This is the fourth Joseph Peterson’s series examining new TLDs by country. Read parts one, two and three.
This article continues our examination of GEO nTLDs – new domain suffixes, released since 2014, that describe a city, region, nationality, or ethnicity. Data is derived (after much slicing and dicing) from nTLDStats.
Last time, we found that domain registrations for most GEOs are overwhelmingly concentrated in a single home country – .LAT and .IRISH being 2 exceptions. Now that we’ve analyzed the distribution of GEOs across various countries, let’s look at the significance of GEOs within national borders.
Country | # TLDs | % GEO | Local | GEOs |
---|---|---|---|---|
New Zealand | 1 | 91.8 | Yes | .KIWI |
South Africa | 4 | 35.8 | Yes | .CAPETOWN, .JOBURG, .DURBAN, .LONDON |
Belgium | 5 | 33.2 | Yes | .VLAANDEREN, .BRUSSELS, .GENT, .PARIS, .AMSTERDAM |
Austria | 4 | 26.6 | Yes | .WIEN, .TIROL, .BERLIN, .BAYERN |
Germany | 6 | 23.4 | Yes | .BERLIN, .BAYERN, .KOELN, .HAMBURG, .COLOGNE, .SAARLAND |
UK | 4 | 17.4 | Yes | .LONDON, .WALES, .SCOT, .CYMRU |
Russia | 3 | 16.6 | Yes | .рус (.XN–P1ACF), .MOSCOW, .москва (.XN–80ADXHKS) |
Turkey | 1 | 13.4 | Yes | .ISTANBUL |
Netherlands | 2 | 12.2 | Yes | .AMSTERDAM, .VLAANDEREN |
Australia | 3 | 10.9 | Yes | .MELBOURNE, .SYDNEY, .KIWI |
France | 4 | 10.5 | Yes | .PARIS, .BZH, .ALSACE, .CORSICA |
Spain | 2 | 8.6 | Yes | .BARCELONA, .LAT |
Ireland | 2 | 8.4 | Yes | .IRISH, .LONDON |
Mexico | 1 | 5.6 | Yes | .LAT |
Japan | 6 | 5.6 | Yes | .TOKYO, .NAGOYA, .YOKOHAMA, .OKINAWA, .KYOTO, .OSAKA |
Canada | 1 | 4.5 | Yes | .QUEBEC |
USA | 3 | 3.3 | Yes | .NYC, .MIAMI, .VEGAS |
Brazil | 1 | 1.9 | Yes | .RIO |
Switzerland | 3 | 1.2 | No | .BERLIN, .PARIS, .LONDON |
Ukraine | 3 | 1.2 | No | .рус (.XN–P1ACF), .MOSCOW, .москва (.XN–80ADXHKS) |
Armenia | 3 | 0.4 | No | .рус (.XN–P1ACF), .москва (.XN–80ADXHKS), .MOSCOW |
Hong Kong | 2 | 0.4 | No | .CYMRU, .WALES |
Hungary | 1 | 0.4 | No | .WIEN |
Chile | 1 | 0.4 | Yes | .LAT |
Malaysia | 2 | 0.2 | No | .VEGAS, .RYUKYU |
Romania | 2 | 0.2 | No | .BERLIN, .LONDON |
Denmark | 1 | 0.2 | No | .LONDON |
UAE | 1 | 0.2 | No | .LONDON |
Singapore | 1 | 0.2 | No | .VEGAS |
Israel | 1 | 0.2 | No | .VEGAS |
Greece | 1 | 0.2 | No | .LONDON |
Lithuania | 2 | 0.1 | No | .MOSCOW, .TOKYO |
Thailand | 1 | 0.1 | No | .PARIS |
The country whose nTLD registration volume is most dependent – by far! – on GEOs is New Zealand. .KIWI alone accounts for 91.8% of all nTLD domains registered in that country. However, this aberration is the result of a controversial registry campaign, in which some 200,000 .KIWI domains were registered suddenly. We therefore ought to ignore .KIWI for statistical purposes.
Overall, Europe is unquestionably the GEO nTLD capital. Out of a dozen nations where 8% or more of registered nTLD domains are GEOs, only 2 – South Africa and Australia – are disconnected from the greater European region (including Russia, Turkey, and the UK). Europe’s predominance is partly a result of supply, not demand. Registries designed more GEOs for European markets than anywhere else: 27 out of 45 ranking GEOs correspond to this region.
After the .KIWI disqualification, it seems the high water mark for GEOs is roughly 1/3 in both South Africa and Belgium. Close behind, Germany and Austria see 1/4 of their nTLD registrations go to GEOs. Crucially, each of these nations is home to multiple local city GEOs. Yet they all participate by registering GEOs corresponding to other countries too.
To guarantee a large GEO footprint, is it enough for a country to be targeted by several local GEOs? Nope. Just look at the USA, which encompasses .NYC, .VEGAS, and .MIAMI. For whatever reason, GEOs in the USA account for only 3.3% of registered nTLD domains.
With hundreds of English keyword nTLDs clamoring for attention, one could argue that language and oversupply have played a part in diminishing the American GEO footprint. Then again, english keywords apply equally to the UK, where GEOs claim a 17.4% share – more than 5 times higher than the USA. Residency requirements for .NYC are undoubtedly a limiting factor.
Success?
Here’s a question: Are GEOs more or less successful than non-GEO nTLDs? If we look at total registration volume, we’ll see a chasm between the 2 groups: 26,694,862 non-GEOs compared to 889,579 GEOs. Thus, only 3.2% of registered nTLD domains are GEOs. What does that really tell us?
First of all, GEOs are outnumbered – 62 GEO nTLDs versus 1151 non-GEOs. Hold on, though. Doesn’t that mean 5.1% of nTLDs are GEOs? In that case, GEOs are underperforming with only 3.2% of registrations, right? No. Unfortunately, it’s absurdly simple for companies to manipulate registration volume. With 1-penny domains and registrar stuffing, .XYZ, for instance, greatly inflated its volume. .BERLIN also jumpstarted its total with large-scale freebies back in 2014. The most inflated nTLDs, being the largest, greatly distort these statistics or, at least, undermine their reliability. Consequently, we must try something else.
Alright, which nTLDs rank well in national top 100 lists? (nTLDStats tracks the biggest 49 registrant nations.) Out of 62 GEOs tracked by nTLDStats, 45 of them have charted (72.6%). In comparison, only 344 out of 1151 non-GEOs (29.9%) are anywhere to be found in those country rankings – even with 4819 chart positions for 49 countries. By this standard, GEOs are outperforming non-GEOs in striking fashion.
Is such a test fair? Well, it’s fair in the sense that it highlights the kind of success GEOs aim to achieve – large local presence somewhere in particular … not everywhere. .BARCELONA ought to be judged by its footprint in Spain primarily and not measured against the worldwide volume of .CLUB or .SITE or .VIP. In the previous article, we saw that 11.6% of ranked nTLDs are GEOs. Considering only 5.1% of nTLDs are GEOs in the first place, it could be said that GEOs are over-performing by a factor of 2.3. But GEOs are doing even better than that!
Best Rank | GEOs | non-GEOs | % GEO |
---|---|---|---|
1 | 6 | 9 | 40.0 |
2 | 4 | 11 | 26.7 |
3 | 7 | 9 | 43.8 |
4 | 5 | 6 | 45.5 |
5 | 3 | 11 | 21.4 |
6 | 5 | 8 | 38.5 |
7 | 2 | 4 | 33.3 |
8-10 | 0 | 30 | 0.0 |
11-20 | 2 | 46 | 4.2 |
21-30 | 3 | 50 | 5.7 |
31-40 | 3 | 35 | 7.9 |
41-50 | 3 | 30 | 9.1 |
51-60 | 1 | 27 | 3.6 |
61-70 | 0 | 21 | 0.0 |
71-80 | 0 | 20 | 0.0 |
81-90 | 0 | 17 | 0.0 |
91-100 | 1 | 10 | 9.1 |
Returning to those national top 100 nTLD lists, let’s record the best ranking found in any country for a given nTLD. Accordingly, there are 15 nTLDs that rank #1 somewhere among the most active 49 registrant nations. 6 of them are GEOs. For a different set of 15 nTLDs, #2 is their best ranking. Of those, 4 are GEOs. And so forth.
For cardinal rankings #1 through #7, look how high the percentage of GEOs is! At the very top of national lists, 21.4% – 45.5% of nTLDs are GEOs. To be exact, if an nTLD peaks at #1 – #7 somewhere, then there’s a 35.6% chance it’s a GEO. Remember, only 5.1% of existing nTLDs are GEOs. So they’re overrepresented at the very top of national lists by a factor of 7.
More than 1/3 of ranked GEOs (1/4 of tracked GEOs) peak at #1, #2, or #3. More than 2/3 of ranked GEOs (1/2 of tracked GEOs) peak between #1 and #7. (Once again, for reference, 62 GEOs are tracked by nTLDStats; and 45 of those have charted.) It’s surprising how rarely GEOs fail to reach the top of the charts. 11 peak between #8 and #50, and only 2 peak between #51 and #100. That’s 93% of the dartboard; but, among the nTLDs stuck there, only 4.3% are GEOs. For 299 non-GEOs, a ranking of #8 – #100 in some country is the best they can boast of. Another 807 non-GEOs haven’t managed a #100 ranking anywhere at all.
Referring again to the first table, GEOs have charted in 33 countries out of 49. Although .PARIS ranks in Thailand and .LONDON ranks in Greece, let’s set aside countries where the ranking GEOs aren’t local. That leaves a small set of 19 countries with ranking local GEOs. In these cases, GEOs perform very well as fractions of the country’s total registration volume: Median = 10.9%; Q1 = 5.6%; Q3 = 23.4%.
If every country had access to a handful of local GEO nTLDs – as South Africa, Belgium, Germany, Austria, Australia, and the UK now have – would those nations replicate these high registration rates? It’s a fair question. If more nTLDs are released in a few years, there’s a rational case to be made for including more GEOs among them.
Ultimately, of course, registration volume cannot be trusted as an indicator of TLD success. Without quality registrations meant for usage, volume may be nothing but a speculative bubble or an inflated registry promo.
drew says
Joseph, this has been an informative and interesting set of analyses. For GEO TLDs, I’d be curious to know if there is a lot of SLD overlap between different locales. Like, is the word ‘shop’ or ‘visit’ or ‘restaurant’ common among different GEOs, or is there more variance? A lot of unique domains in a particular GEO that relate to the actual place might indicate a healthier namespace than one with lots of generic registrations.
One thing I’d suggest for further research is to reconsider what you count as the # of available TLDs. Out of the 1151, many of them are Brand or Closed TLDs, which skew the % quite a bit. Roughly 350 TLDs have only one domain in the zone file (nic.TLD), and over 500 have less than 10 registrations (many of these are the Brand TLDs). Discounting these would bring the % of GEO TLDs closer to 10% of all available nTLDs.
Please keep this up. It’s refreshing to read well thought out and researched pieces on the domain industry.
Joseph Peterson says
@drew,
Peer review – bravo! Glad someone is reading this thoroughly enough to notice problems.
That’s a fair criticism – about # of available TLDs. My database has a field for marking dot brands, but I evidently forgot to exclude them. Doing so would have been as simple as adding 1 line to my query: “and not `Dot Brand`”. Ugh! Also, I ought to go back and label other TLDs that appear generic but are nonetheless operated as closed name spaces. As you say, including those TLDs – which are not publicly registrable – causes some of my numbers to be off.
A crude way to do set these aside might have been to insist on a minimum number of registrations. But that’s too arbitrary. Depending on the minimum threshold, the numbers would vary quite a bit. And we might throw a lot of babies out with the bathwater.
There are so many different policies at work with 1000+ suffixes and registries each having their own peculiar restrictions, marketing strategy, pricing structure, discount history, and (let’s not forget) rollout dates, that it’s impossible to control for all variables and get a genuine apples-to-apples comparison. Many of these peculiar registry traits are “averaged out” when looking at dozens or hundreds of TLDs. The problem you’ve pointed out isn’t of that kind, since it affects the 2 sets – GEOs and non-GEOs – unequally.
Still, a perfect solution isn’t achievable because the dividing line can be blurry and subjective. While some nTLDs are obviously closed, many others are semi-closed. .BIBLE, for instance, is open only to people who pass some vaguely defined religious test – a subset of christian sects. Others like .CAR, though theoretically open, are priced high enough that they exclude a major class of registrants other nTLDs depend on for volume – and, hence, rankings. .NYC is semi-closed too because of local residency requirements.
Of course, I made a mistake by not taking dot Brands into account. Not offering an excuse for that error – only rhapsodizing about the challenges of drawing conclusions about groups of TLDs. Grouping them always fails to some degree, one way or another.
I’ll try to account for the misleading closed keyword nTLDs before the next article. If anybody has a list handy, please email it to me, since that will save me half an hour.
drew says
It’s regrettable that there isn’t an easy to grab list out there which clearly labels the various states of the TLDs (open, closed, application based, etc.) to make it easier to slice and dice. Like you said, there are so many variables at play with the 1,000+ TLDs that make any comparison is tricky and full of caveats. And limiting on zone file size has its own issues, like you point out with .CAR.
Joseph Peterson says
@drew,
About 2LD overlaps – strings to the left of the dot that are registered in multiple TLDs – yes, I’d like to analyze that data.
Is it true that lack of overlap is a sign of healthy interest in a name space? I do follow your train of thought here, but I’m not so sure. Many of the most desirable names would be generic terms or phrases, applicable all over the place. Meanwhile, there are a lot of seemingly random strings that could be registered uniquely in a given TLD; but, if generated robotically, they’d hardly be a sign of health.
We’d have to control for language, since .BERLIN and .KOELN will have more in common with one another than they will with .LONDON; and .WALES and .CYMRU would overlap more than .CYMRU and .KYOTO.
To draw any conclusions about name space health, we’d have to have some sense of what “normal” is, first of all. And establishing that is not so easy, given how small the family of GEOs is – especially when subdivided by language. Too tiny a sample size, perhaps. The best comparison might be with ccTLDs, since they’re more established and add a larger pool of data.
Garth says
I wouldn’t class .kiwi as a GEO. Not a place name.
Joseph Peterson says
@Garth,
My definition includes ethnicities / nationalities as GEOs – e.g. .KIWI, .CYMRU, .LAT, .IRISH, .SCOT, etc. See previous article:
https://domainnamewire.com/2017/01/31/analyzing-new-top-level-domain-registrations-country-part-3-geos/
There’s no underlying “truth” to what a GEO is. Just a matter of drawing a circle around various scattered TLDs. People will choose to lasso different cows.
Sure, I could analyze the 5 different types of GEOs separately: (1) cities; (2) states / provinces / regions; (3) countries; (4) nationalities; (5) ethnicities. But that’s too many articles. And, after subdividing into tiny sets, it’s harder to draw statistically meaningful conclusions.
Rubens Kuhl says
.kiwi is clearly focused on New Zealand, so even though not a city or region name, it’s focused on a cultural identity. It is as a “fruity” domain as .apple … 😉
Calltometal says
Is .RYUKYU Malaysian? I thought this was a term for the Okinawan islands in Japan. Part of the Ryukyu Kingdom (琉球) in older history.
Joseph Peterson says
You’re correct.
Joseph Peterson says
@Calltometal,
See discussion of .RYUKYU in previous article. (Link above.)
Calltometal says
Sorry Joseph. I read the chart wrong. I see that connection, now.
Thank you and amazing data analysis!
Calltometal says
Sorry, Joseph. I see that connection now. Thank you for pointing that out. Amazing data analysis, by the way!
Yann says
Hi Joseph
Again, this is a great article. I do have one question though: is there any reason for not including (or mentioning) the Dutch GeoTLD .frl?
Joseph Peterson says
@Yann,
Only human error. When going through 1000+ nTLDs and marking them as GEOs, .FRL escaped my attention. Possibly I missed it because it was an acronym rather than a keyword.
.FRL ranks #4 in the Netherlands, taking up 6.78% of nTLD registrations in that country, where 99.2% of .FRL registrations are registered. .AMSTERDAM ranks #2 with 12.0%.
So the row in the first table ought to read:
Country Netherlands
# TLDs 3
% GEO 19.0
Local Yes
.AMSTERDAM, .FRL, .VLAANDEREN
And we’d also change a row in the 2nd table (for cardinal rank #4) as follows:
Best Rank 4
GEOs 6
non-GEOs 6
% GEO 50.0%
Arthur says
And with 19% , catapulting Netherlands from place #9 to place #6 in terms of %geo of ngTLDs.
Joseph Peterson says
Yes, that’s so.
.shopman says
Joseph, The % of GEOs in nTLDs can be quite misleading as the market size differs markedly in every country so in a tiny market with low uptake, the % can be quite exaggerated.
I will take the example of South Africa. The combined registrations for .JOBURG, .CAPETOWN and .DURBAN is probably roughly 10,000 regs. Given that these were “home gTLDs”, the uptake is not surprising but South Africans are not so active in other geo TLDs or even nTLDs. For example, South Africa only has 82 .LONDON domains (there is an East London in ZA), 11 .BERLIN(there is a Berlin in ZA) domains, 9 .VEGAS domains etc. That doesn’t represent a “general trend” towards embrace of GEOs. Given that embrace of other nTLDs is also lukewarm(in the low hundreds), I am even surprised the % of GEOs in that country is only 33% instead of something like 70%. So here, too, you could be looking at an aberration and the local GEOs have significantly contributed to the disproportionate GEO and indeed nTLD footprint.
Are GEOs less successful? Of course and for various reasons. Generic TLDs like .XYZ, .SHOP, .CLUB, .VIP, .ONLINE offer users greater scope of use without any geo-restrictions so it is a no-brainer that they would be hugely successful globally relative to limited city GEOs. Even if you remove the “distortion” in numbers of the nTLDs and I guess this is .XYZ “distortion”, we still remain with over 26 million generic ngTLDs versus 880,000 GEO ngTLDs. That does not necessarily mean users don’t like GEOs; their scope is simply very limited. If you live in Berlin and are targeting the larger German or DACH market, you are better off with a .DE, .EU, .COM or .shop rather than a .BERLIN.
“Considering only 5.1% of nTLDs are GEOs in the first place, it could be said that GEOs are over-performing by a factor of 2.3. “
Isn’t this comparing apples and oranges. If GEOs represent 5.1% of nTLDs in every city globally, it could hold true. But local GEOs are only significant in their respective cities.
I guess you could get the same result if you compare rankings for ccTLDs vs .COM or .NET in their respective countries.
You simply can’t compare GEOs and other generic gTLDs without a risk of misrepresentation as both are targeting very different markets. GEO nGTLD registries can always push several 1000s regs to a few users but users are already adequately covered by ccTLDs for local markets. Any more GEOs would simply be redundant and would be piling up massive registration costs on end users who are already well covered.
Rubens Kuhl says
There is a false premise in this sentence:
“Generic TLDs like .XYZ, .SHOP, .CLUB, .VIP, .ONLINE offer users greater scope of use without any geo-restrictions so it is a no-brainer that they would be hugely successful globally relative to limited city GEOs.”
The false premise is of localization being a limiting factor; there are two types of uses where that doesn’t hold up:
– Local businesses.
– Global businesses with strong localization.
Google redirects from google.com to google.ccTLD in almost all markets where they are active; they haven’t made yet such a move with a City TLD, and I don’t think they will, but I wouldn’t be surprised if Uber starts doing such a move with City TLDs…
Local businesses also like identifying themselves with the local region, see the strong affinity ccTLDs have throughout the world.
The ones that don’t like local TLDs:
– Those that are not big enough to invest in localization
– Those that are somewhat shame of where they are.
Joseph Peterson says
@shopman,
Let me try to unpack your comments, 1 paragraph at a time:
1. Yes, the GEO percentage can be high in some countries. That’s not “misleading” as you call it. Rather, it’s the reality we’re measuring.
2. You point to South Africa and other countries with multiple local GEOs as an “aberration”, but your reasoning seems to be only that they do have a high percentage of GEO registrations. Yes, of course, the more local GEOs released for a country, the greater the chance it will have registered a larger number of GEO domains. Again, that’s not a mistake; it’s the precisely what I’ve set out to measure.
You are right that including GEOs, where they exist, will cause a “disproportionate GEO and indeed nTLD footprint”. Which is why, when comparing separate countries on the basis of total nTLD registration volume, it’s important to exclude the GEOs. If you read articles #1 and #2, where I examined nTLD volume in aggregate, you’ll find I did filter out the GEOs.
3. Here you’re exhorting people to use .DE instead of .BERLIN and expressing a hunch that versatile non-GEOs will do better than GEOs. You’re welcome to your opinions.
4. You’re confused about what the 5.1% refers to. That’s a fraction of nTLDs, not a fraction of nTLD domain registrations. But your main point seems to be that we can’t call GEOs a success or even assess their performance unless they’re found in large numbers “in every city globally”. Considering only a few countries have any local GEOs, I guess knowledge is impossible then.
5. You say we can’t compare GEOs and non-GEOs because they are “targeting very different markets”. Nonetheless, I have compared them.
Sorry if this seems dismissive. I wanted to do my best to answer you, rather than ignore your post. If you can point out any specific errors in my article, I’m happy to listen.