Naming takes a week or it takes four months, and the difference is almost never the quality of the shortlist. It is that the founders never agreed what a good name has to do, so every candidate gets argued on a different basis. One person is defending a meaning, another is defending a sound, a third has quietly decided the domain is the whole question.
The name is judged by people who never hear the story.
Every founder has a story about where the name came from. Nobody else will ever hear it. A name is met in a subject line, a podcast mention, a badge at a conference, a colleague saying it once in a meeting.
That is the standard to judge against. Not whether the name captures what the company means, which no name does, but whether it survives being said badly by a stranger who is thinking about something else.
Names do not carry meaning at the start. They collect it. The good ones were mostly unremarkable words before the company made them worth something, and the work of making a name mean anything happens after it is chosen, which is covered in what startup branding does, and what it doesn't.
Order the constraints by what being wrong costs.
The reason naming stalls is that all the criteria get treated as equal. They are not. Some mistakes cost you a little friction for years, and some end the project.
In cost order: sayable, spellable, clearable, ownable. Test them in that order, because the cheap tests eliminate more candidates than the expensive ones do.
Meaning is the criterion founders weight hardest and it belongs near the bottom. A name that describes the product precisely is describing the first product, and the first product is the one most likely to change. A name built on a feature, a city or a file format is one you pay to get out from under later. An empty name you can fill is worth more than an accurate name you outgrow.
Sayable and spellable are free to test, so test them first.
Say each candidate on a phone call to somebody who has not seen it written. Then ask them to type it. That is the entire test, and a list of twenty does not take long.
A name people cannot spell after hearing it leaks word-of-mouth referrals, in small amounts you never see. A name people are unsure how to pronounce costs you something worse: they avoid saying it, which is the one thing you need them to do.
Invented names tend to survive this better than clever ones, and short beats long. A name that has to be spelled out loud has already failed.
Clearable and ownable are the ones that kill a name late.
Clearable means somebody in your line of business is not already using it, and a lawyer answers this, not a search engine. The expensive version of this mistake is a rebrand after the name is on the contracts and inside the product.
Ownable means you can get a domain and the handles without paying a fortune or accepting something people will mistype. An exact match is not required. A short prefix or a plain second word is fine. What is not fine is a domain that differs from the spoken name in a way you have to explain.
Check the handles at the same time, and say the candidate out loud to somebody who speaks another language if you sell outside one country. Both are quick, and both can end a shortlist late.
Do the clearance check on your top two before you fall in love, not after. That order is the whole trick.
Pick in a week, then stop discussing it.
Set the deadline before you start, because naming expands to fill whatever time it is given and the discussion gets less useful after the first few days. Generate widely for two days, cut with the four tests, take the top two to a lawyer, then choose.
Nobody on the team will be fully happy, and that is the normal outcome rather than a sign you chose wrong. A name is judged by what the company does under it. Ship it, put it on everything, and spend the remaining energy on Brand, which is the part people form an opinion from.
Tell us what you're building.