We analyzed 5,000 DTC brand names. Only 27% are invented words.

Of 5,018 real direct-to-consumer brand names, only 27 percent are invented words. Nearly two thirds are built from words that already existed.

That number cuts against the most common piece of naming advice there is, which is to coin something new, a blank slate nobody owns yet, the next Kodak or Verizon. The shelf mostly did the opposite. BrandNames read those 5,018 names, spread across nine consumer categories and ninety-one subcategories, and measured each one down to its sounds and its structure. The instinct to invent a word turns out to be the minority move, and it is the first of several places where what the market actually built parts ways with what the guides recommend.

This is not a controlled experiment. There is no pile of failed names to compare against, so nothing here proves that one pattern wins and another loses. It is a record of what the market reached for, thousands of times over, once the naming was done and a brand had to live on a shelf.

The corpus is the real direct-to-consumer brands in the BrandNames reference set, spanning food and drink, beauty, apparel, home and garden, accessories, health and fitness, tech, kids, and pets. Length and word count come straight off each name. Name type is a reading of what a name is doing, an invented word set against a borrowed one. The sound figures map each name's phonemes onto the published sound-symbolism research, so the text flags any figure that rests on that reading rather than a hard count.

Most names are borrowed, not invented

Start with the split that surprises founders most. Of the 5,018 names, just over a quarter are coined, words that did not exist until a brand made them. Olipop, Oatly, Poppi, Beis, Solgaard. Nearly two thirds are ordinary words the language already had. The small remainder, under one name in ten, are founder or family names.

Coining a word can work, but it is the rarer path, and it runs into two facts at once.

The first is memory. When Lerman and Garbarino compared real-word brand names against invented nonwords in Psychology and Marketing in 2002, the real words were recalled more reliably. A made-up word has to teach the brain a new sound before it can mean anything, while a borrowed word arrives already known. That is a head start most new brands would rather not give up.

The second fact is supply. The obvious real words are gone. Beebe and Fromer studied 6.7 million United States trademark filings for the Harvard Law Review in 2018 and found that 81 percent of the thousand most common words in American English are already registered as single-word marks. You cannot name a coffee brand Coffee or a sleep brand Sleep. Someone owns the easy ones.

Borrow a real word from a category that has nothing to do with the product.

So the names that get built thread between the two. Not an invented word, and not the obvious one. They borrow a real word from a domain that has nothing to do with the product. Allbirds sells shoes. Osprey and Antler sell luggage. Matador sells gear. Each one took a word the buyer already owned and aimed it somewhere unexpected. That distance does a second job. A word pulled from far enough away, Osprey for a backpack, sits clear of the things it literally describes, which is part of what lets a brand register it and own it outright. Read by mechanic instead of by the six buckets in the table, this kind of borrowing is the most common move of all, an existing word aimed at a product it does not describe, tagged on roughly one name in four. The work is in the aim, not the invention.

The typical name is eight letters, and half are one word

For all the energy that goes into naming, the result is usually small. The median name in the corpus is eight letters long. The single most common length is five, and half of all the names land inside a tight four-to-eight-letter band. Just over half, 53 percent, are a single word, and only about one in ten stretches to three words or more.

The typical name is short. Each extra word spends a little of the speed that made the first one work.

Single-word names are smaller still, a median of six letters. These are the names that read in one glance and survive being said across a noisy room. Greats. Knack. Rains. Studs. Away.

The pull in the other direction is real and worth naming. When a single word feels too thin, the reflex is to add a second word to explain it, then a descriptor to qualify the second. The corpus shows where that ends. The longest names run past twenty letters, almost all of them a brand plus a category plus a hedge, like The London Grooming Company or Medicine Mama's Apothecary. Each added word buys a little clarity and spends a little of the speed that made the first word work.

There is no rule here, only a center of gravity. If your shortlist sits well outside eight letters and one or two words, the names on it are not wrong, but they are unusual, and unusual should be a choice you made on purpose rather than one that crept in a descriptor at a time.

The shelf sounds light, not bold

Names carry a feeling before they carry a meaning. The sounds in a word arrive with associations already attached, fast or slow, heavy or light, and that reflex has a century of research behind it. Run every name in the corpus through that read and one association dominates.

Light is the single most common sound a direct-to-consumer name carries, the strongest note in 31 percent of them. Add the names that read smooth or fast and nearly two thirds of the shelf lands in the same gentle, quick register. The heavy and forceful sounds that founders reach for when they want a name to feel serious are the minority, together under a sixth of the corpus.

Most names sound light. The heavy, forceful ones are the minority.

This is a property of the read, not a law of the universe, and the direction is hard to miss. It also fits the goods. Most direct-to-consumer categories sell softness and ease, a better night's sleep, a gentler cleanser, a snack that feels like a treat. The names sound the way the products want to feel. The lesson for a shortlist is not to chase light for its own sake. It is to check that the sound a name makes points the same way as the thing it sells, because a heavy, blunt name on a soft, easy product works against the very thing it is meant to move.

One name in three opens on a hard sound

One lever in the corpus looks deliberate. A third of the names, 34 percent, begin on a hard plosive, one of the sounds b, d, g, p, t, or k, made by stopping the air and releasing it. Bombas. Poppi. Casper. Glossier.

A third of names open on a hard sound that lands clean and sticks.

Those openings hit sharp and land cleanly, which is part of why they stick. Roughly a third of names begin this way, more than any other kind of opening sound. Names that open on a soft hiss, the s and z and sh sounds, are far rarer, under one in ten.

A hard opening is not mandatory, and plenty of strong names start soft. But when two names on a shortlist are otherwise close, the one that opens on a plosive holds a small, free advantage in how easily it lodges, and it costs nothing to prefer it. The deeper read on why, and how to apply it to a real shortlist, is in the guide on how to name a brand.

The flashy tricks almost never get used

Naming lore loves a trick. The palindrome that reads the same backward, like Sonos. The doubled syllable, like a Lulu or a Toto. The alliteration that chimes, like Le Labo or Blu Dot. They are satisfying to spot, and they are close to absent on a real shelf.

Alliteration is the most common of the three, and even that turns up on only about one name in ten. Palindromes account for about a third of a percent of the corpus, seventeen names out of more than five thousand. Repeated-syllable names are just as scarce, another seventeen. A founder who anchors a shortlist on finding the perfect palindrome is hunting for a pattern that almost no built brand actually used.

The reason is not that the tricks fail. Sonos is a fine name. It is that they are constraints, and a constraint throws out far more candidates than it keeps. Demanding a palindrome rules out nearly every available word in service of a flourish most buyers will never consciously notice. The names that got built mostly spent their effort on sound and sayability instead, and let the symmetry go.

Most names are easy to say, but a fifth fight it

Underneath all of this sits the plainest test of all, whether a stranger can say the name on the first try. By the corpus read, most pass. Six in ten alternate their consonants and vowels cleanly, the rhythm that lets Banana or Casper roll out without a stumble. Only about 7 percent are genuinely cluster-heavy.

But more than one in five names, 22 percent, carry a run of three or more consonants with no vowel to break it. Look at which names those are and a pattern jumps out. BKR. JBL. ASRV. CDLP. They are mostly initials and stylized contractions, names built to look compact in a logo rather than to be spoken aloud. That is a real trade, and sometimes the right one for a brand that lives mostly as a mark on a product. But it is a trade, paid in the half-second a new buyer spends working out how to say the thing, and worth making on purpose rather than by default.

What the shelf is telling you

The advice points at ambition, the invented word and the striking trick. The 5,018 names that got built point the other way, at borrowed words that stay short and sound light. None of that is hard to state. It is hard to do well, because the good borrowed words are the ones nobody else thought to borrow, and finding them is the part a rulebook cannot hand you.

That is the harder half of naming, and it is where a shortlist is won or lost. The patterns above are the floor the market already rewards. What lifts one name above the rest is the specific word it borrows, and whether the sound of that word points toward the thing you sell or away from it.

Frequently asked questions

Should a brand name be a real word or an invented one? Usually a real word. Invented words are 27 percent of the corpus against nearly two thirds built from existing ones, and real words recall more reliably (Lerman and Garbarino 2002). The complication is supply. With 81 percent of the thousand most common English words already trademarked (Beebe and Fromer 2018), the workable real-word names are the ones borrowed from an unrelated category, not the obvious ones.

How long should a brand name be? Shorter than most shortlists assume. The median is eight letters, the most common length is five, and just over half are a single word. Longer names are not penalized in the data, but they are uncommon, and the extra length usually arrives as an explanatory word rather than a longer root.

BrandNames reads founders' real shortlists the way it read these 5,018 names, against the sounds and the patterns that built brands actually share. The doors open soon. Drop your email to get one note when they do.