discuss After reviewing ~10,000 recent .com drops, one pattern kept showing up

[DomainGemsAI]

Over the past couple of weeks I’ve been going through a fairly large batch of recently dropped .com domains (around 10k names). The goal wasn’t to estimate resale value or anything like that. I was mostly curious about something simpler: which structural patterns actually survive once you start filtering aggressively? After looking through the list, a few things stood out.

First, CVCV patterns still seem to hold up surprisingly well. That classic consonant-vowel-consonant-vowel structure kept passing early filtering much more often than random four-letter combinations. Not shocking, but interesting to see how consistent it was. Names like Luma, Vero, Rivo, Nexo follow that structure — they’re simple, pronounceable, and pass the “say it out loud” test instantly.

The second thing that surprised me was that length wasn’t the biggest elimination factor. Phonetic friction was. Some short names still fail immediately if they’re awkward to pronounce — things like hard consonant clusters, strange vowel placement, or combinations that make you pause when saying them out loud. Even a short domain can feel “heavy” if the sound doesn’t flow. The third pattern I noticed was that the stronger names tended to have flexible meaning. In other words, they could realistically work across several sectors — SaaS, AI tools, fintech, marketplaces, consumer apps, etc. The more industry-agnostic the name felt, the stronger it looked overall.

Curious how others here think about this. When you're evaluating brandable domains, what do you usually prioritize first — phonetics or raw length? For context, I’ve been experimenting with running these kinds of pattern checks through a dataset pipeline I’m building to study drop-list behavior.

 

[Ali Mustafa]

Over the past couple of weeks I’ve been going through a fairly large batch of recently dropped .com domains (around 10k names). The goal wasn’t to estimate resale value or anything like that. I was mostly curious about something simpler: which structural patterns actually survive once you start filtering aggressively? After looking through the list, a few things stood out.
First, CVCV patterns still seem to hold up surprisingly well. That classic consonant-vowel-consonant-vowel structure kept passing early filtering much more often than random four-letter combinations. Not shocking, but interesting to see how consistent it was. Names like Luma, Vero, Rivo, Nexo follow that structure — they’re simple, pronounceable, and pass the “say it out loud” test instantly.
The second thing that surprised me was that length wasn’t the biggest elimination factor. Phonetic friction was. Some short names still fail immediately if they’re awkward to pronounce — things like hard consonant clusters, strange vowel placement, or combinations that make you pause when saying them out loud. Even a short domain can feel “heavy” if the sound doesn’t flow. The third pattern I noticed was that the stronger names tended to have flexible meaning. In other words, they could realistically work across several sectors — SaaS, AI tools, fintech, marketplaces, consumer apps, etc. The more industry-agnostic the name felt, the stronger it looked overall. Curious how others here think about this. When you're evaluating brandable domains, what do you usually prioritize first — phonetics or raw length? For context, I’ve been experimenting with running these kinds of pattern checks through a dataset pipeline I’m building to study drop-list behavior.

Good observations. In my experience phonetics usually matter more than length & if a name is easy to say, remember, and brand, it already passes half the test.

 

[Bravo Mod Team]

the biggest elimination factor. Phonetic friction was

This is one of my biggest eliminating factors, as well.

Right up there with common words and non-negative associations.

Below those are: easy to spell, memorable, etc.

 

[DomainGemsAI]

Good observations. In my experience phonetics usually matter more than length & if a name is easy to say, remember, and brand, it already passes half the test.

Thanks for sharing that — it lines up quite closely with what I started noticing while going through the lists. One thing that became very obvious while filtering was how quickly the brain rejects names that create even a small hesitation when spoken. Even when a domain is short, if you have to pause for a moment to figure out how it sounds, it almost immediately drops in perceived quality. That’s probably why structures like CVCV keep surviving the early filters so consistently. They tend to pass the “radio test” almost automatically because the sound pattern feels natural. I’m starting to think that pronounceability might actually be the real underlying metric, while length is just a secondary signal people use when scanning domains. Curious if in your experience certain letter combinations tend to work better than others when it comes to brandables.

 

[DomainGemsAI]

This is one of my biggest eliminating factors, as well.

Right up there with common words and non-negative associations.

Below those are: easy to spell, memorable, etc.

That’s a great point about associations. While going through the dataset I noticed something similar — even when a name is structurally solid, it can fail very quickly if the word carries any kind of negative or awkward meaning in common usage. So the elimination process started to look something like phonetic flow first, then association or meaning, followed by ease of spelling and memorability. What surprised me was how many names actually fail right at the phonetic stage before anything else even matters. Once a name sounds slightly uncomfortable to say, most of the other qualities become irrelevant. I’m thinking of running another pass on the dataset focusing more on association and semantic signals to see how often structurally strong names get filtered out for that reason. It would be interesting to see how often that becomes the deciding factor.

 

[Embrand]

Some good points, but length is still quite important. DomainAgents regularly report statistics about this. Here is the latest observation from their January newsletter:

 

[william]

10K is a large batch? I think most domainers go through more than that daily.

 

[Ali Mustafa]

Thanks for sharing that — it lines up quite closely with what I started noticing while going through the lists. One thing that became very obvious while filtering was how quickly the brain rejects names that create even a small hesitation when spoken. Even when a domain is short, if you have to pause for a moment to figure out how it sounds, it almost immediately drops in perceived quality. That’s probably why structures like CVCV keep surviving the early filters so consistently. They tend to pass the “radio test” almost automatically because the sound pattern feels natural. I’m starting to think that pronounceability might actually be the real underlying metric, while length is just a secondary signal people use when scanning domains. Curious if in your experience certain letter combinations tend to work better than others when it comes to brandables.

Yes i have noticed combinations with soft consonants (L, M, N, R, V) tend to feel smoother and more brandable.

 

[DomainGemsAI]

Some good points, but length is still quite important. DomainAgents regularly report statistics about this. Here is the latest observation from their January newsletter:

Show attachment 296951

That’s a fair point, and I agree length still matters. I probably should have framed it more as “length alone wasn’t the first eliminator in my pass.” A short name still has an advantage, but if the phonetics are awkward, that advantage seems to fade fast. So for me it started looking less like length vs phonetics, and more like length works best when the sound already feels natural.

 

[DomainGemsAI]

10K is a large batch? I think most domainers go through more than that daily.

Fair point — “large” is relative here. For full-time domainers going through lists daily, 10k probably isn’t much. I meant it more in the context of a focused pattern review rather than raw daily volume. The idea was to look closely enough at one batch to notice what kept surviving repeated filters.

 

[StarsEquipe]

For me length really matters, but pronounceability is what gives a domain what we call stickiness.