How long is a .com domain? We counted them.
We are adding .com domains to Hostprobe.io by name length. Along the way, we found a surprisingly clear peak: 11 characters is the most common length in our current dataset.
HostProbe Team
Data notes
3 min read
We have started adding .com domains to Hostprobe.io based on domain-name length.
Why length? Mostly because it felt like a relevant way to prioritize the enormous .com namespace. Shorter domains tend to include established brands, common words, abbreviations and names people are actually likely to search for. ford.com, disney.com and microsoft.com are rather more interesting starting points than some randomly generated 37-character domain.
Our first batch therefore contained .com domains with names of eight characters or fewer.
While preparing and scanning those domains, we started wondering what the overall distribution actually looked like.
So we counted.
.com domains by name length
The number below is the length of the domain before .com.
| Characters | Domains in our dataset |
|---|---|
| 2 | 1,165 |
| 3 | 30,403 |
| 4 | 682,620 |
| 5 | 2,970,012 |
| 6 | 4,940,014 |
| 7 | 5,211,151 |
| 8 | 5,977,893 |
| 9 | 6,528,887 |
| 10 | 7,107,470 |
| 11 | 7,240,735 |
| 12 | 7,147,201 |
| 13 | 6,775,008 |
| 14 | 6,232,767 |
| 15 | 5,538,707 |
| 16 | 4,785,720 |
| 17 | 4,072,843 |
| 18 | 3,388,041 |
| 19 | 2,766,764 |
| 20 | 2,236,341 |
There is a surprisingly clear answer:
11 characters is the most common .com domain length in our current dataset.
The population rises rapidly through the short names, reaches just over 7.24 million 11-character domains, and then begins a remarkably steady decline.
Our decision to start with names of eight characters or fewer therefore cut the distribution off while it was still climbing.
That first group contains approximately 19.8 million .com domains.
But how full is the namespace?
This led us to another entertaining calculation.
A normal DNS label can use letters, numbers and hyphens. Ignoring some special edge cases, there are 36 possible non-hyphen characters and 37 possibilities for characters inside the label.
Because a domain label cannot begin or end with a hyphen, the theoretical number of names of length n is approximately:
36² × 37^(n−2)
That makes the short-name namespace look very different:
| Length | Approx. possible names | Domains in our dataset |
|---|---|---|
| 2 | 1,296 | 1,165 |
| 3 | 47,952 | 30,403 |
| 4 | 1,774,224 | 682,620 |
| 5 | 65,646,288 | 2,970,012 |
| 6 | 2,428,912,656 | 4,940,014 |
| 7 | 89,869,768,272 | 5,211,151 |
| 8 | 3,325,181,426,064 | 5,977,893 |
At two characters, our dataset contains names corresponding to a huge proportion of the theoretical namespace.
By eight characters, there are more than 3.3 trillion possible names, while our dataset contains just under six million.
And it gets ridiculous very quickly after that.
This illustrates why simply saying that “all the good .com domains are taken” is both understandable and mathematically rather funny. The namespace is unimaginably large. Humans just happen to consider an extraordinarily tiny part of it useful.
What comes next?
Length also gives us a convenient way to expand Hostprobe.io systematically.
We have started with the shortest .com domains. The next groups contain:
- 9 characters: 6,528,887 domains
- 10 characters: 7,107,470 domains
- 11 characters: 7,240,735 domains
That is another 20.9 million domains just in those three groups.
We will continue adding longer names as Hostprobe.io’s .com coverage grows.
One important caveat
These numbers are not the number of all .com domains currently in existence.
They are the .com domains present in our dataset.
Our dataset is large, but we cannot currently say exactly how complete it is. There will be domains missing from it, and the Internet is obviously not static: domains are registered, deleted and changed continuously.
As Hostprobe.io obtains additional domain data, we will merge it into our known domain universe and keep these statistics updated.
That is actually part of the fun of building Hostprobe.io: as the dataset improves, we get to watch our picture of the Internet improve with it.