How to Brainstorm Brand Names

Why “let’s think of names” produces nothing

Put four people in a room with a whiteboard and the instruction to think of names, and you will get the same session every time. Someone suggests a compound of the two most obvious words. Someone else suggests a Greek god. A third person says they do not love either, and within ten minutes the group is discussing the merits of three candidates instead of producing thirty more.

Two things went wrong, and both are structural rather than personal. The group had no constraints, so every candidate was equally arguable, and it started judging before it had finished generating. Fix those two and the same four people will fill a page. The rest of this guide is that fix: methods supply the constraints, and the session format keeps judgement out until it is useful.

The seven methods

Each method is a different way of manufacturing candidates. Run at least four of them, since a hundred names from a single technique is one idea repeated rather than a wide search.

1. Root miningTake the concept at the heart of your product and dig for its roots in Latin, Greek, or any language you have a native speaker for. Roots give you syllables that feel meaningful without spelling anything out, and they are usually short. The trap is that a root which reads as elegant to you may read as a common word to somebody else, so every candidate from this method needs a cross-language pass before it goes anywhere near a shortlist.try: lumen, clarus, kairos, ferro, novus, altus
2. CompoundingFuse two ordinary words into one. This is the most productive method by volume and the easiest to run with a group, because everyone can play. The output tends to be long, so favour pairs where both halves are one syllable, and read each candidate aloud before you write it down: compounds that look fine can be unsayable.try: stackline, dropfield, quickvault, sunpath
3. TruncationCut a longer word or phrase down to its first syllables. Clipping is how a descriptive placeholder becomes a brandable name without losing its thread, and it usually leaves you with something short and ownable. Check that the remaining fragment does not accidentally spell something in another language, which is the one failure mode this method reliably produces.try: federal express to fedex, san francisco to cisco
4. AffixingAttach a prefix or suffix to a real word: get, try, use, hey at the front; ly, ify, io, hq, base at the back. It is fast, it makes taken words available again, and it is the most dated-looking method on this list, because every affix has a decade stamped on it. Use it to widen the pool, and expect most of what it produces to be cut.try: trystack, sortly, paperbase, hqgrid
5. Semantic fieldsPick a world with nothing to do with your product (birds, mountains, navigation, mythology, minerals, tools) and mine it systematically. This is where distinctive consumer names come from, because a borrowed word arrives with a picture attached and no competitors in your category. Work the whole field rather than the first three words you think of: the good ones are usually eight or nine entries down.try: kestrel, obsidian, meridian, anvil, cairn
6. Phonetic mutationTake a word you like and change how it sounds rather than how it is spelled: swap a vowel, double a consonant, move the stress. Mutating sound gives you something pronounceable and ownable at once. Mutating spelling alone (dropping vowels, swapping c for k) gives you a name people mishear and mistype forever, so avoid that variant of the trick.try: valor to valora, orbit to orbita, forge to forja
7. Borrowing far awayTake an ordinary word from a category unrelated to yours and use it straight. Done well it produces the most memorable names in existence, because the mismatch itself is the hook. Done casually it produces a name whose exact-match domain sold years ago, so treat availability as the first filter rather than an afterthought.try: rivet, harbour, compass, thicket, ember

Which methods suit you depends on where you want to sit between describing your product and abstracting away from it. That axis, and what each end costs, is in descriptive vs. abstract brand names. If you would rather start from a finished list than a blank page, there are fifty worked examples in brand name ideas for tech startups.

The session, in six steps

This runs in about two hours plus a night’s sleep. The order is the point: every step that generates comes before every step that judges.

1. Write the brief before anyone says a wordOne page: what the product does, who it is for, the three adjectives you want the name to carry, the markets you will sell in, and the things the name must NOT say. A brief that names its constraints is worth an hour of arguing later, and it is what turns a naming session into a search instead of a mood board.
2. Generate alone before you generate togetherGive everyone the brief and twenty minutes of silence with the seven methods. The research on group idea generation is consistent on this point: people produce more and more varied ideas working alone and pooling afterwards than they do in a free-for-all, where one loud voice sets the direction and everyone else edits. Pool the lists only when the timer stops.
3. Ban judgement until the target is hitAim for a hundred candidates before anyone is allowed to say “I do not like that”. Evaluation during generation is the single most common way naming sessions fail: the group converges on the first respectable option, and the twenty better ones never get written down. Volume first, taste later.
4. Cut to twenty by reading aloudRead the whole list out, one candidate at a time, with no discussion. Anything nobody can say cleanly, anything that needs spelling twice, and anything that makes the room wince gets struck immediately. This pass costs ten minutes and removes most of the list.
5. Cut to five with the checkable propertiesNow the criteria change from taste to fact. Domain availability, spelling, memorability, and how the word behaves in your other markets are things you can measure rather than debate. Score the twenty, keep the five with real gaps in their favour, and let the numbers do the arguing.
6. Sleep on the last five, then decideA day later, say each of the five out loud as if answering the phone at that company. Names that felt clever yesterday often sound effortful today, and the one you still say without hesitating is usually the answer. If two are genuinely tied, break the tie with a method rather than another meeting.

Step five is where this guide hands over. The full battery of evaluation is in how to test a brand name, and if you end up deadlocked between two finalists, the tiebreak ladder is in how to choose between two business names.

Where an AI generator actually helps

Generators are good at exactly one part of this: widening a pool that has gone stale. Ask for forty compounds in a semantic field and you will get forty in seconds, some of them better than the ones you were going to write. That is a real contribution to step three, and it costs you nothing but the discipline not to treat the output as a shortlist.

What a generator cannot do is tell you whether a name is any good, because the properties that decide that (a live domain lookup, an ambiguous spelling, a collision in a market you sell in) are facts about the world rather than patterns in text. Generate wide, score narrow. The longer argument is in why AI name generators fall short.

Four ways a naming session goes wrong

These are the failure modes worth watching for, because each one is invisible from inside the room.

• Falling in love before checkingThe most expensive mistake in naming is emotional investment in a candidate whose domain sold in 2011. Check ownability early, while a name is still one of a hundred, not after the team has been saying it for a fortnight.
• Naming the feature instead of the companyA brief written from today’s product produces a name that describes today’s product. If there is any chance the business changes shape, generate against the problem you solve rather than the thing you currently ship.
• Letting a thesaurus set the registerSynonym-hopping produces long, formal, interchangeable words: optimal, premier, dynamic, synergy. If the list starts looking like a corporate slide, switch methods entirely and mine a semantic field instead.
• Confusing volume with progressA hundred candidates from one method is not a wide search, it is the same idea a hundred times. Run at least four of the seven methods before you start cutting, and count how many distinct approaches survive to the shortlist.

A fifth, rarer trap: brainstorming a company name when the thing that actually needs naming is the product, or the other way round. If that question is still open, settle it first with naming a product vs. naming a company.

Frequently asked questions

How do you come up with a brand name?

Work from methods rather than inspiration. Write a one-page brief, then generate candidates with several construction techniques: mining word roots, compounding two words, truncating a longer phrase, adding a prefix or suffix, mining an unrelated semantic field, mutating the sound of a word you like, and borrowing a word from a distant category. Generate a hundred before you judge any of them.

How many brand name ideas should I generate?

Aim for around a hundred candidates from at least four different methods, then cut to twenty by reading them aloud and to five using checkable properties like domain availability and spelling. The number matters less than the ratio: most of what you generate should be thrown away, and a shortlist you never had to cut is a sign the search was too narrow.

Should I brainstorm brand names as a group or alone?

Both, in that order. Have each person generate alone against the same brief, then pool the lists. Research on group idea generation consistently finds that people produce more and more varied ideas working individually first, because in an open session one voice sets the direction and everyone else edits rather than invents.

Should I use an AI name generator to brainstorm?

Use it for volume, not for verdicts. A generator is a fast way to widen a pool that has gone stale, and it cannot check live domain availability, spelling ambiguity, or how a name behaves in another language. Generate wide with whatever tools help, then score the shortlist against checks that produce facts.

What do I do after the brainstorm?

Turn taste into evidence. Score the shortlist on domain availability, memorability, phonetics, and cross-language safety, then compare the survivors side by side. Namoly does that part free and explains every score. Trademark screening and social handles are separate steps that Namoly does not check, so plan for those before you commit.