Naming a business or product is the rare decision that is both permanent-ish and made under maximum ignorance: you choose it before customers exist, and you’ll say it ten thousand times after. Most naming advice makes it worse by treating the choice as pure inspiration (“brainstorm until it feels right”) or pure paranoia (“check 40 databases before you love anything”). What actually works is a process — one that generates widely, filters mechanically, tests cheaply, and clears the legal path before emotional attachment sets in. Here’s that process, field-tested across startups, products, and rebrands.
Before Generating: Write the Brief (30 minutes)
Names are judged against criteria, so set the criteria first — or the loudest voice in the room will set them mid-argument. One page:
- What the name must do: lead with the positioning. Should it describe the offer (clear but generic), suggest a quality (evocative), or be an empty vessel you’ll fill with meaning (distinctive but expensive to establish)?
- Where it must work: spoken on podcasts and sales calls? Typed as a domain? Pronounceable in the languages of your target markets? Each context kills different candidates.
- What it must survive: the company’s likely evolution. “QuickBooks” survived expanding beyond bookkeeping; “Burbn” had to die for Instagram to live. Avoid names that hard-code today’s single feature.
- Hard constraints: budget for the domain, categories you’ll trademark in, names that are simply off-limits (too close to competitors, unfortunate meanings).
Generate Wide: 100+ Candidates (one week, background process)
Quantity is the whole game at this stage — the fatal error is falling in love with candidate #4 of 12. Work the five standard name types deliberately:
- Descriptive: says what it is (General Motors, The Home Depot). Clear, instantly understood, hard to trademark, easy to confuse with rivals.
- Suggestive: evokes the benefit without literalism (Slack, Pinterest, Shopify). The sweet spot for most startups — meaning plus ownability.
- Invented: coined words (Kodak, Xerox, Zillow). Maximum trademark strength and domain availability, maximum spend required to attach meaning.
- Arbitrary: real words from elsewhere (Apple, Amazon, Caterpillar). Distinctive and ownable in your category; demands confidence.
- Founder/place: (Ford, Patagonia). Automatic authenticity, harder to sell or scale past the person.
Generation tactics that reliably produce non-obvious candidates: mine the vocabulary of your customers’ goal (what words live around the outcome, not your product?); raid adjacent languages, mythology, and old technical jargon; compound two short words (Face+book, Snow+flake); truncate and mutate (Cisco from San Francisco); and use AI brainstorming for volume — it’s excellent at “50 more like these three” once you have seeds. Keep a running list; the best candidates usually appear on day four, after the obvious layer is exhausted.
Filter Mechanically: 100 → 10 (one afternoon)
Now apply the cheap, unarguable tests in order — each one deletes candidates without debate:
Test the Finalists Like a Marketer, Not a Poll (one week)
Skip the “which do you like?” survey — people vote for the familiar and comfortable, which selects for forgettable. Test what actually matters:
- Recall: show ten people the shortlist once, in passing, and ask a day later which they remember. Memory is the name’s real job.
- Assumption mapping: “what do you think a company with this name sells, and what would it charge?” — reveals whether each name pulls perception toward or away from your positioning.
- The live mock: put each finalist on a fake homepage header and a fake ad. Names live in layouts, not lists; several strong candidates die and one dark horse usually wins at this step.
- The repetition test: have your team use each finalist in real sentences for two days — “welcome to X,” “thanks for calling X.” Awkwardness that surveys miss surfaces in the mouth immediately.
Clear the Path Before You Commit (days 1–10, in parallel)
For the final two or three candidates, run the legal and practical clearance before announcing anything, in escalating order of cost: a knockout trademark search in your national database (free, an hour, catches obvious blocks); domain and social-handle availability — remembering that the exact .com is negotiable (“get” / “try” / “use” prefixes and alternate TLDs are now normal, though a name whose every variant is taken is telling you something); and finally, for the winner, a professional trademark search and registration in your core classes and markets. The few hundred to few thousand this costs is the cheapest insurance in branding — renaming a live business later costs orders of magnitude more in money and accumulated recognition.
Decide, Then Stop Relitigating
Choose with the brief, not the room’s mood: score the finalists against the original criteria, let the founder or brand owner make the final call (naming by committee averages its way to beige), and then — this is the discipline almost everyone skips — declare the decision closed. Every name feels wrong for the first month; “Google” sounded absurd until it didn’t. A name becomes good through consistent use, accumulated associations, and the quality of the thing it points to. Six months of confident, consistent deployment will do more for the name than another quarter of second-guessing ever could.
The meta-lesson of naming: the name matters less than the process’s failure modes. A mediocre name cleared, tested, and used consistently for years quietly wins; a brilliant name that’s unspellable, legally encumbered, or relitigated every quarter quietly bleeds. Run the process, make the call, and start pouring meaning into it. (For how a name fits into the larger system — positioning, identity, story, and distinctive assets — see our complete guide to brand building.)
