What Makes a Good Company Name: 7 Properties That Separate Names That Work
Most founders approach company naming as a creative exercise -- brainstorming on a whiteboard, testing names on friends, checking if the domain is available. The result is usually a name that feels fine in the room but creates friction in the market: it gets misspelled, mispronounced in international markets, confused with competitors, or simply fails to carry the brand positioning the founders intended.
The problem is that naming is treated as taste when it is actually a measurable property of sound. Names are decided in approximately 150 milliseconds -- before the listener has processed a single word of meaning. What registers in that window is entirely phonetic: the acoustic signature of the consonants and vowels, the rhythm and syllable structure, the sound associations built up over a lifetime of exposure to language.
This is not a theoretical observation. It is a research finding from psychoacoustics and cognitive linguistics, replicated across dozens of studies over four decades. The names that endure -- Stripe, Slack, Zoom, Notion, Apple, Intel -- do so because they score well on specific, measurable phonetic properties. The names that fail often do so for equally specific, measurable reasons.
Seven properties separate company names that work from those that do not.
Sound symbolism is the phenomenon whereby specific phonemes carry consistent meaning associations across languages and cultures. Hard stops (/k/, /t/, /p/) project precision, authority, and decisiveness. Fricatives (/s/, /f/, /sh/) project speed, elegance, and forward motion. Nasals (/m/, /n/) project warmth, reliability, and approachability. Open vowels (/a/, /o/) project confidence and energy.
A name with high phoneme fit uses sounds that reinforce the brand's intended positioning. A name with low phoneme fit creates an acoustic signal that contradicts it -- and the contradiction is felt by the audience before they can articulate why the name seems off.
The most durable brand names occupy a specific territory first described by naming strategist David Placek: unexpected but immediately believable. They do not describe the product. They do not feel invented for the sake of novelty. They feel like something that could have existed for a long time but was never applied in quite this way.
A name that is too descriptive ("Taskflow", "CloudStore", "BrightPath") is forgettable because it blends into the category. Every competitor sounds similar. A name that is too abstract -- arbitrary coinages with no phonemic anchor -- provides no handle for memory to grip. "Zoplux", "Greptia", and "Forvix" fail for this reason.
The ideal is a word or root from an adjacent domain, applied with confident precision to a new category. Stripe is a visual mark applied to payments. Slack is the amount of rope that is not taut, applied to communication. Figma is a figure drawing term applied to interface design. Each carries semantic tension: the referent is not obvious, but once the connection is clear, the choice feels inevitable.
Processing fluency is the cognitive ease with which a stimulus is perceived and understood. Names with high fluency are decoded quickly, feel familiar even on first encounter, and are associated with competence and trustworthiness. Names with low fluency require more cognitive effort, create uncertainty, and are more likely to be misspelled and mispronounced.
Fluency is a function of three factors: phonotactic regularity (do the sound combinations follow expected English patterns?), syllable structure (are the stress patterns predictable?), and orthographic transparency (does the spelling predict the pronunciation?). Names like "Zoom," "Stripe," and "Slack" score near-perfect on all three. Names like "Xaaxe," "Vyyr," and "Ttpwr" fail all three simultaneously.
The practical test is simple: say the name out loud to five people who have never seen it written. Can they spell it after hearing it once? Can they say it correctly after seeing it once? A name that fails either test has a fluency problem that will create friction at every stage of business development.
Most software products reach international markets within the first two years. A name that contains phonemes absent from your target markets' languages, or that means something unfortunate in a major language, is a structural problem that cannot be solved with marketing spend.
The highest-risk English phonemes for international audiences are the American /r/ (consistently mispronounced or dropped by Mandarin, Japanese, and Korean speakers), the /th/ cluster /ð/ /θ/ (absent from most European, Asian, and Latin American languages), and the /w/ onset (uncommon in Mandarin, often replaced with /v/ in European markets). A name built primarily on these phonemes will sound different in every non-English market it enters.
This does not mean avoiding all imperfect phonemes -- it means making the risk decision consciously rather than discovering the problem after launch materials are printed. Zoom passes this test well: /z/, /uː/, and /m/ are all phonemes that exist in or can be cleanly approximated by most major language families.
A name chosen at founding becomes a liability when the company grows beyond its initial category. Category-descriptive names ("SaaS Tools Inc", "CloudBackup", "MobileFirst") create a positioning ceiling: the name works when the product does exactly what it says, but becomes an anchor when the company expands.
The test for scalability is whether the name could plausibly be the name of the company you intend to be in ten years, not just the company you are today. Amazon was not named "BooksOnline." Apple was not named "PersonalComputers." The name works because it does not describe the product -- it claims a position in the buyer's memory that can expand to support whatever the product becomes.
For most software companies, this means avoiding names built on the current product category, the current technology paradigm (AI, blockchain, cloud), or the current go-to-market focus (SMB, enterprise, consumer). These are constraints that will change; the name will not.
Technology-specific names: "AIbolt," "CloudPath," "BlockVerify" -- name the method, not the outcome. When the technology changes, the name is wrong.
Category-descriptive names: "TaxSimple," "HrPlatform," "MarketingOS" -- describe the current product exactly. Any pivot makes the name misleading.
Market-specific names: "SMBConnect," "EnterpriseGrowth" -- built around the initial customer segment. Limits upmarket or downmarket expansion.
A name must survive at least three distinct sentence contexts in which founders and customers will use it. If it sounds wrong in any of them, it will create friction every time that context occurs.
The three contexts to test:
Slack passes all three. "We are Slack" works. "Slack raises $160 million" works -- this actually happened. "Join Slack today" works. The name carries its weight in every sentence despite -- or because of -- the surprise of a common English word used in an unexpected register.
Names that fail the first context tend to be too casual or too playful for business use. Names that fail the third tend to be too formal or too hard to say quickly in a call-to-action. Names that fail the second tend to be too category-descriptive -- they sound like a feature, not a company.
The best phonetic name in the world is worthless if it cannot be secured as a trademark and domain. The availability check is not the starting point -- it is the final gate that eliminates candidates that pass all six other tests but are already claimed.
The correct sequence is: run tests one through six first. For every name that passes all six, check availability. Running the availability check first inverts the process: you end up optimizing for what is unclaimed rather than what actually works, and produce a shortlist of names that are available precisely because they have the low distinctive quality that makes them easy to overlook.
What to check, in order:
Why most names fail
Most founder-generated names fail because they were evaluated on a single criterion -- does it sound good? -- and this is both the wrong question and an unanswerable one without structure. "Sounds good" is a subjective judgment that depends entirely on what the listener already knows about the company, the founder's familiarity with the name from weeks of exposure, and the pressure of the naming process itself.
The seven properties above are not subject to familiarity bias. They are measurable features of the name's phonetic structure, evaluated against the brief's requirements, independent of the founder's emotional investment in any particular candidate.
The second reason names fail is anchoring: the first name generated in a whiteboard session becomes the implicit benchmark for all subsequent candidates. Every name that does not sound like the first name seems wrong, even when the first name is objectively worse. The way to defeat anchoring is to generate a large enough pool -- 300 candidates or more -- before applying any qualitative filter, so the shortlist reflects the full distribution of possible names rather than the biases of a single session.
The third reason is the availability-first mistake: running domain and trademark checks before phonetic evaluation. This produces a shortlist of names that are unclaimed precisely because they lack the distinctive qualities that make a name worth claiming.
The right evaluation order
A systematic naming process applies the seven properties in the order that maximizes signal-to-noise ratio:
This sequence is why systematic naming processes consistently produce better shortlists than whiteboard sessions: the evaluation order is designed to find the names that actually work, not just the names that were thought of first or happened to have the .com available.
Voxa runs 300 to 1,500+ candidates through all seven properties automatically. The 14-dimension phoneme scoring engine measures fit, distinctiveness, fluency, cross-language risk, and context performance for every candidate -- against your specific brief, category, and competitors. Delivered as a ranked PDF in approximately 30 minutes.
Get my naming proposal →