Naming is the most permanent decision in marketing and the one people spend the least structured time on. Logos get redesigned, positioning gets revised, websites get rebuilt. The name survives all of it, printed on everything, spoken in every introduction, typed into every search bar. I have named companies, products, and campaigns of my own and for clients, and I have developed strong opinions about how to do this without either agonizing forever or grabbing the first clever pun that survives a domain search.
Here is the process, including the filters most people skip and regret skipping.
First, Know What a Name Has to Do (and What It Does Not)
Founders overload names. They want the name to explain the product, signal the values, encode the origin story, and be poetic. A name cannot carry all that, and trying makes names worse. In my view a name has exactly four jobs:
1. Be sayable and spellable on first contact. If people hear it at a dinner and cannot find it later, everything else is irrelevant.
2. Be ownable. Legally, digitally, and in search results.
3. Be empty enough to fill. The meaning comes from what you build; the name is the container. Some of the best brand names on earth meant nothing on day one.
4. Not fight the positioning. The name does not need to state what you do, but it must not contradict it. A playful name on a trust-critical service creates friction every single day, because naming is downstream of positioning, not a substitute for it.
Everything beyond those four jobs is bonus, not requirement.
Generation: Volume First, Judgment Later
The worst naming sessions mix generation and evaluation: someone proposes, someone critiques, and after an hour you have four bruised ideas and no list. I separate the phases completely.
For generation, I work through territories rather than free-associating:
- Descriptive territory: what it literally is or does.
- Metaphor territory: things that share a quality with the brand (speed, warmth, precision, origin).
- Invented territory: constructed words, blends, truncations, borrowings from languages connected to the brand's story.
- Founder and place territory: names drawn from real people, real geography, real history. Growing up between cultures gave me a bias here: names rooted in a real place or heritage carry texture that invented words rarely match, and they come with a built-in story.
Target: a hundred candidates before judging any. It sounds excessive. It is the only way to escape the gravitational pull of your first three ideas.
The Filter Gauntlet: Where Names Earn Survival
Now judgment, applied as sequential filters so weak names die cheap.
Filter one: the radio test
Say each name aloud to someone who has never seen it spelled, then ask them to type it. In 2026 I extend this: say it to a voice assistant and see what it transcribes, because an increasing share of brand discovery happens through spoken queries and AI intermediaries. A name that machines mishear is a name that leaks customers to whatever the machine heard instead.
Filter two: the search landscape
Search every surviving candidate and study what already owns that word. You are not just checking for competitors; you are checking the semantic neighborhood you would be moving into. Fighting an established meaning for your own name is the hardest version of the owning your name problem, and I have watched brands spend years losing that fight. Distinctive beats descriptive here: an unusual name can own its results in months, while a generic name never will.
Filter three: entity cleanliness
The 2026-specific filter. Ask the major AI assistants about each candidate name. If the name collides with an existing entity the models know well, your brand will spend years being confused with it in exactly the surfaces where entity clarity now decides visibility. A clean entity space, a name the machines have no prior about, is the naming equivalent of buying land before the road is built.
Filter four: legal and practical ownership
Trademark screening in your actual classes and markets, sensible domain availability (the exact .com is nice, a clean modifier is fine, a compromise that requires explaining is not), and handle availability where you will actually operate. Do the professional trademark work before emotional attachment forms, not after.
Filter five: the languages of your future
Check meaning and connotation in every language your customers speak, plus the ones your expansion plans imply. Working across Caribbean and international markets taught me this filter is not theoretical: a name can be neutral in English and unfortunate in Spanish, Dutch, or Papiamentu, and you want to laugh about that in a naming session rather than in a market launch.
Choosing: Kill the Vote, Run the Trial
Committees choose safe names; that is what committees are for. Instead of a vote, I run the two or three finalists through a live trial: use each name in real sentences for a week. Mock up the email signature, say "thanks for calling [name]" out loud, write the one-line introduction, draft the headline of the launch post. The right name starts feeling inevitable in use. The clever-on-the-whiteboard name usually starts feeling like a costume.
One warning from experience: do not seek broad feedback on names. Everyone has an opinion, all opinions anchor on familiarity, and crowds reliably sand the distinctive candidate down to the forgettable one. Small trial, real usage, decisive owner.
After the Choice: The Name Is a Commitment Device
Once chosen, the work inverts: stop evaluating and start investing. Spell it one way forever, define it with consistent structured data and profiles, and pour meaning into the container through everything you build. Names appreciate the way trust does: slowly, through repetition and kept promises, until one day the word that meant nothing is the most valuable asset you own.
Choose a word you can keep saying for twenty years. Then spend twenty years making it mean something.