Joseph Peterson continues his deep dive into the languages of new top level domains.
Almost all new domain endings (nTLDs) are meaningful terms – from English suffixes like .WORLD and .ROCKS to the Spanish .UNO or German .JETZT. So far, we’ve counted how many nTLDs belong to each language. We also assessed which languages are over / under-represented within the nTLD program as it was conceived. But that’s just supply. What about actual market demand? Instead of merely counting the TLDs, we must look at domain registrations.
Let’s get started:
|Language||Domains||TLDs #1||TLDs #2||TLDs #3||TLDs #4||% GEO (DNs)||% GEO (TLDs)|
|Italian||35917||27||27||0||0||< 0.1%||< 0.1%|
This article is packed with stats. Before we get bogged down in explanations, please take a moment to scroll to the bottom … just to get a taste for the menu. When you forget what something means, come back here for the glossary.
The most obvious question – what’s up with #1 through #4? This same quartet of related statistics will appear in every table. So let’s deal with the shorthand right away. Within this article, only nTLDs with more than 100 registrations are included. .XYZ and .OOO, however, are set aside because those 2 suffixes don’t really belong to any language. Further, the TLDs are grouped in 4 ways:
- All the nTLDs that can be interpreted using a given language. (Some keywords are meaningful in 6+ languages. For instance, .SCIENCE counts as both English and French. To see how French or how English it is in practice, see my previous article. This measurement gives the biggest language count; and it can be viewed as an overestimate in some cases.)
- Same as #1 but excluding GEOs like .BERLIN, .NYC, .AMSTERDAM, .IRISH, etc. Far more GEOs were released for countries like Germany and Austria than for, say, the USA; and this adds quite a bit to the counts for German-language nTLDs. If these GEOs are included, then we can’t differentiate between language and locality as factors in market demand. (Earlier I wrote a couple of articles about GEOs, in case you’re curious.)
- All the nTLDs that are unique to a given language. By this measure, .SCIENCE is ignored altogether, counted neither for French nor English. Obviously, by excluding multilingual keywords, we’re missing a big part of the picture. Like #1, this count involves some distortion – but in the opposite direction. In order to avoid an overestimate, this count accepts being (potentially) an underestimate.
- Same as #3 but without GEOs.
Make sense? Above, you’ll also see the number of registered domains for each language. Although this count seems straightforward, it’s actually a complex estimate. Instead of tallying up every domain registered within all those TLDs that can be French – 100% of .SCIENCE + 100% of .BOUTIQUE + 100% of .RESTAURANT + etc. – we’re endeavoring to count only domains registered in TLDs that are indeed being interpreted as French. So any multilingual suffix would have a language share between 0% and 100%. The procedure for making this crude estimate was described in my previous article.
As a further wrinkle, similar estimates were made by excluding TLDs according to criteria #2, #3, #4. Those numbers are used in subsequent calculations but not displayed. Note: Raw registration volume data for country-TLD pairs is derived from nTLDStats, cached during January; and it has since been processed extensively for language and GEO labeling.
One very important question for the nTLD program – especially heading into a 2nd round of ICANN applications – is this: If more and more meaningful domain endings continue to be released, will buyers keep buying the new nTLDs as they pour in? Will the uptake be consistent, or will it drop off? How much supply can the market soak up before demand is saturated? Sooner or later, a maximum number of domain registrations must be reached, for there are only so many human beings on planet earth, only so many online projects in need of a domain. Brand protection may absorb .MARKET + .MARKETS, .PHOTO + .PHOTOGRAPHY without diminution; but if keyword nTLDs keep piling up, eventually registration volume will begin to sink, as consumers choose between options rather than gobbling up every new item.
If we’re looking for evidence of oversupply and dilution, then we’d expect to find it in a shrinking domain-per-TLD ratio within some particular sector – e.g. country or language or topic. Conversely, a high domain-per-TLD ratio might indicate unmet buyer demand – or, at any rate, room to stuff in more suffixes.
A single GEO nTLD for Kurdistan seems to be ample for Kurdish, given only 402 registrations. Meanwhile, the market appetite for English keyword nTLDs arguably has room for dessert. After all, with 30k – 48k domains registered (on average) for each of 281 – 378 English nTLDs, there’s a case to be made for adding 1 more. If the next registry achieves the same average result of 30k – 48k domains, they could be profitable.
Next in line would be French, but 11k domains per non-GEO suffix wouldn’t turn a profit so easily as 30k – 48k. This statistic is important for deciding what new nTLDs, if any, to apply for during the next round of ICANN applications. Despite there being 4 times as many Arabic as French speakers worldwide, from the ratios above it’s abundantly clear that the Arabic-language market isn’t clamoring for more nTLDs.
Chinese has the biggest ratio. Why? Anybody who has paid much attention to the domain market will remember the Chinese surge of 2015, which was a huge speculative bubble. During that time, many domains were registered in nTLDs – mostly at very low prices. Is that what we’re seeing here? Possibly but not necessarily. Keep in mind, this count doesn’t represent all Chinese-style domains nor all domains registered in China but, rather, domains registered in Chinese extensions – .WANG and .REN and .XIN, plus more than a dozen IDN options like .网址 and .信息. Before assuming you know what’s happening in China, read this. Even with all the Chinese registrations we’ve seen, it’s possible to view China as an underserved market.
|Language||% DNs #1||% DNs #2||% DNs #3||% DNs #4||% TLDs #1||% TLDs #2||% TLDs #3||% TLDs #4|
|Russian||0.5%||< 0.1%||0.9%||< 0.1%||1.2%||0.9%||1.6%||1.2%|
|Portuguese||0.3%||0.3%||< 0.1%||< 0.1%||6.3%||6.6%||0.5%||0.3%|
|Dutch||0.2%||< 0.1%||0.1%||< 0.1%||0.6%||0.0%||0.5%||0.0%|
|Hindi||< 0.1%||< 0.1%||< 0.1%||< 0.1%||0.4%||0.5%||0.3%||0.3%|
|Kurdish||< 0.1%||< 0.1%||< 0.1%||–||0.2%||0.0%||0.3%||0.0%|
|Tatar||< 0.1%||< 0.1%||< 0.1%||–||0.2%||0.0%||0.3%||0.0%|
|Basque||< 0.1%||< 0.1%||0.1%||–||0.2%||0.0%||0.3%||0.0%|
|Welsh||< 0.1%||< 0.1%||0.1%||–||0.2%||0.0%||0.3%||0.0%|
|Breton||< 0.1%||< 0.1%||0.1%||–||0.2%||0.0%||0.3%||0.0%|
|Afrikaans||< 0.1%||< 0.1%||–||–||0.2%||0.0%||0.0%||0.0%|
|Nepali||< 0.1%||< 0.1%||< 0.1%||< 0.1%||0.2%||0.2%||0.3%||0.3%|
|Arabic||< 0.1%||< 0.1%||< 0.1%||< 0.1%||0.6%||0.7%||0.8%||0.9%|
The table above seems clear enough. One detail worth noting: Roughly 1/4 of nTLD registrations are for language-less suffixes like .XYZ. Yet the percentages above add up to 100% because only nTLDs with language labels are considered.
|Language||% DNs /|
|% DNs /|
|% DNs /|
|% DNs /|
|Nepali||< 0.01||< 0.01||< 0.01||< 0.01|
When the ratio of domain share to TLD share is > 1, that implies demand exceeds supply. Notably, this is only true for Chinese suffixes. Obviously registry operators in 2013 didn’t anticipate the 2015 Chinese surge; otherwise they would have launched more China-focused nTLDs – whether in roman letters or as IDNs.
When the ratio of domain share to TLD share is approximately 1, we can infer that registries estimated market demand correctly and released the proper number of nTLDs. Wait a second! What? Isn’t it well known that registries anticipated greater market demand for new gTLDs than we’ve witnessed so far? Yes, that’s true. My claim isn’t that the absolute numbers are what they ought to be … nor what registry operators expected. What I’m saying is that registries correctly gauged the percentage of nTLDs that ought to be English – even if they misjudged demand overall.
When the ratio of domain share to TLD share is less than 1, supply exceeds demand – at least, relative to average nTLD performance. Because Chinese and English registrations dominate, the ratio for other languages is somewhat depressed. Acknowledging this, we can multiply by some extra factor when looking at the rest. Accordingly, we see that German, Japanese, and French occupy approximately the right share of nTLDs, given their market demand. That’s why their numbers are moderately high. Russian too – but only when GEOs are included … suggesting the market has room for more non-GEO keyword nTLDs in Russian.
|Language||% Sites||% DNs /|
|% DNs /|
|% DNs /|
|% DNs /|
High-flying domain registration stats ought to be tethered to real-world statistics. How well are Spanish nTLDs doing relative to the number of Spanish-speaking internet users? How are French nTLDs faring compared to the fraction of French websites? The bigger the ratio above, the more active a language community is in registering nTLDs.
What are the hottest nTLD languages? English is #2, followed – after a considerable gap – by German and French, both of which rely heavily on local GEOs. But Chinese is out in front by a mile. That’s not only because Chinese suffixes are registered in large numbers. It’s also because astonishingly few Chinese-language websites exist relative to China’s population.
|Language||Internet Users||Users per DN #1||Users per DN #2||Users per DN #3||Users per DN #4|
The contrast between a language’s websites and its online population is quite revealing. In particular, there’s a vast discrepancy for Chinese. As a result, when we look at the ratio of internet users to nTLD registrations, China drops back from the #1 lead we just saw. Yes, registration volume for Chinese nTLDs is very high when measured against the number of existing Chinese websites. But Chinese speakers have far fewer websites available to them than people do in the West.
If, instead of using websites as a benchmark, we measure nTLD reg volume against online participation, English beats Chinese by a factor of 3 – 7. On average, 1 English nTLD domain has been registered for every 54 online English speakers, compared to 1 Chinese nTLD domain for every 340 Chinese speakers online. When German GEOs are included, it seems German speakers and Chinese speakers are equally active in registering nTLDs.
Note: That’s per capita. We’re not counting distinct domain buyers. So a small number of speculators amassing large domain portfolios can greatly distort these statistics. As can registry manipulation. Ultimately, registration volume is NOT a reliable indicator of market demand. So take these numbers with a grain of salt.