How to name a startup: the complete guide

To name a startup, work in four passes: decide what the name needs to communicate, generate a large pool of candidates — fifty to a hundred is a realistic minimum, not overkill — cut the pool with a few ruthless filters (can people say it, spell it, and can you actually own it), then test the survivors on real humans before you register anything. The most useful principle in naming is that quality comes from volume: your fortieth idea is usually better than your fourth, because the first ideas you generate are the same first ideas everyone else generates. Everything below is a way of making those four passes concrete.

How do you brainstorm startup names?

Before generating anything, write one sentence describing what the name has to do. “Signal speed to developers.” “Feel calm and trustworthy for a finance app.” “Be pronounceable in five languages.” That sentence is your filter later, and it stops you from falling in love with a clever name that’s clever about the wrong thing.

Then generate in deliberate styles rather than staring at a blank page. Four approaches cover most of the territory:

Coined words. Invented names like Spotify, Klarna, or Zapier are blank canvases: nothing to unlearn, easy to trademark, and the exact-match domain is usually free. The trade-off is that you start at zero meaning and have to fill the name with your product. The reliable way to coin words that don’t sound like pharmaceutical brands is to build from meaningful fragments — inventing names from Latin and Greek roots walks through the method, and it’s how names like Novartis and Verizon were built.

Real-word combinations. Take two ordinary words and collide them: Facebook, Snowflake, MailChimp, DoorDash. These names arrive pre-loaded with meaning and are easy to spell because both halves already live in people’s heads. The risk is genericness — “CloudSync” describes a category, not a company — so aim for combinations with a little friction or surprise in them.

Metaphors. Instead of describing what the product does, name the feeling or the image it evokes. Amazon (vast), Oracle (knows things), Stripe (clean, minimal), Anchor (stability). Metaphors age well because they aren’t chained to your current feature set; a descriptive name like “PDFTools” dies the day you outgrow PDFs.

Foreign words and mythology. Nike, Samsara, Asana. These sit between coined and real: distinctive to your audience, but with roots you can tell a story about. Check pronunciation and unintended meanings in your target markets before committing.

Work each style separately for fifteen or twenty minutes and you’ll have a genuinely varied pool instead of forty variations of one idea. AI generators are excellent at exactly this fan-out if you brief them properly — vague prompts get vague names, so it’s worth learning how to prompt AI for name ideas rather than typing “name for my startup” and accepting whatever comes back.

What makes a startup name good?

Two qualities do most of the work: brandability and relevance. A brandable name is distinctive, short, easy to say out loud, and easy to spell after hearing it once — the classic “radio test.” A relevant name gives people at least a thread of connection to what you do, even if it’s only a mood. You rarely get to maximize both, and the central strategic decision in naming is where you sit on that spectrum. Brandable vs descriptive domain names covers the trade-off in depth, and what makes a name brandable breaks brandability into testable properties — our methodology shows how Domain Yoga scores them.

A quick gut check that catches most bad names: say the name aloud in the sentence “I work at ___” and imagine spelling it over the phone. If either feels awkward, the name will tax you a little every single day you own it.

How do you check if a name is available?

Availability is a stack, and you should check it in order of how hard each layer is to fix:

1. Trademark conflicts — the only layer that can legally force a rename (more below).
2. The domain — check your preferred extension plus the handful you’d accept.
3. Social handles and package registries — the usernames on the platforms that matter to your audience, plus npm, PyPI, crates.io, or the App Store if you’re shipping there.
4. Plain search results — Google the name. If page one is dominated by an established product, a common dictionary phrase, or something unsavory, you’ll be fighting that forever.

Checking names one at a time is where most people burn their evenings, because good names fail availability checks at a brutal rate — expect the large majority of a fifty-name pool to be taken or trademark-adjacent in some form. The sane workflow is to check your whole shortlist at once and let availability prune the pool for you; checking domain availability in bulk covers the tools and the workflow.

One reframe worth internalizing early: “the .com is taken” is not the end of the process, and treating it as a veto will cost you your best names. Nearly every short English word and most good two-word combinations were registered under .com years ago, often by squatters who will never build anything. Your real options are broader: a modern TLD that fits your product (Vercel ships on .app links, thousands of dev tools live on .dev and .io), a modifier (getdropbox.com was Dropbox’s home for years, and Notion still runs on notion.so alongside notion.com), or simply a different name from your pool. The .com matters most if your audience types URLs from memory — which, for a product discovered through search, links, and app stores, almost nobody does anymore.

Do you need to worry about trademarks?

Yes — before you register anything, not after. A trademark conflict is the one naming problem that compounds: the longer you build on a conflicting name, the more expensive the forced rename becomes, and “I didn’t know” is not a defense. The good news is that the basic screen takes under an hour: search the USPTO database (and the EUIPO if you’ll operate in Europe) for your name and close lookalikes, in the trademark classes that cover your actual product category. Trademark rights are scoped by category — Delta Air Lines and Delta faucets coexist fine — so a hit in an unrelated industry is usually survivable, while a similar-sounding name in your own category is a hard stop even if the spelling differs.

You don’t need to file a trademark on day one, but you do need to confirm nobody can make you stop using the name. Trademark basics before you register walks through the searches, what “confusingly similar” actually means, and when it’s worth paying a lawyer instead of self-serving.

How do you test a shortlist of names?

Once availability and trademarks have cut your pool to three-to-five finalists, resist the urge to just pick your favorite. You are the worst possible judge of your own names — you’ve stared at them too long, and you know what they’re supposed to mean. Cheap tests that produce real signal:

• The radio test, live. Say the name to five people, then ask them to type the URL. Every misspelling is a lost visitor forever.
• The recall test. Tell someone the name and what the product does; ask them for the name two days later. Names that survive this are rare and precious.
• The context test. Mock up the name in the places it will actually live: a landing page hero, a tweet, an app icon, an email signature. Names behave differently in context than on a list.
• A tiny smoke test. A one-page site or a small ad experiment per finalist can tell you which name people actually click — imperfect data, but data.

The full protocol, including how to keep the tests honest and how to weigh conflicting results, is in how to test a name before committing. Budget a week for this stage; it’s the cheapest insurance you’ll ever buy.

Which domain extension should you choose?

The TLD question generates more anxiety than it deserves. Search engines treat generic TLDs equally — .com, .dev, .app, and .yoga all rank on the merits of the site, not the extension. What actually differs is audience perception and price. A rough decision guide:

• Get the .com if it’s affordable — it remains the default assumption for mainstream consumer audiences, and it protects you from a competitor sitting on it.
• .io, .dev, .app, and .so are fully normalized for developer tools and startups; nobody in that audience blinks.
• Industry TLDs (.design, .studio, .finance, .yoga) work when the extension genuinely describes you — they add meaning instead of merely substituting for .com.
• Country-code TLDs are fine for clever names but check the registry’s rules and renewal pricing first; they’re run by national registries with their own policies.

The most fun option in the modern TLD era is making the extension part of the name itself — the way domain.yoga reads as “domain yoga,” or bit.ly spells “bitly” across the dot. Domain hacks explained covers when that cleverness pays off and when it backfires.

Whatever you pick, register the obvious misspellings and the .com later if it becomes cheap — but don’t let a $30,000 squatter price on the .com kill a great name when the .dev is $12.

What are the most common startup naming mistakes?

The same failures repeat across thousands of launches:

• Naming by committee. Consensus produces the least objectionable name, not the best one. Gather input widely, but let one person decide.
• Choosing the name that describes today’s product. You will pivot, expand, or reposition; a name welded to one feature becomes a liability. Burbn had to become Instagram.
• Ignoring pronunciation. If your own team says it two different ways, customers will say it zero ways — they’ll avoid saying it at all.
• Falling for cleverness over clarity. Deliberate misspellings (Lyft, Tumblr) worked for a few famous companies and quietly killed hundreds of others that lacked their marketing budgets. Every “we dropped a vowel” name pays a spelling tax forever.
• Skipping the trademark search because the domain was available. Domain registrars do not check trademarks for you.
• Waiting for the perfect name. A good-enough name shipped this month beats a perfect name shipped never. Names accrue meaning through use — “Google” and “Spotify” meant nothing until the products made them mean something.

When should you rename a startup?

Sometimes the right move is admitting the name is wrong. The legitimate triggers are few: a trademark conflict you can’t win, a name that actively confuses or embarrasses (users consistently can’t say it, spell it, or it means something unfortunate in a market you’ve entered), or a pivot so complete that the old name misleads. Twitter-scale rebrands aside, plenty of companies renamed early and thrived — Google was BackRub, Instagram was Burbn, Slack was born inside a company called Tiny Speck.

What doesn’t justify a rename: mild boredom with your own name, or a shinier domain becoming available. Renaming costs you accumulated recognition, links, and SEO equity, and the cost grows every month — which is exactly why the testing stage above exists. If you’re genuinely facing the decision, when and how to rename a project covers how to judge the trigger and how to execute the migration without losing your audience.

What’s the short version?

Define what the name must do. Generate far more candidates than feels reasonable, across coined words, combinations, metaphors, and borrowed roots. Filter hard on say-ability, spell-ability, trademarks, and availability. Test the finalists on strangers, not friends. Pick, register, and start pouring meaning into it — the product makes the name, far more than the name makes the product.

If you’d rather compress the brainstorming and availability passes into minutes instead of evenings, Domain Yoga generates name ideas in every style above and checks domain availability across TLDs as it goes — so the names you fall for are names you can actually have.