How to Choose a Domain Name: A 7-Step Framework

Most advice on choosing a domain name is a list of adjectives: make it short, make it memorable, make it brandable. True, and useless — nobody has ever been stuck because they forgot short names are good. The hard part is a process: how to get from a blank page to a name you can defend in two years.

The short version: what actually matters

Before the process, the honest hierarchy. Domain-name criteria are not equally weighted, and treating them as a flat checklist is why people end up with names that satisfy every rule and still feel wrong.

Step 1 — Write the one-line brief first

Do not open a domain search yet. Open a text file and write three lines:

1. What this is, in plain words. Not the pitch — the category. “Scheduling software for physical therapy clinics.”
2. Who says the name out loud, and where. A developer typing it into a terminal? A clinic owner hearing it from a colleague at a conference? A consumer seeing it in an app store? These have different winners.
3. What this could plausibly become in five years. This is the line that kills descriptive names. If the answer is broader than the brief, the name has to be broader too.

That third line does most of the work. books.com would have been a catastrophic name for Amazon. Naming the thing you sell today is the single most common way founders build in a rebrand.

Step 2 — Generate far more candidates than feels reasonable

The number one cause of a mediocre domain is a shortlist of four. Founders think of three names they like, find all three taken, get demoralized, and register the least-bad variant with a hyphen in it. The fix is volume: you want fifty to a hundred raw candidates before you evaluate any of them.

Productive generation methods, roughly in order of hit rate:

• Compound two ordinary words. The most reliable source of good available names. One word carries meaning, one carries feel: firehydrant.com, notion.so, stripe.com. Build two columns — one of category words, one of quality words — and read across.
• Take a real word slightly out of context. Words that exist but are not owned by your category. This is where most of the strong short names hide, because the obvious in-category words went twenty years ago.
• Use the metaphor, not the thing. What does your product do, physically? A lock, a bridge, a lens, a ledger, a runway. Metaphor names age far better than descriptive ones.
• Borrow from adjacent languages and mythology — but say it out loud first, and have a native speaker check it. This method generates both the best and the worst candidates.
• Modify a taken name you love. Prefixes that read as natural English (get, try, use, join, hey) are the most forgivable modification, because people say them anyway.

Write everything down, including the bad ones. Bad candidates are raw material — half of the strong finalists in any naming exercise are mutations of something that looked terrible in round one.

Step 3 — Run every candidate through six hard filters

Now evaluate. These are pass/fail, not preferences. Apply them in order and stop at the first failure — it saves time.

Filter 1: The radio test

Say the name out loud to someone. Do not spell it. Ask them to type it. If they hesitate, ask a clarifying question, or get it wrong, the name has failed — and it will fail the same way for every customer who hears it from a friend. The usual culprits are homophone risk (site/cite/sight), doubled letters, -ly versus -li endings, and any word whose spelling differs between American and British English.

Filter 2: Length and syllables

Under fifteen characters before the extension, and ideally two or three syllables. Syllable count predicts memorability better than character count. Two syllables is the sweet spot for a name people repeat.

Filter 3: No hyphens, no numbers, no doubled letters at the seam

Hyphens are the clearest amateur signal in a domain, and they are invisible in speech — you have to say “dash” every single time. Numbers create an immediate ambiguity (is it 4 or four?). Doubled letters at a compound seam, like sunnnetwork.com, break silently and constantly.

Filter 4: Trademark clearance

Search the name in the trademark databases for the markets you will operate in — the USPTO search for the US, the TMview database for international coverage. You are looking for existing marks in your class of goods or services; identical names in unrelated categories usually coexist fine. This is a screening step, not legal advice — if the name is going to carry a funded business, pay a trademark attorney for a real clearance search before you spend on brand assets.

Filter 5: Extension availability

Check whether your preferred extension is genuinely obtainable — registered, parked, listed for sale, or free. A name where the .com is an active competitor in your category should be discarded regardless of how good it sounds. Checking availability properly is its own subject, because registrar search results are not always what they appear.

Filter 6: Five-year fit

Read line three of your brief again. If the name only describes what you sell today, cut it. This filter eliminates most descriptive names and it should.

Step 4 — Decide the extension

The extension debate is over-litigated. Here is the practical version. .com is the default assumption for a general or non-technical audience — it is what people type when they guess, and its dominance in registered domains is well documented. If your customers are consumers, small business owners, or anyone outside tech, a .com is worth real effort to get.

Outside that audience, the penalty has largely evaporated. .io and .dev read as native to developers. .ai is now unremarkable for AI products. .co works as a general-purpose fallback. Plenty of serious companies operate on all of them.

Step 5 — Clear it legally and socially

You have two or three finalists. Before registering, spend twenty minutes on each:

1. Search the plain name. Not the domain — the word. What else already owns this term? An existing company, a band, a film, an unfortunate slang meaning? You are checking what you will be competing against for your own branded searches.
2. Check the trademark databases properly for your goods or service class, in every market you plan to sell in.
3. Check the social handles. Consistency across the domain and the major platforms is worth a lot, and finding out afterwards that your handle needs an underscore is an avoidable annoyance. A short suffix used consistently (getname, nameapp, namehq) is a reasonable compromise.
4. Check the name in other languages, particularly if you have any international ambition. Machine translation catches the obvious problems; a native speaker catches the ones that matter.
5. Check for a bad history. If the domain was previously registered, look it up in the Internet Archive to see what it used to host. A domain that spent three years as a spam blog carries baggage that is tedious to clean up.

Step 6 — Test the name on real humans

Founders are the worst possible judges of their own names, because they have heard the name a hundred times and it now sounds inevitable. Three cheap tests:

• The dictation test. Say it once in a normal sentence — “I work at Kestrel” — and ask people to write it down. Count the errors.
• The recall test. Show five people your shortlist alongside a few decoys. Come back the next day and ask what they remember. The one they recall is usually not the one you like most.
• The sentence test. Write three real sentences you will say constantly: “We built this at ___.” “You can find us at ___.com.” “___ handles your scheduling.” Names that are awkward in these sentences will be awkward forever.

Step 7 — Register it properly

You have a winner. A few things that are easy to get right now and expensive to fix later:

• Register it in your own name or your company’s, at an account you control. Not your agency’s account, not a contractor’s. Recovering a domain from a former vendor is a genuinely miserable process.
• Use a real email address you will still have in five years, and never one on the domain itself — if the domain lapses you lose the mailbox you need to recover it.
• Turn on auto-renew and registrar lock, and put the expiry date in a calendar you actually read. Lapsed domains get picked up within minutes by automated buyers.
• Enable WHOIS privacy unless you have a reason not to. It is free or near-free at most registrars and it cuts the volume of spam and fake-renewal scams substantially.
• Consider registering the obvious typo and the .net/.org only if the brand matters enough to justify the annual cost. For most early businesses it does not; buy them later once the name is worth defending.
• Register for multiple years if you plan to keep it. It is cheaper per year at some registrars and removes one recurring way to lose the name.

The five types of name, and when each wins

Every domain falls into one of five categories. Knowing which you are choosing clarifies most disagreements.

The questions people get stuck on

“Everything good is taken”

Every short single-word .com is registered — that has been true for years. But the space of two-word compounds is effectively unbounded, and that is where nearly all of the good names launched in the last decade came from. If your shortlist is empty, your generation step was too narrow, not the namespace.

“I cannot find the perfect name”

There is no perfect name. Names acquire their quality from the company behind them — Google, Yahoo, and Uber all sounded ridiculous before they meant something. What you are actually looking for is a name with no active defects: spellable, clear, legally safe, and expandable. Past that bar, ship, and stop optimizing.

“How long should this take?”

A week is plenty and a month is too long. Naming is a task with a sharp diminishing return: the first two days produce most of the good candidates, and everything after that is founders relitigating personal taste. Set a deadline before you start.

The name matters. It just does not matter as much as the time people spend on it suggests — and the difference between a fine name and a great one is far smaller than the difference between a fine name and one with a hyphen in it.