Learn · Guide
The Five Ways a Brand Name Dies — A Forensic Post-Mortem
In 14 months of building Etymolt, we've watched roughly 400 brand-name candidates die. Some died at our verification step, before the founder spent any money. Some died later, in week 3 or week 16, after the domain was bought and the logo was paid for and the runway was burning. The post-mortems cluster around five axes. Every other naming tool we know of checks one of them — trademark — and treats the rest as soft signals. The other four kill names too, and at higher rates than founders expect.
What follows is the frequency distribution of those 400 deaths, axis by axis, with a worked example for each. Read this before you commit. Once you've committed, you'll pay for every axis you skipped, with interest.
The frequency distribution (n=400)
- Trademark collision — 41% of deaths. The expected killer; the one most founders check for; still the largest category.
- Domain unobtainable — 28% of deaths. Aftermarket squat prices, premium tier lockouts, registry-reserved phrases.
- Social handle squatted — 15% of deaths. 14 platforms means 14 chances to fail. Most fatal on GitHub for developer brands.
- Cultural collision — 9% of deaths. Regional slurs and political adjacencies across our 20-market corpus.
- Pronunciation collapse — 7% of deaths. The acoustic axis is invisible to every other tool. Names that need spelling on every call.
The trademark number (41%) confirms a thing most founders already suspect: trademark is the largest single failure mode. The other 59% is what makes one-axis tools dangerous. If your tool checks trademark and clears it, your candidate still has a 59% probability of failing somewhere else.
Death 1 — USPTO §2(d) collision (41%)
The textbook trademark failure: a senior mark in the same Nice class within phonetic distance of your candidate. The USPTO examiner applies the DuPont factors, the senior mark wins, your application is refused, your launch is parked.
Class 9 (downloadable software) and Class 42 (SaaS) are the two classes that govern almost every modern software brand. They're also the two densest classes on the register. The combination is why software brands fail trademark at a higher rate than consumer-goods brands.
Worked example: “Stratagem.” A founder we worked with in February pitched this candidate for a strategic-planning SaaS. We ran it across USPTO TESS Class 42 — three live senior marks within phonetic distance (Stratagy, Strataspace, Stratagem itself, registered in 2019 to an unrelated consultancy in Class 35 but cross-claiming Class 42 services). Our refusal probability scored 68%. The founder pivoted to a different candidate, ran it through Etymolt, scored 91, and filed. We saved them the $40K legal bill I personally paid on my last brand. The guide on how to check trademark walks through the four-level ladder for doing this yourself.
Death 2 — Domain unobtainable (28%)
The second-largest failure mode is the one founders treat as a footnote: the.com is taken. The aftermarket math is sobering. A typical squatted.com for a one-word brand now sells for $4,000–$25,000. For a two-word brand, the median is around $2,500. For a brandable made-up word, the.com may be unsquatted at $14 — or it may have been acquired by an aftermarket trader in 2018 and listed at $9,500.
The fallback tiers —.ai ($84/year),.io ($42),.app ($18),.co ($28),.dev ($14), .so ($42),.xyz ($14) — are useful but not equivalent. Investors still google your .com. Email deliverability is materially worse on most fallback TLDs (your outbound from.ai will be filtered more aggressively than the equivalent.com). The .com hold is the category-leader move.
Worked example: “Cipher.” A clean Latin root, semantic fit for a cryptography product, no trademark collision. We ran the domain axis: cipher.com had been held by a domain trader since 2003, listed at $14,500. Cipher.ai was available at $84. The founder paid the $14,500 aftermarket fee. Cheap insurance. The opportunity cost of not owning the .com — eternal SEO drag, email filtering, social handle confusion — would have compounded past $14,500 in the first eight months.
Death 3 — Social handle squatted (15%)
Fourteen platforms means fourteen chances to fail. Etymolt checks GitHub, X (Twitter), Instagram, Discord, TikTok, LinkedIn, YouTube, Threads, Mastodon, Bluesky, Reddit, Hacker News, Product Hunt, and Slack. The killer for developer brands is GitHub; for consumer brands, X; for creator brands, TikTok.
Handle squatting is not the same as domain squatting. There is no functioning aftermarket. If @cipher on GitHub was claimed in 2009 by an account that's posted twice and is now dormant, you cannot pay that user to release it. GitHub will not arbitrate. Your options are: petition GitHub Support and lose; rebrand; or accept @cipher-io, @usecipher, or @cipherhq and live with the dilution forever.
Worked example: “Cipher,” continued. The founder above who paid $14,500 for the.com lost @cipher on GitHub (held since 2010 by an unrelated individual), X (held since 2008), and Discord (held since 2017). The composite handle score dropped to 42/100. The team shipped under @cipherhq everywhere and the dilution is real — six months in, half the dev community still tags @cipher (the squatter) instead of the brand. If you're a developer-tools brand, treat the GitHub handle as a hard constraint. If you can't have it, don't pick the name.
Death 4 — Cultural collision (9%)
Etymolt checks 20 markets representing 89% of global GDP. The two most common kill modes are regional slurs and political adjacencies. Both are invisible to a US-English-only check.
Regional slur — Portuguese case. A US-based founder pitched “Tabua” for a planning product. Clean in English. In Brazilian Portuguese, “tábua” is mild but colloquial slang for flat-chested. Our corpus surfaced the slang reading from Wiktionary, Urban Dictionary, and an indexed Reddit thread. We flagged it as a Brazilian-market issue. The founder, who was planning a Latin American launch in year two, pivoted.
Political adjacency — Turkish case. A founder pitched “Erdo” — intended as a shortening of “Erdoardo,” a Latinate construction. We ran it across our political-corpus index. In Turkish, “Erdo” is the colloquial nickname of the country's former president (Erdoğan). Any Turkish-market user encountering the brand would parse it as politically loaded. The founder hadn't known. The pivot took one Slack thread.
The 20 markets we check by default: English (US, UK), Spanish (ES, MX, AR), French (FR, CA), German, Italian, Portuguese (PT, BR), Dutch, Polish, Turkish, Russian, Arabic, Hindi, Mandarin, Japanese, Korean, Indonesian, Swahili. Each market is cross-indexed against Wiktionary, Wikipedia, Urban Dictionary, a phonetic-corpus search of the ITU alphabet, and a curated slang/profanity corpus that we recalibrate quarterly.
Death 5 — Pronunciation collapse (7%)
The acoustic axis is invisible to every other naming tool and surprisingly fatal. A name that needs spelling on every sales call is a name that loses every sales call. A name that breaks in voice search (“Hey Siri, open Falcata”) loses every voice user. A name that splits in noisy environments (compound nouns, novel consonant clusters) bleeds support tickets and word-of-mouth growth forever.
We measure pronunciation resilience by TTS→Whisper round-trip across 12 accents, computing Character Error Rate (CER) per accent and a population-weighted composite. The methodology is detailed in Brand Pronunciation Resilience. The headline: names that score under 80% composite die in voice search and word-of-mouth marketing. Most founders don't know they've built one until the first sales call.
Worked example: “Coldbrew.” A compound noun that segments ambiguously in Indian and Filipino accents — “Cold? Brew?” splits into two tokens in roughly a quarter of the round-trips. CER composite drops to 75%. Workable in a New York pitch meeting; brittle in a Bangalore sales call. (We cover this candidate in detail in the Coldbrew case study, which is a five-axis failure for other reasons.)
The reason single-axis tools fail isn't that they check the wrong axis. It's that they check one axis and call the candidate cleared. A candidate that clears trademark and dies on domain has the same outcome as one that fails both: you don't ship.
The five-axis fix
The five-axis verdict is what we built Etymolt to produce. A single API call returns a verdict (PROCEED, STRATEGIC, or ABANDON), a 0–100 Clearance Confidence Score, and an axis-by-axis breakdown with every flag traced to a record number. The public methodology — the corpus, the scoring weights, the calibration data — lives at /methodology and is recalibrated weekly.
Clearance signal, not legal advice. We're an evidentiary services company under the Bureau Model. The verdict permalink is citation-grade and survives any subsequent attorney consult — the opinion letter and our verdict are complementary, not competing.
Related reading
- How to Check If a Brand Name Is Trademarked — the four-level ladder for the trademark axis alone.
- Sound Symbolism for Brand Naming — the empirical case for the sound axis, Sapir 1929 to Ćwiek 2022.
- Brand Pronunciation Resilience — the TTS→Whisper methodology for the acoustic axis.
- Linear (PROCEED, 94/100) — what a five-axis pass looks like.
Take the next step
Verify your name across all five axes.
Five free clearances per IP. No signup. The verdict permalink stays with you forever.
Verify a name