What makes an invented name easy to say and spell
10 min readUpdated
Read these two aloud: 'Velora' and 'Vlkrth'. Both are invented. Both are almost certainly available. One of them you could tell somebody over the phone.
The difference is not taste, and it is not luck. Pronounceability follows patterns strict enough to write down, which is worth understanding whether you are naming something by hand or judging what a generator hands you.
Spelling follows its own patterns, and they are the ones people forget. A name can be perfectly easy to say and still arrive misspelled every time somebody hears it, which costs you exactly as much.
Vowels set the rhythm
English syllables are built around vowels. Every syllable needs one, and the spacing of vowels through a word is what gives it rhythm.
A useful measure is the vowel ratio: vowels divided by total letters. Real English words mostly land between 30% and 50%. Below that you get consonant sludge — 'Strndth'. Above it you get baby talk — 'Aiaeoa'. Names in the comfortable band read as words even when they are invented.
| Name | Vowel ratio | Reads as |
|---|---|---|
| Velora | 50% | a word |
| Stripe | 33% | a word |
| Vlkrth | 0% | a password |
| Aeiouia | 100% | a mistake |
Ratio alone is not enough — the vowels have to be distributed, not clumped. 'Aeiobrt' is 43% vowels and unreadable, because all of them are at the front.
Consonant runs, and why context decides
Three consonants in a row is where most invented names fail. But not always, and the exception matters more than the rule.
English is full of three-consonant runs nobody notices: 'street', 'sprint', 'thrive', 'functionality'. Your ear accepts them because the words are familiar and the specific combinations are ones English uses.
The same run inside an invented name is much harder work. Two things differ: the word is not familiar, so you have no stored pronunciation, and the run may be a combination English never uses. 'Nct' is fine inside 'functionality' and hopeless in 'Nctara'.
This is why a good check is not 'reject three consonants'. It is: trust runs inside a real word the user supplied, and be strict about runs that appear at a join or in invented material.
The join is the weak point
Most constructed names are a stem plus something. The place they break is where the parts meet.
Take 'place' and add 'ery': 'placery' joins cleanly, consonant meeting vowel. Add 'ry' to 'plac' instead and you get 'placry', where 'cr' at the seam gives you a stumble.
Three failures cover almost every unusable join:
- Vowel collision — 'a' meeting 'e' gives you 'aeventi'. English has vowel pairs, but they are learned pairs like 'ea' and 'ou'. Arbitrary ones read as typos.
- Consonant pile-up — three or more consonants meeting across a join, none of them a cluster English opens or closes words with.
- Accidental doubling — 'k' meeting 'k'. Some doublings are natural ('ll', 'ss', 'tt'), most are not.
A name can be perfect on both sides of the join and unusable because of it. This is the single most common way generated names fail, and it is why where you add the letters is a decision worth making deliberately.
Endings carry more weight than beginnings
The last two letters do a disproportionate amount of work. They tell the listener what kind of thing the name is, before they know anything about you.
- '-a' and '-o' feel open and product-like: Vercel, Figma, Notion sit in this family.
- '-ly' and '-ify' read as tools, and read as 2013. They are not wrong, they are dated.
- '-r' and '-x' feel technical and clipped.
- '-ery' and '-ory' feel like a place or an institution.
None of these is better. But they are not neutral either, and picking an ending is a decision about what your name implies whether or not you make it deliberately.
Spelling: the decisions you are making for the listener
Being easy to say and being easy to spell are different properties, and the second one is where most invented names quietly leak. A listener who hears your name has to reconstruct a spelling from sound alone. Every point in the word where two spellings are equally plausible is a coin flip you have handed them.
So the useful question is not 'is this pronounceable' but 'how many spelling decisions does this force'. A name with none is typed correctly first time. A name with three will arrive in four different forms, and one of those forms is somebody else's domain.
The sounds with more than one spelling
A handful of sounds account for nearly all of it. An f might be f or ph. A k might be c, k or ck. An s sound might be s, c or z. A long e might be e, ea, ee or ie. A doubled consonant might be doubled or not, and a listener has no way to tell.
Silent letters are the worst of the set, because they are unrecoverable from sound — nobody hears the h in a name you spell with one. If a candidate contains a letter that does no audible work, it is a letter your customers will leave out.
When an unusual spelling earns its place
Sometimes it does. An unconventional spelling can make a name distinctive, ownable and easier to trademark, and there are famous companies built on one. The test is whether the spelling delivers something.
The failure case is specific and common: the spelling exists only because the normal spelling's domain was taken. That is not a branding decision, it is a workaround wearing one, and it will be explained on every phone call forever. Before accepting it, compare it against a genuinely new word — the invented alternative is usually both cleaner and more available.
Budget one surprise, not five
Distinctiveness comes from breaking a pattern, so a name with no surprises at all tends to be generic. The mistake is treating that as permission to break several.
An unexpected opening consonant is fine. An unusual vowel is fine. A silent letter is defensible. An ending nobody uses is defensible. Put all four in one word and you have not made a distinctive name, you have made a puzzle — and each element is individually arguable, which is exactly why this happens in real naming sessions.
The pattern behind most strong invented names is familiar phonetics plus one deliberate oddity. Decide which single element is doing the distinguishing, and make everything else predictable.
A name should work before the story is told
There is a particular kind of name that only makes sense when explained: it is a Latin root joined to a Japanese word, it is the founders' initials rearranged, it means something in a language nobody in the market speaks. The explanation is genuinely interesting, and no customer will ever hear it.
The origin story is a nice thing to have and a bad thing to depend on. Judge every candidate as a stranger meets it — cold, without context, on its own. If it needs the story to work, it does not work.
The opposite is also true and easier to fix. A bare invented word often feels strange in a document full of other candidates and perfectly normal next to a product description. Test finalists in a realistic sentence, headline and icon before rejecting them for feeling unfamiliar — unfamiliar is what every new brand name is.
Check what the sound already means
An invented word is invented in your language, not in every language. A sound you assembled from scratch may already be a common surname, a place, a slang term, a medication, or an ordinary noun in a market you intend to sell to.
None of that is necessarily disqualifying — plenty of brands share a string with something unrelated. But it decides what the name communicates, and it is better to know before you commit than to discover it from a customer. Search the bare word, then search it alongside your sector, then translate it in two or three large languages.
Length is about effort, not characters
Rules like 'brands should be six letters' are confident and useless. Businesses succeed at every length. The question worth asking is how much work the name costs to say, remember, type and display — and character count is a poor proxy for all four.
Shorter is not automatically better. Four-letter names are memorable and essentially all registered; the ones still free tend to be unpronounceable. Six to nine letters, two or three syllables, is where invented names work best: long enough to be distinctive and available, short enough to say in one breath and fit in a logo.
Syllables matter more than letters. 'Zapier' is six letters and two syllables and easy. 'Schmidt' is seven letters and one syllable and hard, because the work is front-loaded into a consonant cluster. Eight letters with two obvious beats will beat five letters that have to be spelled out, every time.
Compression has a floor. Removing letters to chase a shorter name eventually destroys the pronunciation or the meaning, and a five-letter name customers have to ask about is cognitively longer than the eight-letter one it replaced. Edit length; do not worship it. Stop at the shortest version that a stranger still reads correctly.
Test it in the small spaces
Length problems are visual as often as they are spoken, and they only appear at real sizes. Put each finalist into a mobile header, an app icon, a browser tab where it will be truncated, an email address, and a social avatar. Something that reads beautifully at 48 points can be illegible at 16, and an app icon holds about four characters.
The brand and the domain are separate lengths
These influence each other and they are not the same decision. A slightly longer brand whose exact-match .com is free is usually cleaner than a short brand paired with a modifier — 'usevelo.com' is shorter than 'velora.com' by nothing that matters and worse by a whole instruction.
The test that decides
Every rule above is a proxy for one real test: say the name out loud to somebody who has never seen it written, and have them type it.
Run it properly. Three to five people, one at a time, no showing them the spelling first, and have each of them write down what they heard. Then run it in reverse — show the written name to somebody new and ask them to say it. A name that survives both directions is a name that survives a recommendation.
Say the whole address, too, not just the brand. 'Velora dot com' and 'get Velora dot com' are different tests, and the second one has a modifier in it that somebody has to remember.
Then treat the mistakes as data rather than as a verdict. If four out of five people write the same 'wrong' vowel, that spelling is the one your customers find natural, and it is worth checking whether it is available. If listeners disagree about the opening sound, the opening is the part to change. Most pronunciation problems are fixable with one edit, and the test tells you which edit.
If they get it right first time, the name works. If you find yourself saying 'with a y' or 'one word', you have found a cost you will pay in every conversation, every podcast mention, and every business card, for as long as the company exists.
A name you have to spell out is a name that leaks traffic forever.
In short
Sayable names keep vowels spread through the word, avoid consonant pile-ups at joins, and end on something that sounds like a product. Spellable names force no guesses — no silent letters, no sound with two plausible spellings, one deliberate surprise at most. The phone test, run in both directions, settles anything the rules leave open.
Frequently asked questions
Can an invented name be genuinely easy to spell?
- Yes, and the ones that are follow English spelling conventions rather than fighting them. Novelty and clarity are not opposites: the novelty can live in the combination of sounds while every individual sound keeps its most obvious spelling.
How many syllables should an invented name have?
- Two or three, in almost every case. One syllable is nearly always registered already. Four or more starts to get shortened by customers, and the shortening is usually not the part you would have chosen.
Is a shorter name better?
- Only up to the point where it stays readable. Pronunciation, memorability, distinctiveness and fit all outrank character count, and a short name that has to be explained is longer in every way that matters.
How many finalists should I take into the spoken tests?
- Three to five. Fewer than three and you are not comparing anything; more than five and the people helping you stop distinguishing between them.
Sources
These support factual background only. The guide itself is original writing, and nothing here should be read as legal advice.
Related guides
Brand Naming
50 ways to turn an ordinary word into a brand name
Fifty edits — openings, endings, vowels, consonants, expansions, compressions, semantic and phonetic moves — and how to keep an invented name from sounding random.
9 min read
Domain Names
Prefix, suffix, or change the word itself
The four ways to reshape a word you already like — add to the end, add to the front, add to both, or alter the letters already there — and when each one works.
8 min read
Domain Names
How to choose between two names you both like
Six tests that separate two candidates when intuition keeps looping — including what actually makes a name memorable, and how to weigh a great brand against a compromised address.
9 min read
Domain Names
Every good .com is taken. Here is what actually works.
Roughly every dictionary word is registered. The way out is not a cleverer search — it is changing the shape of what you search for.
12 min read
Domain Names
From shortlist to registered, without regret
What to check between finding an available name and paying for it: the spoken tests, the searches, trademarks, handles, history, and the settings that matter at checkout.
9 min read