How Does Duolingo Make Money?

For five minutes every day, when I can pull myself away from my Pokemon, I go on Duolingo to practice my French. It is slick and it is free. You may have never heard of Duolingo, but it is no less than a revolutionary step forward in a woefully underserved market. It is likely the cutting edge of crowdsourcing, and that is because it is the latest project of one man, Luis von Ahn.

How Does Duolingo Make Money

Image: Pola Esperanto-Junularo

Luis von Ahn

I’m a huge believer in putting my conclusion right up front, but for once I can’t. Because Duolingo is a not merely a vision, it’s a journey.

You’ll like Luis von Ahn. The thirty-eight-year-old Guatemalan was a professor at Carnegie Mellon and even if you don’t use Duolingo, he touches your life almost daily. Every time a website offers you a squiggly “are you a robot” identity check? That is him. Not him in your PC screen dotting the pixels, but it is his slightly misshapen love child.

How does Luis von Ahn make money

Image: ITU Pictures

To give you a quick sense of the kind of person he is, here are the first ten rules in the original employee manual at Duolingo:

  • Say “Good Morning, X” to everybody when you arrive.
  • By implication, you must arrive in the morning (before 12:00), unless you were at the office the night before working past midnight (“the Android exception”).
  • Loose lips sink ships.
  • Loose tweets sink fleets.
  • Loose pants make friends.
  • Be nice to our users.
  • Do not eat other people’s lunch.
  • Hector is a co-worker, not a toy.
  • It’s ok to bring doggies to the office, but they must be house-trained. If your dog pees, they can’t come for a month, if they poop they can’t come for 2 months. Reptiles are not allowed. Dinosaurs will be considered on a case-by-case basis.
  • All suspicious activity must be reported to Franklin.

ReCAPTCHA the Castle

CAPTCHA is a backronym for Completely Automated Public Turing test to tell Computers and Humans Apart. The principle behind them is simple: offer up a task that people find easy and that bots find hard. What are these bots up to? On the benign side they might be creating sock puppet Twitter accounts to make someone appear more popular. Less benign is the use of bots to hack systems.

My WordPress site has had over forty malicious login attempts. They are likely trying common default passwords, and they gain access to other people’s property all the time. Make your passwords hard to guess guys!

Luis von Ahn didn’t invent CAPTCHA the concept, but he did revolutionize it. He posed the insightful question if 200 million individuals are using this daily, and each expends 10 seconds of time, are people wasting 500,000 hours of their lives on this?

If you’re wondering 500k hours is 57 years being used up EVERY DAY.

This thinking led to ReCAPTCHA – instead of getting people to identify something random, how about getting them to identify something useful.

ReCAPTCHA makes money

Image: Tony Hirst

Google Books and Google Street View

As of October 2015, Google have scanned almost 30 million books. This is not all the books out there, which Google estimates at over 130m, and likely never will be due to copyright issues.

Of the digitized books over 30 years in age, 30% of the scanned text is illegible. By presenting two words to users, one that the system knows, and one that it doesn’t, ReCAPTCHA can test whether or not you are a bot, and gain confidence regarding the spelling of the word it does not know.

Back in 2010, 350k websites were using this technology, and it was deciphering 100m words a day, or 2.5m books a year.

Google acquired ReCAPTCHA in 2009. I’d love to tell you how much they paid but it was not disclosed. So far even Oracle is yet to find a way to claim this one.

You may have seen a newer version of ReCAPTCHA with images of street signs and door numbers. Yep Google are using this system to improve the depth and accuracy of its Street View. Anywhere an image exists that can be usefully turned into text, Luis von Ahn made it possible.

Crowdsourcing

Is this innovative crowdsourcing or is this surreptitious slave labor? Since bot checks are essential you should be able to guess which way I lean.

Duolingo, Luis von Ahn’s follow up project, posed a new question to his PhD student Severin Hacker: How do we get 100 million people to translate the web into every major language for free?

Duolingo is simple

I hate the term obvious – as a rule it’s only valid in hindsight. Truthfully, the only lesson hindsight teaches is I wish I had done it sooner!

Duolingo has over 110 million users, each of whom translate sentences in order to learn a foreign language.

The Learning of Languages

I did okay at school. However, one blemish was an A grade in my GCSE French (think high school diploma level). The top mark in the UK for many qualifications is A* – yep with that silly little star by it.

A group of us were talking to our French teacher, and one of them joked about my grade. She said “Yea, he had the vocabulary but it never quite clicked for him”.

People in Anglo Saxon countries are often not very good at second languages, especially compared to say Sweden, where most take up a third language. I have a limited competence in talking about cultural differences and causes, but I am going to say language is generally not taught well in schools. Part of me suspects non-English speakers benefit from the way television and music can be a rather good teacher.

I liked my French teacher, but my GCSE in French was incredibly heavy on grammar. There was a repetitious focus on the difference in spelling verbs depending on whether the subject was I, or we, or a group of guys, or a group of girls. Only later did it become clear that many of these forms were spoken identically. I mean fait or fais – for someone only needing enough to travel, who cares?

Most people can get away with not knowing these nuances. Language you get at schools typically teaches what is easy to examine, not what we are likely to use. This creates road blocks because what you use inevitably forms a motivator or demotivator for further learning.

I attended Cantonese (Chinese) Sunday school until 15. The teaching of Chinese is even worse than French. The core of it is memorization of characters.

duolingo chinese

I had a search of the internet, and came across quite a few tattoo lists of Chinese characters and their meanings. But bad news guys, uh, as a rule they didn’t mean what they said! Think of it like simple words having one syllable. Complex words may have more. Its rare for the fulsome spiritual words people want to be tattooed to them to be as simple as Man and Woman.

Here’s a stupid story. We had some Italian exchange students visit my school when I was 14. One of them asked me how to write something basic in Chinese, so I did. Her friends got excited “Write my name! Write my name!”

They were cute girls, but I had no idea how to write Marcella in Chinese. Generally, these sort of transformations are quasi random based on similar sounding syllables.

I did what any guy in that situation would do. I wrote out a load of nonsense and left them happy. I really hope none of them are walking around with Sickly Green Goldfish written on their arm.

Learning for a Purpose

The optimum way to learn a language is living for a year in that country. It is immersive and you learn exactly what you need.

Moving backwards from that, the next best is a knowledgeable and charismatic person providing one on one tuition. This teacher likely costs over $50,000 dollars a year.

I have discussed how language is taught at school. It is one person teaching a class. Let’s say even that is too costly. An alternative is an online course. Rosetta Stone is an excellent product and used by the United Nations and many multinational companies. It has millions of users and costs over $500 a year. After that, the price point and effectiveness decline in tandem.

The world has over 1.2 billion people trying to learn another language. Most of them cannot afford $500. Luis von Ahn and his team took something that was a problem, and turned the problem into a solution.

duolingo will you dance with me

In my post on WhatsApp, I write about how Time is Money. Luis von Ahn is the leading authority in helping people pay for something with their time, and the original business model revolved around giving users language training in exchange for their time.

According to Luis von Ahn, the cost of translating Wikipedia into Spanish might cost $50m. At this price it simply does not happen. With a 100k Duolingo users, it is possible in 5 weeks. With 1m users, it can happen in 80 hours. It seems fair to describe this as a revolution in translating languages.

Duolingo

Duolingo offers short lessons consisting of a series of small tasks. Translate this sentence. Read out this sentence. Identify these word pairs. Mostly I ignore the gamification features, but they do form a large part of the product.

By translating the same text many times and using machine learning, the level of accuracy it delivers is extremely close to that managed by a professional human translator.

What gets fascinating with the question How Does Duolingo Make Money, is that having found an excellent form of monetization, they stepped away from it. Although Duolingo continues to translate articles for websites like Buzzfeed, it made a decision to not take the company further in that direction. The problem was simple, if translation became the purpose, it would harm the language lessons it provided.

The speaking exercises do not make money, but they are integral to the teaching. Cut them out or leave them in? Duolingo the teacher could operate with a small staff. But translation is direct sales driven. Add 200 sales people to a staff of 60 or try and remain a small entrepreneurial community?

Duolingo decided its purpose was teaching, not translation.

Instead Duolingo has developed a $49 language test, to compete with the many $200 language tests out there. It is computer science led, and therefore unlike typical language tests. It explores your level of competency in ways that indicate your overall ability. It can be taken at home as long as you let a webcam film you the whole time. This test is now accepted by many prestigious universities.

duolingo stages

How Does Duolingo Make Money?

  1. Provide a gamified language teaching application for zero cost to the student.
  2. Gain over 110 million users.
  3. Generate some revenues by translating websites and books.
  4. Cover the rest of its expenses by offering a low cost language proficiency test.
  5. Keep gaining more users by being focused on their needs as students.

Duolingo isn’t perfect. You likely need some basic language knowledge to start with. It mostly avoids the grammatical intensity of many other methods – great for me but less so for professional purposes. But it is awesome. It makes just enough money to pay its small staff of around 60 people. Its purpose is to teach and to translate almost costlessly.

Duolingo has opened 110 million doors, and I applaud them for their achievement.

One last dumb story about Duolingo. Early on, they hired a design firm to help develop the product and their identity. The co-founder Severin Hacker said “I don’t much care what you do in terms of style, but please don’t use green. I hate green.”

As you can see they chose a green owl mascot just to annoy him!

Interested in why Credit Karma is the closest thing to a cat that has learned to love mice? Then read on here.

EDIT: Duolingo have started showing some app download adverts after you complete a round of questions. This makes me a little sad as it suggests the testing business model has not quite taken off. The launch of their chatbot language learning assistant is interesting as of course I write about that in my post on Whatsapp.

Yuen Lo

One Comment

  1. So in summary Duolingo makes money only from the $49 test and advert pop-up? How else can they make money

Leave a Reply

Your email address will not be published. Required fields are marked *