How popular are Duolingo courses really?

[SweNedGuy]

First of all: I never tried to be complete, only to give a few examples. The main idea was to choose a few languages which are very popular on Duolingo and a few others which are less popular (but with a significant number of learners). I also wanted to include some non-European world languages. This accounts for my 'sampling'.

• The first column (next to the language label) shows the Duolingo language subscription count, which you can see when adding a new language course (after opting for English as teaching language). We should keep in mind that this number is generally cumulative. Duolingo started marking this number as 'active learners'. Inactive trees are removed after a month, however the activity threshold is only that learners finish at least one lesson. Users can remove language trees and may also unregister from DL. This reduces the subscription count.

• The second column shows the DuoMe total of active learners for that course. You can see this by typing for instance : https://duome.eu/en/es
  Active learners have at least attained level 2 in a course (60 XP). The limiting factor is that learners are included on DuoMe only after they have once attained a 100 day streak. The active learners number may not be cumulative: if you unregister from DL, you will disappear from DuoMe. If you eliminate the tree after completion, it will disappear in DuoMe. While you keep your profile private, you will not be shown in the list. (That may be transitory)

• The DuoMe statistics page shows other useful numbers: The third column shows active users having surpassed level 10 (only 3000 XP are needed). This differentiates sustained learners from 'dabblers'. The next column shows users having attained XP-level 25 which corresponds to 30.000 XP.

• The fifth column shows the number of golden owls.

Finally I have calculated the percentage of active learners in DuoMe relative to the total number registered in DuoLingo. The very last column is the percentage of DL golden owls among the number of active learners (on DuoMe).

Example of the DuoMe statistics output (for Spanish)
185214 Students
185214 Trees
147402 @ L10+ ········· 79.58%
84779 @ L25 ········· 45.77%
13386 Golden Owls ··· 7.23%
482 Golden Trees ····· 0.26%

Some observations:
The 100 days streak inclusion threshold for DuoMe is the main limiting factor for being considered a "committed active learner".
Any course an active learner adds, will soon be included in the DuoMe statistics (60 XP: a few lessons or a placement test are enough).

The most popular courses (Spanish, French) have a low percentage of committed active learners. That is also the case for non-European language courses, especially Hindi has a very low 'active learners' percentage. The less popular courses (Swedish, Dutch) do better. German holds a middle ground. It's a Germanic language as English is supposed to be, but the grammar is hard ... Italian has also more devoted learners.

Yet if users are committed, the 'owl'-ratio can be significant. Less popular languages and Italian do better. Chinese and Hindi are best in class among the rest. The frequent course additions and changes to the Spanish and French courses from English come at a price. Most learners don't try to regain the owl any more ...

Internal links:
Second batch of languages from English (with comments) (related and benchmark courses)
¿Son realmente muy populares los cursos de Duolingo? (Popularity of courses from Spanish)
Les cours de DuoLingo, sont-ils vraiment populaires? (popularity of courses from French)
Hoe populair zijn de Duolingo cursussen vanuit het nederlands? (popularity of the few courses from Dutch)

[pawndemic]

Very interesting statistic. Apparently Swedisch and Italian are thrilling to learn. But I guess it is hard to conclude anything about popularity.
Many language learner quit very fast. That is the reason why it is so hard to find a good language partner (tandem). Most of them quit in the first weeks.
I ask others, which had the same experience. Maybe they followed false promises and got quickly disappointed. So maybe an early quit is not only Duolingos fault.
Another problem is that Duolingos courses quite different in their quality. Some have stories, others not. Some are more or less well supported with a moderator (in the sentence forum) some are not. Some are updated more frequently others aren't and so far and so on.

If I had to guess I would assume that most of the Duolingos user are either "gamer" or people who use Duolingo as a supplement. I don't know what motivation the "gamers" have . Maybe they learn to maximize their lingots or their XPs, their crowns, or they search the fastest walkthrough. So maybe getting an owl is not their main target.

Among the language learner could be a chunk of people who like to dabble in diferent languages with no intention to finish something.

[LizLizLiz]

This is very interesting information -- especially considering Duolingo is now publicly traded (as of this past summer). The number of active, committed learners would be an important number for anyone considering investing in the company.

It would be interesting to look through the Duolingo 10K annual report to find out if this information is conveyed somehow. Here is a link to that report: https://duolingo.gcs-web.com/sec-filing ... -22-000039

[Basler Biker]

[Basler Biker]

The high "commitment" rates for SV, NL, DE.. can for sure partly be explained by the fact that these countries

• accept a high number of immigrants
• AND those have the "inburgeringsplicht" so they need to know the language before they can relocate/settle/work there
  can't judge about the other countries

[Basler Biker]

Documents such as the 1OK are made by the companies (or an external consultant on their behalf)
Numbers such as income, profit, subscribing/paying users would be hard to fake, as they appear in a "balance" and its details about income, but for number of users DL has never given any proof. The numbers shown in the tables below don't lie, and even if you would multiply them by a factor 10 (assuming e.g. that duome only fetches a "statistical subset of DL data" just to be indicative on "trends", then still these numbers never will reach the hights claimed by DL.

One can understand that 10K is business-talk to keep the stakeholders happy "that all is going well"
Yet the shares going down in less than a year, from 202 to 72 USD today, proves the opposite.

But the good news is - for those who believe in a revival - when shares are low then BUY.

Nasdaq DUOL 1Y

[SweNedGuy]

There also is a category "alumni" of users covered by DuoMe with a current streak below 100 days but who have previously had a 100+ days streak. That doesn't in itself explain the anomaly.
One is tempted to assume that many users lose their streak while on holiday. I see no other plausible reason why the number of users with streak length nearing 340 days would suddenly fall off a cliff only to resume to its plausible trend value at a much higher streak length. Whenever DuoLingo goes down, a streak preservation is usually guaranteed.

[BalkanGod]

I think that the lack of consistent use for the French and Spanish courses is due to the following:

• People using duo for a vacation/a goal then they have the vacation/give up

• XP-grifters that do the basic categories then delete the language to inflate the leaderboards

• Some people just outgrow Duo. When one gets further along in the process, Duo just
  becomes too elementary for one's goals.

Some languages also have better courses than others as previously mentioned. The Russian course is plagued with errors as an
example while the Dutch course is well-made.
Some languages also have a cult following due to pop culture such as Norwegian, Japanese, and Korean.
Does someone have the statistics for the smaller European languages like Danish, Romanian, or Czech? They have
a much smaller learner count, but are the learners for the languages more active? I wonder whether Duo adding
stories for Italian made an impact for the amount of Italian users?

[SweNedGuy]

In this second and last batch of language courses learnt from English, I extend the popularity checks to some related or benchmark languages other users were interested in.

With a comparable population as Denmark, Norwegian has a relatively high Duolingo subscriber base, yet both the percentage active learners and the owl-ratio don’t match that of Swedish.
Danish has a low DL subscriber base with however a higher percentage of active learners. The owl ratio is above average but doesn’t match the 21% of Swedish.

Now follow two Slavic languages both using a Latin alphabet (with quite a few diacritical marks). Polish has a much higher subscription base than Czech. The percentage of active learners is above average and so is the owl-ratio. Czech has a high percentage of active learners, but the owl-ratio is mediocre.

The two missing Romance languages are Portuguese and Romanian. The former has a high subscriber base (though not to the tune of Spanish or French). The active learners percentage is above average and with an owl ratio above 20%, Portuguese learners are among the most dedicated. Due to its low DL subscription base, Romanian has a high percentage of active learners. The owl ratio is however just average.

Irish has a DL subscriber base of 1.2 million, more than the number of native speakers. The percentage of active learners and the owl-ratio of our Celtic language example are however mediocre.

The vibrant pop-culture of Korea has inflated the DL subscriber base. Very few subscribers turn into learners, with a dreadful 0.29% active learners percentage. The owl-ratio also is low.
Arabic has a quite high DL subscriber base. Both, the percentage of active learners as the owl-ratio are below average. Learning Arabic is not ‘a walk in the park’.
Despite the low subscriber base of Indonesian, the percentage of active learners is average, while the owl-ratio is low. This Australasian language uses Latin script and its syntax is very easy. Unfortunately, these trump cards don’t pay off for Indonesian.

[GermanApple]

It's interesting how few golden owls there are for these languages. For example, only about 8% of Korean learners on Duolingo have a golden owl. For the amount of time the Korean course has been on Duolingo, you would think there would be more. And the K-Pop fans would try to learn a fair amount of Korean to listen to their favorite songs, right?

I'm interested to know what the golden owl ratio is for languages like Navajo and Hawaiian, short courses that some people do just to get a golden owl. The subscriber bases are also some of the smallest, but they could be the most dedicated.

Anyway, thank you for all this data. This must have taken a long time to do.

Have a good day,
German Apple

[SweNedGuy]

Below are the language courses from English, I didn't check yet on DuoMe. They are sorted by decreasing DuoLingo subscription totals as of March 16. Does anybody feel the vocation to add the DuoMe data for his favourite course?

Language DuoLingo subscriptions
Russian 4830000
Turkish 2870000
Latin 1410000
Vietnamese 1350000
Greek 1210000
Hebrew 905000
Ukrainian 664000
Hawaiian 588000
Finnish 564000
High Valyrian 519000
Welsh 497000
Yiddish 489000
Scottish-Gaelic 430000
Swahili 394000
Hungarian 356000
Klingon 305000
Esperanto 295000
Navajo 294000
Haitian Creole 159000

[norrsken]

With a comparable population as Denmark, Norwegian has a relatively high Duolingo subscriber base, yet both the percentage active learners and the owl-ratio don’t match that of Swedish.
Danish has a low DL subscriber base with however a higher percentage of active learners. The owl ratio is above average but doesn’t match the 21% of Swedish.

Norwegian's popularity is inflated because the course was released when SKAM was trendy, and it has a reputation as the highest quality user-contributed course on Duolingo. I've heard that Swedish is also inflated because Duolingo is being promoted to immigrants as a substitute for public language lessons (the Swedish For Immigrants program, SFI), which are said to be ineffective for learning the language. Actually, I've also heard of some SFI schools assigning students Duolingo as homework. Danish is seen as the hardest to learn because of its complex pronunciation and relative lack of resources. And everyone calls it the 'potato in your throat' language.

As for Swedish and Danish's high golden owl ratio, the courses haven't been updated since they were launched. They teach less than two thirds of the words Norwegian does and have noticeably short trees.

[Basler Biker]

With a comparable population as Denmark, Norwegian has a relatively high Duolingo subscriber base, yet both the percentage active learners and the owl-ratio don’t match that of Swedish.
Danish has a low DL subscriber base with however a higher percentage of active learners. The owl ratio is above average but doesn’t match the 21% of Swedish.

Norwegian's popularity is inflated because the course was released when SKAM was trendy, and it has a reputation as the highest quality user-contributed course on Duolingo. I've heard that Swedish is also inflated because Duolingo is being promoted to immigrants as a substitute for public language lessons (the Swedish For Immigrants program, SFI), which are said to be ineffective for learning the language. Actually, I've also heard of some SFI schools assigning students Duolingo as homework. Danish is seen as the hardest to learn because of its complex pronunciation and relative lack of resources. And everyone calls it the 'potato in your throat' language.

As for Swedish and Danish's high golden owl ratio, the courses haven't been updated since they were launched. They teach less than two thirds of the words Norwegian does and have noticeably short trees.

Yes, I agree about Swedish, a shorter but stil very usefull tree gets completed more easily. I have been collecting the best bits from DL, and my own links to all kinds of resources. My favorites being the two 8sidor.se and Readlang combined.

It would be fine if you - affiliated to the Swedish flag - could have a look at Swedish from English forum here, and give me (and readers to come...since the announcement of DL forums to close, they register here at a rate of 50/day...) some feedback and additional hints, tips and tricks that you would know of.

Vanliga hälsningar.

[Basler Biker]

Below are the language courses from English, I didn't check yet on DuoMe. They are sorted by decreasing DuoLingo subscription totals as of March 16. Does anybody feel the vocation to add the DuoMe data for his favourite course?

Language DuoLingo subscriptions
Russian 4830000
Turkish 2870000
Latin 1410000
Vietnamese 1350000
Greek 1210000
Hebrew 905000
Ukrainian 664000
Hawaiian 588000
Finnish 564000
High Valyrian 519000
Welsh 497000
Yiddish 489000
Scottish-Gaelic 430000
Swahili 394000
Hungarian 356000
Klingon 305000
Esperanto 295000
Navajo 294000
Haitian Creole 159000

May I be frank, yet polite of course. I am quite a statistics man too, but numbers don't mean anything if labels are not explained what they "hold". E.g. what is a DL subscription? I had 6 DL aliases, learning a mix of different languages. Does that make 6 subscriptions? Or is it DL app downloads, which are meaningless either. Or is it "active learners" - again a very "undefined" concept. So I wonder what "subscriptions" is supposed to mean

The other thing being that you ask us to give feedback on "our" language...what number are we supposed to give. From the people who register here we can count their "affiliation" flags (they are shown down the main forum structure - either the registered ones, or the activated ones (=90% of the registered aliases) but for the rest I would not have a clue how to find out, based on duome data, how many people are learning, or what they learn... since we can't query the signatures.

[SweNedGuy]

May I be frank, yet polite of course. I am quite a statistics man too, but numbers don't mean anything if labels are not explained what they "hold". E.g. what is a DL subscription? I had 6 DL aliases, learning a mix of different languages. Does that make 6 subscriptions? Or is it DL app downloads, which are meaningless either. Or is it "active learners" - again a very "undefined" concept. So I wonder what "subscriptions" is supposed to mean

The other thing being that you ask us to give feedback on "our" language...what number are we supposed to give. From the people who register here we can count their "affiliation" flags (they are shown down the main forum structure - either the registered ones, or the activated ones (=90% of the registered aliases) but for the rest I would not have a clue how to find out, based on duome data, how many people are learning, or what they learn... since we can't query the signatures.

I tried to be as clear as possible in the opening post: the DuoLingo subscriptions number is what you see for all courses from English when attempting to add a course: it provides 3 significant digits: E.g. 1.35M or 905K.

The 'active learners number' and other data are totals you see in the DuoMe list:
E.g. https://duome.eu/en/sv now gives following numbers:

Statistics
29011 Students
29011 Trees
16002 @ L10+ ········· 55.16%
6076 Golden Owls ··· 20.94%
1050 Golden Trees ····· 3.62%

I use the number of Students, L10+ trees and the number of golden owls. The ratio (now 20.94%) is also used.
The number active learners is divided by the number of Duolingo subscriptions to yield a percentage of how many subscriptions eventually turn into active users.

I also added possible/plausible reasons why this ratio is generally quite low. Yet it still allows comparing among courses.

For low 'owl ratios' I pointed to tree updates which make you lose your owl (and learners may give up on fetching it back)

Personally I once lost an owl in the German from Dutch course after it was upgraded to complete the A2 content.

For these statistics I never use any personal user data. DuoMe makes the compilation from its own database. Fresh data and new eligible learners are retrieved by DuoMe upon querying the DuoLingo user database.

[Corinnebelle]

I remember reading that Duolingo corrected the subscriptions on "add a new course" to reflect how many active users in a language per month. This means even if you for example do one lesson in a language in a month you would be classed as an active user. Of course most people do more than that. But I occasionally do a lesson in something else besides my TL.

It's good you got the L10 statistics as well because I haven't had a hundred day streak until recently yet I got past level 10 many moons before that.

[duome]

I have just added the L25 numbers to give you a bit more data to play with

[SweNedGuy]

I have just added the L25 numbers to give you a bit more data to play with

Thanks, some remarkable conclusions may be drawn from those. Some obvious ones also: courses with stories lead to a higher fraction of L25. Many learners reach L25 even before fetching the owl. Personally I haven't even reached a single L25 yet. ... too much cutting corners

Spanish from English
183472 Students
183472 Trees
146196 @ L10+ ········· 79.68%
83998 @ L25 ········· 45.78%
13248 Golden Owls ··· 7.22%
493 Golden Trees ····· 0.27%

[ARCANA-MVSA]

I would also like to point out that it can make a pretty big difference who is learning with these courses.

Younger people in particular frequently come to DL thinking to learn a language, dabble a bit, and end up losing interest, getting lost in their own lives, or something like.

This would certainly account for the abysmal performance of Korean, as many young folks end up learning the hard way that kpop simply isn't enough to keep them going. (Not saying anything bad about kpop - it's just a fact that there's not much of a demand for the language outside of that, and Korean is a very difficult language for English speakers to learn, so a high drop-out rate would be expected regardless.)

All this is even more prevalent given the fact that Duolingo is sometimes used in schools across the US (particularly for French and Spanish - not sure about any other languages), and here (in most places, if not everywhere) a minimum of two years studying a foreign language is required for graduation. This could include people who have no interest in learning a foreign language, or for whom the language they want to learn is unavailable. As such, there may be many users who use a particular course only for the very bare minimum, and abandon it (or possibly their entire account) as soon as school is done. (And given the sheer length of these trees, it's guaranteed they won't have gotten the owl before jumping ship!)

Clearly French and Spanish didn't take the same blow that Korean did when it came to the drop-out rate, but that might have more to do with the facts that 1) both French and Spanish have more trees (with other languages as base - many of which are quite short) than Korean, meaning more opportunities to gain a golden owl in those languages potentially with less work, 2) both are far easier to learn, and 3) both languages are far more in-demand beyond a singular media sensation, which means they are more guaranteed to attract dedicated learners.

Perhaps both French and Spanish would look more similar to Portuguese if they didn't have a combination of constant tree extensions and disinterested students to deal with.

(This isn't to say anything bad about younger people - especially as I am one of them. It's just an observation from some of my own experience, on Duolingo and in school. )

[Theron126]

I have just added the L25 numbers to give you a bit more data to play with

Thanks, some remarkable conclusions may be drawn from those. Some obvious ones also: courses with stories lead to a higher fraction of L25.

Another factor is that courses with stories are mainly the same as those which had Immersion, which back in the day was how people really racked up XP - both serious learners and gamers looking to farm numbers. That's how I reached high levels in Russian and Norwegian, but since those languages officially didn't have it you needed to know the hack

[Theron126]

It's interesting how few golden owls there are for these languages. For example, only about 8% of Korean learners on Duolingo have a golden owl. For the amount of time the Korean course has been on Duolingo, you would think there would be more. And the K-Pop fans would try to learn a fair amount of Korean to listen to their favorite songs, right?

I think what you're seeing is that the K-pop insanity has the opposite effect. Lots of people really do think it would be fun to learn Korean to listen to BTS or whoever is cool these days. But then they get into it, and they find out that Korean is hard, and they give up. Relatively few of them have the motivation to stick to it.

[ImEatingCookies]

It will be interesting to see what happens to the Ukrainian course. The number of learners has increased drastically during the last month, but most new learners are from the US, probably with very little knowledge of slavic languages. If I remember it correctly, the second largest group were Polish people. If you already speak a slavic language, have the opportunity to use Ukrainian and the need to do so, I'm expecting you to get that golden owl quickly. But the number of people who quit will likely outweight the number of golden owls.

[IceVajal]

Some just started to learn Ukrainian to get the flag and so show support. These might already dropped it.

[SweNedGuy]

Related article (in Dutch) on the popularity of the courses from Dutch. Active learner stats are just as bad as in the courses from English. The owl ratios are at the high side. An additional feature added is the number of learners having reached XP level 25 (those are 30.000 XP). Their number is significantly below the number of golden owls for learners from Dutch. That comes as little surprise, since many take the placement test and there are no stories in these courses. So why dilly-dallying to fetch the owl?

Hoe populair zijn de DuoLingo cursussen vanuit het nederlands?

[Siobhan688078]

Thank you very much for this excellent compilation - some of your explanations were news to me, after all this time! However, I don't usually spend a lot of time with duome, and I still don't really know how to navigate the site properly..... ^O

[Dana_Dany Danuta]

Hi, this is a very interesting statistic regarding the popularity of some language courses. I think it may depend on the quality of the courses.

Some are really very good, more frequently updated, well supported with moderators and have stories (German, French, Spanish).

I am surprised by the popularity of the Italian language, because it is short and has no history, but maybe immigrants contribute to these statistics, or there is simply a fashion for a some language!

[Siobhan688078]

Hi, this is a very interesting statistic regarding the popularity of some language courses. I think it may depend on the quality of the courses.

Some are really very good, more frequently updated, well supported with moderators and have stories (German, French, Spanish).

I am surprised by the popularity of the Italian language, because it is short and has no history, but maybe immigrants contribute to these statistics, or there is simply a fashion for a some language!

Why don't you just give it a try? Nothing to lose, and you never know, you might get hooked

[Sofia222677]

I am surprised by the popularity of the Italian language, because it is short and has no history, but maybe immigrants contribute to these statistics, or there is simply a fashion for a some language!

If by "history" you mean "stories", actually the Italian course does have them as well (there are fewer sets though). I'm pretty sure it's not immigrants, the number of them coming to Italy has been on the decrease for a few years now, if I recall correctly.
Italy is a popular holiday destination, and its art and culture are appealing to quite a few people.

That being said, I was also surprised the first time I checked the list of courses from English ordered by number of learners. (But over time, this turned into dismay; the 6th most popular course from English — and reportedly the only one of the few original Duolingo trees is shamefully neglected, it's hardly undergone any update.)

[SweNedGuy]

In the opening post I have added a few links to similar studies about the courses from Spanish and from French. Differences can be startling. Spanish speaking learners of English have a lower 'percentage of active learners' than French speaking learners. Moreover the percentage of French speaking English learners eventually fetching the owl is really high, whereas for Spanish speaking learners it is dreadfully low.
The English course for Spanish speakers is among the best on Duolingo, yet it is quite lengthy. Does this justify such different outcome?
Are French speakers more persistent at using Duolingo or better at languages? Does Duolingo better reach its target audience in France rather than in Latin America and Spain?

When you continue digging, you may find that Spanish speakers have more success with Portuguese than have English (or French) speakers. Portuguese vocabulary is a lot closer to Spanish, which may explain a lot. Yet Spanish learners are nearly almost as successful in Italian and French. The lousy learners stats on English from Spanish remain startling. The more obvious conclusion is that the audience of English learners must have a different profile than has the one of Portuguese, Italian or French learners.

[SweNedGuy]

I am surprised by the popularity of the Italian language, because it is short and has no history, but maybe immigrants contribute to these statistics, or there is simply a fashion for a some language!

If by "history" you mean "stories", actually the Italian course does have them as well (there are fewer sets though). I'm pretty sure it's not immigrants, the number of them coming to Italy has been on the decrease for a few years now, if I recall correctly.
Italy is a popular holiday destination, and its art and culture are appealing to quite a few people.
(...)

For generations Italy used to be a country of emigrants. Because of this you find Italians all over Western Europe, North and Latin-America. After two generations, their descendants get really mediocre at the mother tongue of their grand-parents. Why not learn the language of your relatives who stayed in Italy?
If the economy/democracy/society breaks down in the country where you live, it is reassuring to have a place to return to.