How to Name a Business: A Process That Ends in a Decision

Most naming sessions fail the same way. Someone opens a document, types “name ideas” at the top, everyone shouts words for forty minutes, and three weeks later the company still has no name. The problem is not a lack of creativity. It is that generating and judging are being done at the same time, by the same people, with no agreed standard for what a good answer looks like.

Here is a process that separates those jobs and ends with a name you can register.

Step 1: Write a one-page brief before you generate anything

You cannot judge a name without knowing what it has to do. Write down five things:

• What you sell, in six words. Not the mission statement. The thing a customer pays for.
• Who says the name out loud. Customers to each other? A receptionist answering a phone? Only a search engine? This decides how much you should care about spelling.
• Where the name lives. A .com, an app icon, a truck door, a shelf label, a business card. Each has a length ceiling.
• What it must not sound like. List your three closest competitors. Your name should not be a fourth variation on the same idea.
• Markets and languages. If you plan to sell in Spain or Brazil or India, say so now, not after you print the signage.

This page is your scorecard later. Without it, “I like it” is the only criterion, and the loudest person wins.

Step 2: Decide what type of name you want

Name types are not a matter of taste. They have measurably different costs and different legal strength. The naming firm Igor, whose naming guide is the most cited practitioner framework in the field, splits names into four categories.

Type	Examples	Strength	Cost
Functional / descriptive	Trans World Airlines, Digital Equipment, FindWhat	Instantly explains itself	Weak trademark, easy to copy, forgettable
Invented from Greek or Latin roots	Agilent, Aquent, Novartis	Clean trademark, easy domain	“Emotionally void” without a huge ad budget
Invented and poetic	Google, Kleenex, Oreo, Snapple	Memorable, fun to say, strong trademark	Harder to sell internally
Experiential	Explorer, Navigator, Safari	Intuitive, maps to the user’s experience	Overused, crowded, hard to register
Evocative	Apple, Virgin, Yahoo	Biggest differentiator, brand bigger than the product	Easiest to get wrong if positioning is fuzzy

Igor’s blunt point about descriptive company names is worth repeating: a descriptive name is asked to perform one task, explaining what you do, and that “needlessly taxes a marketing and advertising budget.” Descriptive names work well for products inside an established brand (Microsoft Word, Apple Notes). They work badly as the company name.

The poetic invented category is where blended names live. Snapple is literally snappy plus apple. Groupon is group plus coupon. Verizon is Latin veritas plus horizon. Accenture is “accent on the future,” submitted by an employee in Oslo during an internal competition. If you want a name that is fun to say and clean to register, this is the category to aim at.

Step 3: Generate 50 candidates, and do not judge any of them

Fifty is the number because the first ten are always the obvious ones and the good ideas usually sit between twenty and forty. Use several techniques so the list is not fifty versions of one idea:

Word pairs. List 15 words about what you do, 15 about how it feels, and 15 from an unrelated world you like (sailing, geology, baking, old maps). Pair across columns.

Blends. Take two of those words and fuse them. Front of word one plus back of word two is the pattern behind motel, Instagram, and Frappuccino. Our portmanteau generator runs the linguistic version of this, and the business name generator applies brand-oriented filters like length and pronounceability on top.

Clipped compounds. Cut both words short and join them: Microsoft (microcomputer, software), Netflix (net, flicks), Amtrak (American, track), Velcro (velours, crochet). This pattern produces more usable brand names than strict blending does, because the pieces stay recognizable.

Productive suffixes. Add -ly, -ify, -io, -ia, -ix to a root: Shopify, Calendly, Grammarly. Cheap, and now very common, so it will not differentiate you on its own.

Real words, used arbitrarily. Apple for computers. Camel for cigarettes. Legally the strongest category after coined words, and free to generate: open a book, take the nouns.

If you want to widen the input side rather than the output side, the word mixer will throw combinations at you faster than a whiteboard will.

Step 4: Cut 50 down to 5 with three tests

Now switch modes. Judging is a different activity, and it happens after generating, not during.

The say test. Write each name on a card and hand it to five people who have never seen it. Ask them to read it aloud. Pass bar: at least four of the five say it the same way. Any name that produces two pronunciations is a name your customers will argue about forever.

The spell test, also called the radio test. Say the name out loud to five people and ask them to type it as a web address. The practitioner rule of thumb is simple: if three of the five misspell it, reconsider. This is where clever spellings die. Consistent phonetic substitutions like Lyft for lift survive. Arbitrary ones do not.

The search test. Type the name into Google in quotes. If page one is full of other companies, a famous person, or a medical condition, you will spend years fighting for your own name. Then search it in the app store if you are building an app.

Two more filters worth applying at this stage:

• Length. An analysis of more than 400 best-selling brands put the mean at 10 characters, median 9. The top 100 websites average 6 characters in their domain. Around 29 percent of top brands are exactly two syllables. Aim for 5 to 10 characters and 2 to 3 syllables, and treat anything past 12 characters as a warning.
• Meaning in other languages. Mitsubishi renamed the Pajero to Montero for Spanish-speaking markets because pajero is vulgar Spanish slang. Ford renamed the Pinto to Corcel in Brazil for the same class of reason. If your brief lists markets, get one native speaker per market to react to the shortlist. It takes ten minutes and it is the cheapest insurance you will ever buy.

Step 5: Check availability before you fall in love

Do this on all five finalists, not just your favorite, because at least two of them will fail.

Check	Where	What “clear” means
Trademark	USPTO search, EUIPO, WIPO Global Brand Database	No live mark for a similar name in your class
Domain	Any registrar, or our domain name generator for variants	The exact match, ideally .com
Social handles	Instagram, X, TikTok, LinkedIn, YouTube, manually	Same handle across the ones you will use
Business registry	Your state, Companies House, or national equivalent	Name not already registered in your jurisdiction
Plain search	Google, app stores, GitHub	No confusing collision on page one

A note on trademark strength, because it interacts with everything above. US trademark law ranks names on a spectrum: generic (never registrable), descriptive (only with proof of acquired meaning), suggestive, arbitrary, and fanciful. Suggestive, arbitrary, and fanciful are inherently distinctive and can be registered immediately. A blend of two descriptive words usually lands in “suggestive,” sometimes in “fanciful” if the result reads as a coined string. That gives you the interesting rule of thumb: the more your portmanteau hides its ingredients, the stronger it is as a trademark, and the more work you have to do to explain it.

Nothing here is legal clearance. A real search is class by class, jurisdiction by jurisdiction, plus unregistered common-law use. Before you spend money on signage, pay an attorney for an hour.

Step 6: Decide, then stop

Put the surviving names in front of the people who have to live with them. Ask one question: which of these can you say to a stranger at a conference without explaining it? Then pick, register the domain and the handles the same day, and stop looking.

The most common failure at this stage is re-opening the search because someone says the name feels “weird.” Every name feels weird for about three weeks. Google, Kodak, and Xerox all felt weird. The test is not whether it feels familiar, it is whether it passed the say, spell, and search tests and cleared your classes.

A worked example

Say you are launching a grocery delivery service. Brief says: spoken aloud constantly, .com required, US and Mexico.

Word lists give fresh and market. Straight blending produces Fresket, which is short, two syllables, and clean. Clipping word two to the familiar splinter -mart produces Freshmart, which reads instantly but collides with existing grocery marks. Reversing the order gives Markesh. Onset splicing gives Frarket, which fails the say test because of the consonant cluster.

Shortlist: Fresket, Markesh, and two unrelated candidates. Say test kills nothing. Spell test kills Markesh because half the group writes Marquesh. Trademark search kills Freshmart. You ship Fresket, register it, and get back to work.

That whole loop takes a week, not a quarter, because the brief did the arguing for you. If you want a starting pile of candidates for your own word list, the brand name generator will produce them in bulk, and you can bring the shortlist straight into Step 4.