards to a sales listing and that listing answers with an error — often because it blocks automated visitors — the scan previously stopped and learned nothing; it now recognizes the sales listing it was sent to, whatever the listing answers. Third, when a name points at a known parking service in its DNS records — the way most registrars park a name — the scan now reads that directly, which is often the only signal available; the Namecheap parking service was also added to the list of services recognized by name. Fourth, parking pages are now built like ordinary web apps, so they no longer look empty and could pass as real sites; they are now identified by the parking software that builds them rather than by how full the page looks. A page that shows it is a genuine, in-use site is still never marked as parked. This can change the score for some names in either direction — a name correctly identified as parked or for sale no longer counts as established, while a name whose site was simply found at “www” now does — so the model version advances.
Model dv-5.4.1 — 2026-07-21
Trademark matches now recognize the mark’s own owner. An exact match to a brand or product trademark — for example google.com or apple.com — used to be scored as low value and capped at 299, the same as a squatter holding someone else’s mark, even when the domain is the trademark holder’s own established, operating site. Now, when a name that matches a trademark shows evidence of being the holder’s own domain (a live, developed site; a long-held registration that isn’t for sale; or established web authority), it is scored as the valuable name it is, and the trademark appears as a note rather than a cap. The lens for a trademarked name is “how safely could a third party resell this?”: for the rightful holder the value is real but belongs to them, so a note explains that a third party couldn’t safely register or resell it; for anyone else — and for names that merely contain or misspell a trademark — the cap still applies. This changes the score for some names, so the model version advances.
Model dv-5.4.0 — 2026-07-19
Explicit penalty for keyword-stuffed and low-coherence names. A name that reads as SEO keyword-stuffing — it segments into six or more real words (or five across a long label), more than a brandable name carries — is now capped at 599 (Average). A genuinely incoherent name — not recognizable words, hard to pronounce, and carrying no premium keyword — that previously slipped past the low-merit cap only because a sibling extension happened to be registered is now capped at 449 (Weak). Both caps are waived when there is evidence the owner means it: a live developed site, an established registration (three or more years old and not for sale), strong cross-extension demand, rising search momentum, or indexed domain authority — so an established, descriptive business name is never marked down, only a fresh, unused one is. Coined but pronounceable brand names are unaffected. This changes the score for some names, so the model version advances.
Model dv-5.3.2 — 2026-07-11
Browser-verified reachability. Some parking and CDN edges block non-browser clients outright, so the raw scan fetch failed and a name every visitor could see was reported as having no live site. When a name resolves in DNS but the direct fetch fails, the scan now loads the page once in a headless browser and reports what it renders — so a blocked for-sale lander reads as “For sale” rather than “No live site”. Status reporting only; the value math is unchanged.
Model dv-5.3.1 — 2026-07-11
Better detection of parked and for-sale domains. A common current parked-name format (GoDaddy) serves a bare page that redirects in the browser to a same-site lander whose “for sale” content renders only in JavaScript — so a raw scan saw no cue and reported the name as an active site. The scan now recognizes that lander from its parking bundle, and when a suspected lander matches no known marker at all, it is rendered in a headless browser and its “for sale” text read directly — so future lander formats are classified from what they render, without waiting for a new marker. The registration status reads “For sale” rather than “Active site”. Status reporting only — registration status is an overlay beside the score and was never part of the value math, so no score changes.
Model dv-5.3.0 — 2026-06-29
Real-word recognition, broadened. The dictionary behind name reading grew from a small hand-curated starter list to a generated frequency dictionary of tens of thousands of real English and Spanish words (built from published word-frequency corpora and gated to genuine words). So a closed-up multi-word name like “domainvaluescore” is now correctly read as domain + value + score, and far more real-word names earn the brandability and demand credit their meaning deserves — while coined and gibberish names still score low. The detailed report now LISTS the recognized dictionary words (e.g. “domain, value, score”) rather than only counting them — premium commercial keywords shown in bold, each word tagged by where it sits in the name (start / middle / end), with a short legend so the breakdown is self-explanatory. No factor weights, curves, or band thresholds changed.
Model dv-5.2.0 — 2026-06-27
Product-name detection and clearer reporting. Well-known product marks (iPad, MacBook, PlayStation, Photoshop, and similar) are now recognized like famous brands: a name that embeds one is reported as carrying that product trademark and capped the same way a brand match is, so a name such as “AppleiPad” reports both the Apple brand and the iPad product. The keyword evidence now lists every matched term by where it sits in the name — prefix, middle, or suffix — including keywords that fall between two others in a compound, capped at five (a name that segments into more than five words is flagged as unusually long). Report language for trademark findings and score-cap reasons is now matter-of-fact — it states which trademark or product a name matches, rather than describing resale or seizure consequences. No factor weights, curves, or band thresholds changed.
Model dv-5.1.0 — 2026-06-25
Market-trend refresh. The premium-keyword list and the industry map gained 26 durable, researched 2024–2026 terms — AI-era (agent, compute, runtime, tensor, vector, schema, embed), matured fintech and crypto (token, wallet, vault, ledger, treasury, yield, escrow, custody), climate (carbon, climate), health (genome, longevity, peptide), security (sentinel, passkey, verify), and commerce/creator (merchant, checkout, creator). So a name in a category the market now prizes — an AI “agent” name, a “wallet,” a “carbon” name — scores for that demand and is placed in the right industry, the way buyers value it today. The update is baked in at a versioned checkpoint so the score stays reproducible, rather than tracking live search trends; no weights, curves, caps, or evidence collection changed.
Model dv-5.0.0 — 2026-06-25
Name character + market context. The fundamentals now report three explicit, separately-weighted scores — brandability (does it sound like a real company), pronounceability (easy to say aloud), and memorability (easy to recall and spell) — instead of one blended "brandability" reading. Liquidity adds the size of the industry the name points to (kept in a versioned table), so a name in a larger market reads as more liquid. A new name-continuity panel reports who owns the name and for how long, its historic web presence and Internet Archive depth, and present-day DNS, email, and SSL health — shown alongside the score but, like registration status and momentum, never folded into it, so the number stays reproducible. The detailed report now shows the count of recognized dictionary words, the full list of premium-keyword matches, and exactly which sibling extensions are registered, and the industry/continuity tables are version-stamped on every score.
Model dv-4.0.0 — 2026-06-22
Sharper name reading. Length now drives the fundamentals once instead of being counted three times, so quality, liquidity, and scarcity each measure something distinct. A dictionary recognizes real words inside a name — including across supported languages — so a genuine word like "voice" or a two-word name like "localmax" scores for its meaning, while unpronounceable, meaningless strings score far lower. Trademark and abuse risk can now only lower a score, never pad a meaningless one, and names on spam-dominated extensions (kept current automatically) are marked down and capped. A name that is not a word, not a sought-after term, and registered nowhere else is held to the bottom band. Premium results also show a separate, timestamped market-momentum read that is reported alongside the score but never changes it, so the score stays reproducible.
Model dv-3.0.0 — 2026-06-21
Major reframe: the score now measures the intrinsic value of the NAME (length, brandability, keywords, extension, liquidity, trademark risk) independent of who owns it — so it is equally useful to owners, investors evaluating available names, and buyers eyeing names others own. Registration status (available / registered / parked-for-sale / developed) is detected and reported as an overlay; registration age and archive history now only ADD value for a registered name and never penalize an available one, while the live page is reported as context but no longer adds to the score. A name positively known to be available is no longer capped for lacking registration history.
Model dv-2.4.0 — 2026-06-15
Core-provenance cap: when neither a registration date (RDAP) nor archive history (Internet Archive) can be verified, confidence is floored and the score is capped below Premium, so name scarcity alone can no longer certify a high score. Also stamps a real scan timestamp on every result and binds score reproducibility to the recorded inputs.
Model dv-2.3.1 — 2026-06-12
Reachability marking fix: a name whose origin cannot be contacted (edge 520–530 responses) is now correctly reported as not reachable instead of a reachable HTTP-error page. The live-site value was already zero in both cases — only the reported status was wrong.
Model dv-2.3.0 — 2026-06-12
Quality-gated the short-name premium: a short label only earns the full "scarce name" value if it is actually pronounceable. Unpronounceable, vowel-less gibberish (e.g. random consonant strings) now scores far lower on fundamentals, so a made-up name on a weak extension no longer reads as mid-tier.
Model dv-2.2.0 — 2026-06-12
Added a confidence score and value caps, exact-trademark and typo-squat handling, keyword precision fixes, sales comparables, and optional keyed premium signals (domain authority, search volume, and live trademark lookups).
Model dv-2.1.0 — 2026-06-12
Switched to public-suffix parsing so the registrable name (eTLD+1) is valued, added live-site value tiers and keyword-precision fixes, and began showing the reasons behind any value cap on free results.
Model dv-2.0.0 — 2026-06-12
Rebuilt on an investor-framed five-category model — fundamentals, market demand, liquidity, risk, and scarcity & provenance — blending in live Internet Archive history and cross-TLD registration data.
Model dv-1.0.0 — 2026-06-11
First public valuation model, launched with DomainValueScore.com: a 0–1000 estimate of a domain name’s market appeal from its observable characteristics.