Most brand names die quietly.
Not because they were bad, exactly. They were fine. Pronounceable. Trademark-able. The handle was open on Instagram and the .com was within budget. The founder said it out loud to a friend, the friend nodded, and the brand shipped. Two years later the brand is gone, or it's renamed, or it's the one customers can never quite remember when they're recommending it to someone at a dinner party.
The problem isn't that the founder picked a bad name. The problem is that nearly every guide on the internet about naming a brand is solving for the wrong target.
Most guides will tell you to pick a name that's memorable. A name your customers will remember. A name that sounds good. They'll give you a five-step process: define your brand DNA, generate a longlist, run a trademark check, gather feedback, decide.
That process is not wrong. It's just incomplete in a way that costs brands more than they realize. Fifty years of peer-reviewed research in psycholinguistics and consumer psychology says the target shouldn't be memorable at all. It should be polarizing. And the most important decisions about a name happen at the level of sound, before meaning ever enters the room.
This is a guide to what the linguistic research actually says, what it looks like applied to real DTC brands, and what you can do Monday morning without paying a six-figure agency or letting a generator spit out 200 made-up words.
What every other guide gets right (and where it stops)
Spend an afternoon reading the top results for "how to name a brand" and a pattern emerges. The processes are nearly identical. Some have seven steps, some have nine. They all hit the same beats:
- Define your brand identity, mission, values, audience.
- Generate a longlist using techniques like word association, compounding, abstraction.
- Screen for trademark conflicts and domain availability.
- Narrow to a shortlist of finalists.
- Test with focus groups, surveys, or "real-world" exposure.
- Pick and protect.
This is fine. It's a reasonable scaffold for the visible work of naming, the part a founder can do with a whiteboard, a USPTO search, and a Calendly invite to five customers. It catches the easy failure modes. It will keep you from naming your skincare brand the same thing as an existing one in the same class, or from picking a word that means something embarrassing in Portuguese.
The trouble is that this entire scaffold is downstream of a decision that nobody in the SERP makes explicit: what should the name actually do?
If the answer is "be memorable," the generate-and-screen process is well-suited. Memorability is a function of repetition and recognition, and most of the seven-step processes are checks for whether a name is repeatable and recognizable enough to clear the floor.
But "memorable" turns out to be the wrong target.
What the research actually says
There is a small library of peer-reviewed research, going back nearly a hundred years, on how the brain processes names. It almost never appears in mainstream naming guides. When it does, it shows up as a single sentence ("hard consonants feel strong") with no citation. The actual findings are sharper than that, and they point in a different direction than the conventional advice.
Three threads matter most for naming.
Sound symbolism: the brain assigns meaning to phonemes before semantics arrives
In 1929, the linguist Edward Sapir ran a study that became one of the foundational papers in psycholinguistics. He showed subjects pairs of nonsense words like mil and mal and asked which represented a "smaller table" and which a "larger table." Across age groups, more than 80% of people matched mil (front vowel) to "small" and mal (back vowel) to "large." This wasn't cultural. It wasn't English-specific. Front vowels read as smaller, lighter, faster, brighter, almost universally.
A few decades later, the gestalt psychologist Wolfgang Köhler ran a different version of the experiment with shapes. He drew two figures, one all curves, one all spikes, and asked which was "bouba" and which was "kiki." Ninety-five percent of people picked kiki for the spiky shape. The 2022 cross-linguistic study published in Royal Society Open Science validated the effect across 229 languages.
This is sound symbolism: the regularity with which specific sounds carry specific connotations, in the brain of the reader, before semantic meaning is decoded. It happens fast and it happens automatically.
For naming, the consequence is that the phonemes you pick are already telling a story about your brand. Klink (2000) published the consonant mappings in Marketing Letters. Yorkston and Menon (2004) ran a now-famous experiment in the Journal of Consumer Research where they served the same ice cream to two groups and changed nothing except the brand name. One group got Frish, the other got Frosh. The Frosh group rated the ice cream as creamier, smoother, and richer. The back vowel did the work.
The phoneme-class associations are the practical takeaway. They are not subtle, and they have been replicated enough times that you can use them as a working map:
- Voiceless stops (p, t, k): sharp, urgent, decisive, percussive. The reason Pepsi snaps and Kodak clicks. Useful when the brand wants to feel fast, modern, or precise.
- Voiced stops (b, d, g): heavier, more grounded, more reliable. BMW, Dodge, Google. Useful when the brand wants to feel substantial.
- Fricatives (s, f, v, z, sh): smooth, fast, often luxurious. Pathak, Calvert and Lim (2020) showed in International Journal of Market Research that fricatives correlate with luxury perceptions specifically. Sonos, Versace, Visa.
- Liquids (l, r): flowing, fluid, soft. Allbirds, Lululemon. Useful when the brand wants to feel approachable or sensual.
- Nasals (m, n): warm, rounded, intimate. Mailchimp, Notion. Useful when the brand wants to feel friendly or trustworthy.
This is not a recipe. A name doesn't become premium because you stack fricatives in it. The map tells you what the phonemes are doing in the background of the reader's perception, so you can decide whether you want the background to match or contrast with your positioning.
If your name's sound profile contradicts what your category wants from the buyer, you're walking into the room with the wrong handshake.
Processing fluency: the brain confuses easy-to-read with true
The second thread is processing fluency, the subjective ease with which the brain processes a stimulus. The core finding, established in a 2004 review by Reber, Schwarz, and Winkielman in Personality and Social Psychology Review, is that the brain misattributes fluency to liking, to quality, and to truthfulness. When something feels easy to read, we tend to believe it. When it feels hard, we tend to distrust it.
For brand names, this has a quantifiable consequence. In 2006, Alter and Oppenheimer published a study in the Proceedings of the National Academy of Sciences showing that stocks with easy-to-pronounce ticker symbols outperformed stocks with hard-to-pronounce ones in the first weeks after IPO. The pronounceable ticker was being misattributed as the safer investment.
Pronounceability is not the only fluency driver. Length matters. Two to three syllables is the consumer-brand sweet spot, with about 74% of successful consumer brands clustering there. Familiar phonetic patterns help. Words that follow English phonotactic rules (the implicit grammar of which sound combinations are "allowed" in English) process faster than words that don't.
This is part of why Xumo and Qwest feel inherently less reliable than a real English word used unexpectedly. They demand decoding effort. Stripe doesn't.
Optimal incongruity: polarizing beats comfortable, long-term
The third thread is where the research actually flips conventional wisdom on its head. In 1970, Daniel Berlyne published "Novelty, complexity, and hedonic value" in Perception & Psychophysics. He showed that preference follows an inverted-U curve based on novelty and complexity. Too familiar reads as boring. Too novel reads as overwhelming. The peak is in the middle, at optimal incongruity.
Apply this to naming. A name that everybody finds fine is, by definition, sitting on the low-novelty end of the curve. It has cleared the boredom threshold, but it hasn't done anything to claim attention. A name that some people love and other people hate is sitting closer to the peak. It generates engagement on both sides. It is, in marketing terms, polarizing.
Polarizing is not the same as weird. A polarizing name is one that takes a strong phonetic or semantic stance. Liquid Death is polarizing: it names a beverage with the word death, a phonetic and semantic move that the category has spent decades avoiding. Liquid Death sold for $1.4 billion. Pure Mountain Water Co. does not.
The opposite of polarizing is not "safe." The opposite of polarizing is forgettable. A name that no one has a reaction to is a name no one will remember to buy.
Put the three threads together and a different target emerges: a name should be polarizing in stance, processing-fluent in execution, and phonetically aligned with the brand's actual positioning. That is a target the standard seven-step processes are not built to hit.
Anatomy of four working names
The fastest way to internalize what this looks like is to take real brands apart at the phoneme level. Here are four DTC and tech wins, broken down.
Stripe: the stop-liquid blueprint
Say it out loud. Stripe. Four phonemes packed into one syllable: a voiceless stop opener (s-t-), a liquid (r), a short bright vowel (-i-), and a stop-fricative close (-p, -e silent).
The phonetic pattern is CCVCC, opening with a triple-consonant cluster (str-) that is, in linguistic terms, a premium English construction. It's hard to say lazily. You have to commit to the str- cluster. That commitment is part of why the word feels sharp and decisive rather than soft or evasive.
The semantic move matters too. Stripe already exists in your lexicon. You've seen stripes on a tie, on a road, on a credit card. Dropping the word into "we charge credit cards with Stripe" doesn't ask your brain to decode a new symbol. It just lands. That's processing fluency doing work in the background.
Compare to the wave of fintech that came after: Square (softer close, more friendly), Plaid (playful, less percussive), Mercury (mystical, three syllables, soft consonants), Ramp (warm opener, less surgical). All are good names. None are Stripe. Stripe is the coldest, fastest, most decisive of the lot, because the phonetic stack rewards a category that wants to feel like infrastructure.
Liquid Death: the optimal-incongruity stance
Liquid Death is the textbook case of polarizing-beats-memorable. The name violates the category convention of beverage naming, which has historically gone toward freshness, purity, and gentleness (Pure, Fresh, Crystal, Mountain, Spring). It picks the word death, with its hard stop-stop close, and pairs it with liquid, a liquid-heavy three-syllable trochee. The contrast is the point.
At the phonetic level, the name leverages a deliberate anti-fluency move. Death is short, percussive, hard. Liquid is soft, flowing. The collision between them is the brand's whole identity. Mainstream water brands sand the edges off their names. Liquid Death sharpened them.
This is Berlyne 1970 made concrete. The name pushes far enough toward the high-novelty end of the curve to claim attention without flipping over into incomprehensibility. Plenty of people hated it at first. Plenty more loved it. Both reactions paid the same dividend: nobody who heard the name forgot it.
Olipop: the compound trochee with category cues
Olipop is what a soft-sound-symbolism brand looks like done well. Two syllables, trochaic stress (OH-li-pop, stress on the first), heavy on liquids and stops in a friendly arrangement (l, p). The name reads warm and round.
The semantic move is a compound. Oli- gestures at oligosaccharides, the prebiotic fibers that the product centers. -pop tells you it's a soda. The compound is doing what Wisniewski (1996) called conceptual combination: the brain reads two familiar fragments and fuses them into a third concept that didn't exist before. PowerBook does this (power + book). DreamWorks does this (dream + works). Olipop does it with a category cue (functional ingredient) and a category descriptor (soda) in one move.
Compare to the rest of the category. Poppi uses a similar onomatopoeic move. Health-Ade uses a heavy hyphenated compound. Culture Pop stacks two abstract nouns. Olipop wins on processing fluency: the name is shorter, the syllable structure is cleaner, and the conceptual combination is more transparent. You don't have to be told what Olipop does to guess at it.
Sonos: the palindrome with soft sibilants
Sonos is what restraint sounds like. Two syllables, palindrome (the same letters forwards and backwards), opening and closing on the soft fricative s. The phonemic profile is gentle: a sibilant in, an open vowel, a nasal pivot, an open vowel, a sibilant out. Everything is rounded, balanced, ambient.
This is no accident. Sonos sells products designed to disappear into a room, to be heard without being noticed. The name does the same work. There is nothing percussive in it, nothing that demands attention. It carries the brand's whole positioning in five letters.
The palindrome is the structural flourish. Palindromes have a small but consistent edge in recognition memory: the symmetric structure is easier for the brain to encode and retrieve. Civic, Honda's flagship sedan, runs the same trick at a much larger scale.
These four are deliberate choices that align sound, structure, and positioning. Each one would be a worse name if any of the three were off.
The third path
If the seven-step naming guide does not get you here, and the brand-name generator definitely does not, what does?
Historically, there have been two options.
The first is a naming agency. Lexicon, Interbrand, A Hundred Monkeys, Operative Words: the firms that do this work at the highest level charge between $50,000 and $250,000 for a brand name, take three to six months, and run methodologies that look very much like what we just walked through, with additional layers of linguistic screening, trademark clearance, and stakeholder workshopping. They are very good. They are also priced for Fortune 500 budgets, which is why the founders who need them most can almost never afford them.
The second is a random brand name generator. Namelix, Looka, Shopify's tool, the long tail of AI wrappers that have launched in the last two years. These are useful for one thing: producing a high-volume longlist of made-up-sounding words. They are useless for everything that comes after. The generator has no theory of your category, no theory of sound symbolism, no theory of optimal incongruity. It has a vocabulary list and a thesaurus. Most of the names it produces will sound vaguely tech-startup-y because that's what its training data optimizes for. None of them are subcategory-native.
The third path is applied linguistics methodology, run at a founder-accessible price point. That is what we are building at Brand Namers. The Signal & Noise Method takes the research outlined above and turns it into a five-layer system that runs on your specific brief: behavioral intake, register classification, sound-symbolism alignment, processing-fluency scoring, and competitive distinctiveness against the actual brands you'll share a shelf with. The output is not a list of 200 made-up words. It is a curated shortlist of names with linguistic rationale you can hand to a trademark attorney.
This guide is not a sales pitch. It is the public layer of the methodology. Most of what we just walked through, you can run yourself.
What to do Monday morning
If you have a shortlist of brand names you're working with right now, run these four tests before you book any focus groups.
1. The hide-rationale test. Write down your top three names on a piece of paper. Hide the rationale, the explanation, the brand story. Show the names to a stranger. Ask them, in one or two guesses, who they think the brand is for. If they can't guess in two tries, the rationale is doing the work the name should be doing. Cut.
2. The accessible-vocabulary test. Look at each name. Is every component word something the buyer already has in their mental library? Or would they need to look it up? Cleverness is in the connection between the words, not in the obscurity of the words themselves. If your customer needs Google to decode your name, the name is failing.
3. The phoneme map. Write each name out phoneme by phoneme. Count the consonants. Classify them: stop, fricative, liquid, nasal. Look at the vowels: front, mid, back. Does the sound profile match the positioning? A premium yoga brand should not sound like a kitchen knife. A surgical software product should not sound like a candle.
4. The shelf compare. Pull up the websites of your three closest competitors. Read your name in the same line as theirs. Does it sound like a clone of the category, or like a counter to it? If it blends in, it will lose. If it takes a clear phonetic or semantic stance against the others, it has a chance.
5. The pronunciation-collision test. Say your name out loud, fast, in a sentence: "I work at ___." Say it again at half speed. Now say it after a brand it competes with. If the two run together, if a customer hearing it in a podcast ad would mishear it as something else, or if it loses syllables when spoken naturally, you have a fluency problem the spec sheet won't catch. Founders skip this test because their own name has been said correctly in their own head ten thousand times. The buyer is hearing it for the first time, in noise, at speed.
These five tests will not get you a Stripe-grade name on the first pass. They will rule out the names that were going to fail anyway, and they will surface the one or two candidates worth pressure-testing further. That is the work. The rest is patience.
Frequently asked questions
What actually makes a brand name "work" scientifically?
A working brand name does three things simultaneously: it carries the right sound-symbolic associations for the category (so the phonemes don't fight the positioning), it processes fluently enough that the reader trusts it without effort, and it takes a strong enough phonetic or semantic stance to claim attention rather than blend in. The peer-reviewed literature on sound symbolism (Sapir 1929, Klink 2000, Yorkston and Menon 2004) and processing fluency (Reber, Schwarz, and Winkielman 2004, Alter and Oppenheimer 2006) is where the underlying principles come from.
Should a brand name be memorable or polarizing?
Polarizing, by a clear margin. Memorability is a downstream consequence of attention, and attention requires the name to take a stance. Berlyne's 1970 research on optimal incongruity showed that preference follows an inverted-U curve: names that are too familiar are boring, names that are too novel are overwhelming, and the peak sits in the polarizing middle. A name that everyone agrees is fine has, by definition, generated no attention worth remembering.
How long should a brand name be? Is there a syllable count rule?
Two to three syllables is the consumer-brand sweet spot. About 74% of successful consumer brands cluster there. Longer names tend to shorten over time in everyday speech (Coca-Cola becomes Coke, Federal Express becomes FedEx). The deeper rule is that the syllable count should match the phonetic energy of the positioning: a single hard syllable feels surgical (Stripe, Square), two soft syllables feel friendly (Olipop, Sonos), three syllables can feel premium or mystical (Mercury, Patagonia). Length is a tool, not a target.
Are AI brand name generators worth using?
They are useful for one narrow purpose: producing a high-volume longlist of made-up-sounding words you can use as a starting point for further refinement. They are not useful for the work that actually matters in naming. They have no theory of category-specific sound symbolism, no model of your competitive shelf, and no way to evaluate whether a name takes the right phonetic stance. Most of the output sounds vaguely tech-startup-y because that is what their training data rewards. If you use one, treat its output as raw material and apply the four Monday-morning tests above to whatever survives.
Is sound symbolism real or pseudoscience?
It is one of the most replicated findings in psycholinguistics. Sapir's original 1929 study on front-vowel and back-vowel size associations has been reproduced across age groups, cultures, and languages. The bouba/kiki effect (originally Köhler 1929, formalized by Ramachandran and Hubbard 2001) was validated in 2022 across 229 languages in Royal Society Open Science. The application to consumer brands has its own literature, anchored by Yorkston and Menon's 2004 Journal of Consumer Research paper on phonetic effects on consumer judgments. The strength of the effect varies by context, but the existence of the effect is not contested in the field.
How do I check if my brand name is trademark-safe?
Run a free preliminary search on the USPTO TESS database (Trademark Electronic Search System) for U.S. coverage, and the WIPO Global Brand Database for international. Search for your exact name and phonetically similar variations within the relevant trademark class (the goods or services class your business falls under). Once you have a finalist, hire a trademark attorney for a full clearance opinion before you spend money on a logo or domain. The free databases will catch the most obvious conflicts, but they won't catch the close phonetic neighbors that will still get your application rejected.
Brand Namers is an applied-linguistics naming methodology for founders. We're opening early access soon. Drop your email and we'll write you once when we open the doors.