Toucan: effortless language learning and a lesson in building truly empathetic products
Cute bird, or charging unicorn?
Great products don’t just solve problems, they also understand and address the emotional pain those problems create for their users. That’s why I’m excited to dig through Toucan, a Chrome extension that’s looking to address the overwhelming and stressful work of learning a language, essentially by partially translating web pages so that you learn in context without ever really noticing that it’s happening. If nothing else, I’m pretty confident that if you’re learning a language you’ll install the extension after reading this. I wish it had been around when I moved to Berlin two and a half years ago - maybe it wouldn’t have been such a frustrating and defeating effort to unlock the German speaking layer of the city. In a place where people switch to English as soon as you talk to them, full on immersion is actually out of reach and an immersion sidekick like Toucan could help grease the wheels. Toucan is also a good case study for how to sequentially unlock complexity in a product, and provides a tangible way to talk about the amazing future better language translation models can provide. Goodbye dry classroom learning!
Learning a language is overwhelming easy?
Anyone who has learned a new language after the age of 6 knows that it’s incredibly overwhelming. It’s kind of like the theory where you always get promoted one level past where you can actually do the job well - to advance in language learning you essentially always have to be reaching one step past your current capabilities. It’s also pretty widely accepted that immersive learning is the fastest way to go, which usually means going way past your comfort level and just being really confused until your brain picks it up naturally. Knowing multiple languages has many and profound benefits, but the point is that it’s difficult to stick with it long enough to get to the rewards. That’s where Toucan comes in, and to start I just want to show the simple interaction at the core of the product.
With this five second GIF I can basically show you the core functionality of the app. Toucan replaces words in context which allows you to learn immersively but without feeling like you’re being thrown into the deep end. In fact the GIF overcomplicates the core interaction, because most of the time you see the words without noticing or hovering over them. Essentially, if I were to say “Toucans are a species of __” you can guess pretty well that the missing word is “bird”, and your brain can do it automatically without you noticing. So Toucan fills in that gap with a word of the language you’re trying to learn, and voila! You start learning a language without noticing, plus you get this nice feeling of progress as you check off words you know and see a fun celebratory animation. The point is that Toucan is not only solving for learning a language, but also the associated emotional pain of feeling overwhelmed and stuck. That’s one of the reasons why I’m pretty confident when I say they have a killer product team working behind the scenes.
Sequentially unlocking complexity
Toucan was launched just 16 months ago but already has big ambitions and a lot of buzz, after a 4.6M in funding (including a 3M seed round this September) and currently up to 30,000 active installs according to their founder Taylor Nieman. Immature product leaders would likely build lots of features and additional functionality into the core experience, except that by doing so you’d almost certainly remove this lightweight feeling that addresses the typically overwhelming feeling of learning a language. In other words people use products in different mental states, and good product work needs to take that into account. Tax return software has historically been a good example of where this can go wrong - you take people who are already confused and out of their depth and throw a huge number of features at them in a complex UI. It’s not that the software isn’t useful, it’s just not built empathetically. In Toucan’s case even a busy settings page could put people right back in that space of feeling overwhelmed by learning a language. Instead the team is peppering new features across the web like easter eggs to be discovered, for example this little game you can find if you go to a Wikipedia page where you can take a quiz of the words you’ve seen recently.
The core promise of Toucan seems to be that learning progressively in context can make an overwhelming experience feel effortless, and they apply that same insight to how they release new features and design the main focal points of the product. Simple, uncluttered, hardly noticeable. Check out the extension toolbar dropdown. None of the advanced features are mashed into that page - it’s up to the user to find them as they go about their day:
Now to be fair, Toucan isn’t so massively complex yet. It’s not like they’ve managed to fit JIRA or Salesforce level complexity behind a three button toolbar. The point is that they seem to realize that the feeling of ease is the core emotional promise of the app just as learning a language is the core utility promise. You discover the new features in the context of your continued use of the internet. Time will tell if they’re able to maintain this promise as they continue to expand the product’s functionality; I would see it as the core capability they can strengthen to continue distancing themselves from apps like Babbel or even full-on language courses which might be more intense and structured, but suffer from huge churn due to the emotional pain of “sticking with it”.
Monetization: an opportunity to take the promise further
Given the apparent focus on ease, intuitiveness, and effortless functionality, you’d expect the revenue model to be baked directly into the main value moments. And when I think about the value I get from Toucan, it really comes down to the number of new words and phrases I now know without having expended any effort. That’s the magic. Sometimes I even hover over a word I know so well I’m tired of seeing it and hit this little checkmark to get that hit of dopamine that comes with this fun celebration animation.
In contrast to the rest of the product, this is an area where Toucan delivers the value but misses the details that would deliver the emotional benefit. I don’t actually get a feeling of progress from the product, despite the fact that I really am progressing. I check off words when I know them, yet when I go to my dashboard it shows me “words translated” which is basically just the number of words I’ve seen. There’s nowhere I can see the number or list of the words I now know. To be clear I’m not suggesting I think every word needs to be checked off in order for the user to progress (that would also undercut the value of being effortless), but I would assume that behind the scenes there is some model that looks at how often I’m exposed to a word without needing its translation and using that as a metric of “learned.” Whatever the right ways to track it, the feeling of progress I get from the animation when I check a word isn’t carried through to the rest of the experience. And if “checking” words is going to be such a prominent part of the experience, that feedback loop of progress shouldn’t be lost entirely.
It’s not just the core user benefit that revolves around storing value in the product via progress towards a new skill. Being able to cleverly leverage that data into more and more uses is also key to the incredible things I believe Toucan can do in the future. For example almost everyone told me to read simple German articles as a way to advance, which I did but frequently would get so turned upside down that I’d give up halfway through and feel some not insignificant failure. Imagine if Toucan could tell me which articles I’m most likely to understand, or better yet apply this immersive learning in reverse to make any articles accessible. By using my “known” words and phrases Toucan could translate back into English the things I’m not likely to understand so that the resulting page is something I can actually tackle. A usually defeating moment, solved by Toucan.
The advance in natural language processing (NLP) algorithms is pretty clearly the most important technological shift that Toucan is planning to ride, and indeed the implications are amazing. Right now their translations are often a little bit off given a particular context, but that’s a problem that will be solved very soon by relying on more advanced models. And instead of just contextual learning of single words, Toucan will be able to translate sentence fragments and essentially build a massive dataset of how people progress through an immersive language learning journey. It seems pretty likely then that the value they will be able to deliver to a user will be a step change once a user hits a certain amount of stored data, so instead of charging a user for acquiring stored value why not charge them for what they can do with it? Open the floodgates to progress, but charge for quizzes, inverse translations of articles in other languages, maybe even down the road for the ability to hook their Toucan data into more structured coursework as a way to eat Duolingo’s (btw that other cute language learning bird is pretty quietly a >2B company) lunch. Even a traditional in-person tutor platform might benefit from a Toucan integration as a way to build immersive learning into the journey and feed data back to the user and tutor. Using Toucan could be to all immersive language learning as an instant read thermometer would have been to Goldilocks - not too easy, not too overwhelming, just right for your current stage of the language learning process.
There are likely many reasons they’ve started with revenue models which don’t depend on the stored value a user has in the product, not least of which is limiting initial growth by putting valuable features behind a paywall. Their “own a word” feature which allows people to essentially put their name or brand on the translation card for a given word is a clever growth trick if nothing else because they realized that people love sharing fun things they own, and indeed many influential people have Tweeted or shared words they acquired or were given by Toucan’s team. But it actually makes that core experience for the user more cluttered and a little bit more overwhelming (just look back at the GIF above), and while I do agree with Kushaan Shah that it’s a clever and potentially disruptive approach to advertising, I’d also argue it’s a slightly disruptive approach to their core promise to the end user.
The same is true about the concept of subscribing to get new “packs” of words - essentially focused sets of words around a given topic like “meet the parents.” I do understand the intention to allow users to focus on specific jobs to be done for their language learning journey and I do think it’s a fun customization for power users who want to personalize things or maybe to increase virality between people going on a trip, but it doesn’t feel aligned with the idea that for the bulk of users Toucan should be this nearly invisible service running on autopilot to help you learn. And if a user joins Toucan and already knows he/she wants to “meet the parents”, why build friction into the ability to focus the experience around that particular acquisition of stored value? If stored value is Toucan’s currency to build a transformational business, why not make it as easy as possible to grow their stockpile? Again I suspect these early revenue streams are intentionally somewhat perpendicular to the product value exactly to maximize early growth, but given how the focus and emotional benefit of the product I’d be worried about waiting too long to pivot into a more intuitive revenue model.
Toucan takeaway: cute bird, or charging unicorn?
I have to admit I’m a big fan of Toucan because it represents so much of why I love what I do. This seemingly simple browser extension is not only a deceptively powerful way to learn a language but also to see into the future of how language translation technologies will change our lives for the better. The question is, with so much work left to do between now and then, will this be the company that unlocks that future? Is this browser extension just the beginning of an AR world where Toucan is always by our side or in our ears or in our AR glasses, silently turning previously distant skills into joyful immersive learning? To put it another way, is Toucan a good pick for a Google acquisition in a few years? :)
There’s a truly abstract question here which some have asked which is whether you could imagine a future in which a Toucan-like experience could help us learn history or science and so on. In the goldilocks example I suppose I could imagine Toucan taking what you know and using that to make sure you’re only ever one step past your current ability, but it’s just hard to imagine that future fitting into this same immersive experience where you don’t notice it’s happening. Just hovering on a given word to get the historical summary from Wikipedia feels like a very different product, and the world in which Toucan actually replaces the idea you don’t know with an intuitive translation from Wikipedia seems even less likely. I also personally wonder if the total addressable market for language learning isn’t so big as to make that question interesting but somewhat unnecessary for a bullish view of Toucan - with billions of internet users, a fraction of which likely want to learn a new language, it seems believable that the market for this kind of no-brainer experience would be huge.
The main question I would argue is where is the moat? As advanced NLP models are essentially commoditized, the ability to create a Toucan knock-off will seemingly only get easier over time. There are lots of competitors already in the space (Babbel, Duolingo, Busuu) who could create a simple version of this browser extension as an add-on tool to their structured courses without too much effort. So one clear moat would be if Toucan successfully builds a community driven human-in-the-loop ML model where explicit feedback from users and from their “expert community” can be used to dramatically improve a learning process for new users. If they can build best in class models that don’t just train on single translations but actually how people best progress through a journey, then they’re really leveraging unique data, but it seems fair to say that’s a pretty big technical challenge that only time will tell if this team can crack. But perhaps the biggest moat to competition is the core ethos behind the products in this space - where other apps have taken the “eat your vegetables” approach of being ok with high churn as people buy an intention to learn but rarely follow through, Toucan is building a deceptively simple product that is actually betting on the fact that advances in technology will make these manual course-creation products obsolete in the years to come. They’re building towards a future in which churn will be much lower because people hardly notice they’re actually learning. A future where we can have our Kuchen and eat it too. And given how painful and overwhelming learning a new language is, if they can maintain the same quality and empathetic focus they’ve shown so far I think we might be watching this tropical bird start what will be a pretty amazing flight.