Brand Naming · Criteria
What makes a good brand name — the 6 criteria.
Across the corpus of about 1.4 million brand-name candidates Etymolt has scored, roughly 8% reach a PROCEED verdict on the first pass. The other 92% fail one or more of six gates. Those six gates are what make a name good, in the precise sense that a name passing all six survives launch and a name failing any one of them tends not to.
The criteria are widely agreed across the canonical agencies — Lexicon, Operative Words, Eat My Words, and Catchword all converge on something like this list, with slightly different vocabulary. What follows is the operational version, with the evidence from the corpus and the anti-examples that demonstrate each failure mode.
1. Distinctive — not the category word
A distinctive name is legally protectable, semantically own-able, and search-findable. The opposite is the descriptive name — the name that describes what the product is so directly that the USPTO refuses registration under §2(e)(1) and Google returns category competitors instead of you.
Coldbrew Coffee describes the product. The USPTO refuses. Search returns Stumptown, Blue Bottle, and a thousand local roasters. The name is doing none of the three jobs a name should do, and it's why our Coldbrew case study verdict came back at 28/100.
Stripe is distinctive. The word is found-object — repurposed from an unrelated domain (a magnetic card stripe, a band of color). The trademark register at filing showed no live confusable mark in fintech. The word is fully own-able in the payments category. The arithmetic of distinctiveness is asymmetric: distinctive names compound; descriptive names erode.
In our corpus, names that pass §2(e)(1) descriptiveness scoring at >80% confidence reach PROCEED at 4× the rate of names that score <60%. The arithmetic is brutal.
2. Pronounceable — survives noisy channels
A pronounceable name is one a stranger can say correctly after hearing it once. It is also one Whisper can transcribe correctly under a 10dB SNR. The first test is the social-meaning test; the second is the acoustic test. A name that fails the acoustic test is a name that loses sales calls — every conference call, every podcast, every cold-call discovery sound is a quiet erosion of brand equity.
Xobni (Outlook plugin, defunct) was an anagram of inbox spelled backwards. The spelling did not predict the pronunciation for any English speaker. The founders spent ten thousand recorded minutes explaining the pronunciation. The brand died.
Linear survives every accent Etymolt tests against. The pronunciation distribution in our 12-accent benchmark is tight: /ˈlɪniər/ across Northern American, Southern American, Standard British, Australian, Indian English, Filipino English, Singaporean English, with one acceptable variant /ˈlaɪniər/ in a small minority. That is the kind of distribution a good name shows. See Brand Pronunciation Resilience for the methodology.
3. Memorable — sticks after one exposure
Memorability is the property of a name to be reproduced from memory the next day, the next week, the next month. It is partly phonetic — names with strong consonant-vowel contrast stick harder — and partly semantic, because names that connect to an existing concept attach to existing memory structures.
Bombas (D2C socks) is memorable. The hard plosive onset, the open-vowel close, the double-B rhythm, the connection to the Latin for "bumblebee" (the brand's namesake) — all of it stacks. A consumer asked to recall the brand of socks they bought six weeks ago can reproduce Bombas at a rate meaningfully above category baseline.
Highnote (the B2B card-issuer) is memorable for the cohort it targets — fintech operators who recognize the metaphor of a high note as a peak performance. The two-syllable compound, the bright vowel, the metaphor that does work all the way down. Memorability is not always alliteration or rhyme; sometimes it is the metaphor that earns its keep.
4. Available — trademark, domain, handle
Availability is the boring axis that kills the most names. A candidate that scores well on every other axis but collides with a live USPTO senior mark in the same Nice class is not a candidate. A strong candidate whose.com is held by a 2002 vintage squatter at $80,000 is a candidate with an asterisk. A strong candidate whose @brand handle is held by a dormant 2007 X account is a candidate that ships under @brandhq forever.
In our corpus, 41% of candidates fail the trademark axis on first pass. 38% fail the domain axis (no.com under $5,000). 27% fail the handle axis (no @brand on at least one of the canonical 12 platforms). The overlap is significant; the failure rate per axis is much higher than founders expect because the axes are partly independent.
The trademark failure mode is the most expensive — see How to Check If a Brand Name Is Trademarked for the four-level ladder. The domain failure mode is the most negotiable; aftermarket .com purchases at four and five figures are routine. The handle failure mode is the quietest — it does not stop launch, but it dilutes the brand at every social touch for the life of the company.
5. Meaningful — the metaphor does work
A good name carries a metaphor that does work. Not literal meaning — literal meaning is the descriptive trap. Metaphorical meaning — the property of the product re-expressed as a concrete word. Linear is the metaphor of straight, ordered progress. Stripe is the metaphor of the magnetic stripe and the band of color. Mercury is the metaphor of speed and the messenger of the gods.
The Eat My Words framework — the SMILE rubric (Suggestive, Meaningful, Imagery, Legs, Emotional) — names this property explicitly. A name that scores high on imagery is one that triggers a picture in the reader's head. The picture is what makes the name stick. The picture is also what gives the name room to grow — metaphors with multiple branches let the brand expand into adjacent products without renaming.
The opposite is the empty name — the coined word that triggers no picture and carries no metaphor. Coined names can work (Mintlify, Tigris, Convex) but the cost of meaning has to be paid back in category education. The founder who picks a coined name signs up to do that work. The founder who picks a metaphor-rich real word inherits the work the language already did.
6. Cohort-fit — sounds like what the buyer expects
Cohort-fit is the property of the name to belong, phonetically and stylistically, to the register the buyer already reads in. The B2B SaaS cohort reads Latinate two-syllable nouns (Linear, Vercel, Modal, Convex). The D2C consumer cohort reads warmer, often human-sounding names (Casper, Bombas, Glossier, Hims, Allbirds). The fintech cohort reads short, hard-consonant nouns that sound like money (Stripe, Mercury, Brex, Ramp, Bolt).
A name that fits the wrong cohort is not technically bad — it is just expensive. The buyer pauses on first encounter. The pause is a cost. The pause compounds. We unpack the cohort canons in SaaS naming patterns and D2C brand naming.
The check is to read the candidate aloud next to ten names already established in the cohort. If the candidate fits — same syllable count, same register, same consonant density — it has cohort-fit. If it sounds like a different category, the buyer will subconsciously file it in a different category, and the marketing dollar has to undo the filing.
How the six criteria compose
The criteria are not independent. A name that scores high on distinctiveness often scores low on memorability — distinctive coined words are harder to recall than real words. A name that scores high on meaningful sometimes scores low on available — the more meaningful the word, the more likely it is already taken. The good name is the one that finds the local optimum across all six, not the global maximum on any one.
The Etymolt verifier weights the six axes by category. For B2B SaaS, the distinctive and cohort-fit axes carry more weight; for D2C consumer, the meaningful and pronounceable axes lead; for fintech, available (trademark) leads because regulated categories tolerate less risk. The weighting is published in the methodology and revised quarterly against the corpus.
The aggregate score is the Clearance Confidence Score. A score of 80+ is a PROCEED — every axis cleared with meaningful headroom. A score of 60-79 is STRATEGIC — workable with a punch list. A score below 60 is ABANDON — serious blockers present. A score under 40 is ABANDON — the math is hostile and the candidate should be dropped.
Why only 8% reach PROCEED on the first pass
The 8% PROCEED-on-first-pass rate surprises founders the first time they hear it. It is a function of the math of compounding independent failure modes. If each axis fails 20% of the time independently, the joint pass rate is 0.8 to the sixth power, or about 26%. The actual rate is lower (8%) because the axes are not fully independent — descriptive names fail both the distinctive and the available axes; coined names that fail pronunciation often also fail memorability.
The implication is that the right naming workflow expects to verify ten to twelve candidates to find one PROCEED. Founders who plan for that expectation save weeks of emotional drag; founders who do not plan for it spend three weeks trying to repair a single failing candidate before they accept the math.
Related reading
- How to Come Up With a Brand Name — the 7-step process that produces candidates this rubric scores.
- The Best Brand Naming Frameworks — the SMILE/SCRATCH rubric and the Lexicon phoneme model.
- Sound Symbolism for Brand Naming — the empirical case underneath the pronounceable / memorable criteria.
Disclaimer
Etymolt operates under the Bureau Model. We surface clearance signals across the five axes; we do not provide legal advice. A trademark opinion belongs to a licensed attorney in your filing jurisdictions.
Take the next step
Score your candidate against the six criteria.
Five free clearances per IP. Every flag traces to a record number. Three seconds, end to end.
Verify a name