How Preply Built the World’s Largest Language Learning Marketplace

How Preply Built the World’s Largest Language Learning Marketplace Learning is inherently human, especially when you talk about languages, you're talking about cultural nuances. You're talking about human connections. You learn a language to connect with other humans. The tutors are at the core of everything that we do at properly. Without tutors, there's no properly Tutors are the reason that students progress. They are the ones generating amazing lesson plans and providing progress to all of the learners. Now that you have these AI capabilities, you are able to essentially personalize at the end level to every single user based on what they told you. We can generate a learning plan with the help of the tutor very specifically for that learner. Speaker 2 Welcome to Subversive, a podcast dedicated to sharing stories from the best consumer subscription apps in the world. We'll bring you lessons for how to grow your consumer subscription business, including insights and inflection points that led to exponential growth from leaders at category defining companies and innovative startups. Let's get into the show. My guest today is Simone Mizzi, who joined Preply as a Director of Product in 2021 and now serves as the company's VP of Product Growth. In this episode, I talked to Simone about the unique opportunities and challenges of building a human powered marketplace for language learning rather than a digital language learning app like Duolingo or Battle. About how the company's new user onboarding flow has needed to evolve as the company scales, and about how Simone expects AI to change how people learn languages over the next three to five years. All right, welcome Simone to the Subversive podcast. Really excited to have you on today to talk about Preply, which is one of the leading online platforms for language learning in the world and importantly, connects language learners with human tutors. So a little bit different than Duolingo or Babel or some of the other language learning apps out there. But before we jump straight to Preply, we'd love to start just with your background and how you found your way to this company. Speaker 1 Thanks. Very, very happy to be here. Listen, Triply is the second company that I've really worked at. I initially worked at an online travel agency based out of Bangkok called Agoda, part of the Booking Holdings Group. I started there in marketing actually and not in product. And after a couple of years in marketing, the product team actually came to me and said, Hey, Simone, would you be interested in in joining the product team? You seem to like building systems and stuff, so why don't you join us, You know, to the leap? It was, I guess, fairly different types of products than most people think about when you initially think about products, you know, you're like, well, I'm going to be a customer facing product and people are going to touch what I build and and all of that. It wasn't that you know, I was building marketing back end automation, which was a great entry into into product, which I then moved towards more consumer facing product and so on and so forth. We can talk more about that. And then after eight years there, John. John Prepley. Speaker 2 Yeah. So you were at this company, a GOTA for eight years and it sounds like you worked on a lot of SEO and SEM work in particular in the early days. How did some of that work and sort of seeing things from both the marketing and the product perspective help prepare you for a career leading product growth at Prep? Yeah. Speaker 1 So I mean, I, I, I like to think about, I think there's a few different types of product, right? You can think about you've got the B2C and the B2B product and within B2C, you're going to have the customer side and supply side or you're going to have customer facing or more back inside. And I kind of think of it as sexy product versus unsexy product, if you know what I mean, where the sexy product is kind of what you can touch, what you can show to your friends and family. Hey, look, this is what I built with my own hands. I, I was very much on the unsexy side of of the product building these automations like, like I was mentioning just just earlier. But I think honestly, I don't think that I would have started any other way because I, I generally believe it. It's taught me a lot. First, starting with systems thinking. You know, when you build a system that is supposed to automate keyword creation, adds creation on ACM across 50 different languages with hundreds of millions, if not billions of keywords with a system that has many, many different steps in between. You have to think through how step one of the system is going to be impacting step 15 of your system very early on. Otherwise, you know, you're, you find yourself having to, to rebuild everything afterwards, which is obviously not great. And, and you have to think at scale because when you have these hundreds of millions, if not billions of, of entities that, that you deal with in your product, you need to think at scale. And it needs to work in every single language. It needs to work for different attributes of, of this entities. And, and so you definitely have to think through that. And I think these are things that you need to apply in any type of product that you build, whether they're PLG, whether they're back end it. It really just teaches you that that that core, that core thinking, All in all, very helpful even to what I'm doing today, even though that was that unsexy part of of the product. Speaker 2 Well, I was going to say, sexiness is all in the eye of the beholder. Speaker 1 And. Speaker 2 So it's really just about the type of product work you love and, and I can relate because I started my product career at Ibotta. I was one of the first product managers there. And for most of my time at Ibotta, I was doing core product work, consumer facing. So I guess what you would have described as the sexy work, but I found myself gravitating more and more towards the growth product work specifically. So how to optimize funnels, how to improve onboarding experiences, how to deliver value right away in the first minute of the user's experience. So consumer facing, but maybe not as interesting to some people as building entirely new product features. And I, I, I found that it really just depends on what type of product work you love. And I think that's a good segue to July 2021. Why Simon took a leap of faith to join Preply You decide to join Preply. It sounds like at first you were reluctant to leave a go to because you had such a great experience there. But ultimately Preply convinced you. So how did they convince you to, to jump to not only a new company, but an entirely new continent? And how has your role evolved over the last many years as you've been at Preply? Speaker 1 Yes, I can remember exactly where I was when I decided, OK, I'm, I'm, I'm going back. But I think there was a bit of a mixture of of personal and and professional reasons. Like you said, I was very, very happy at a go to the team. There was was amazing, very happy with the products that that I was working on. And I had no intention to leave really until somebody from Prepley reached out and, and it sounded very, very, very interesting. And on, on the personal level, at the time I was 30, I was single. I had been at Nagoda for eight years, in Thailand for 12 years. And something in my mind was, OK, you're so happy here. If you don't try something new now, you never will. Which in and of its own wouldn't be a problem, but I felt like, OK, why not try something new or, or it's never going to or it's never going to happen. So that was part of the reason. Let's try something new. What is the worst that can happen? I'll just come back to Thailand, you know, and, and that is going to work out. And on the professional level, there's on. On the one hand, I guess I had and still have always, I think I would always have some form of an imposter syndrome where I was performing well at Atagoda. But something in my mind was, are you performing well just because you've been groomed here? Like would you be able to perform in, in another place? So, you know, try and challenge yourself a little bit and see that if that works out. And, and, and lastly, I think I just love language learning. You know, I'm French, we're talking in English right now. I lived in Thailand, I was travelling all over the country and communicating with people in Thailand, in Thai and since very, very young, I've always been very passionate about language learning and feeling that oh wow, this is actually, I can work there and it helps so many more people to do that. I think it was also a very, very excited prospect for me. And so take all of that together. July 2021, still in the pandemic decided OK, you know what? Let's go, let's do it. Speaker 2 Well, good for you. I mean, they say that the biggest regrets you have in life are the leaps you don't take. So you take this big leap. And as you said, I mean, language learning has played a huge role in my life. I've majored in economics and Chinese and college, spent almost half my time as a college student in Beijing and other parts of China. We were talking before we jumped on the podcast about your trip to Tanzania. And my wife has spent a lot of her life over there and learn Swahili. So language learning really can be transformative, and one of the things that makes Preply so special relative to some of the other well known language apps out there is that it's a human marketplace. So you're connecting language learners with human tutors. Pros vs. cons of building a language learning marketplace Walk me through some of the opportunities and challenges that are created by a human language learning marketplace as opposed to a digital app like Duolingo or Battle. Speaker 1 Yeah. I mean, I think the term human, I guess it's something that we're going to talk about a lot in in this conversation. And and I think to me, that's that's the biggest opportunity, right? Learning in its own is deeply and inherently human. The motivation that you need to learn something you get from another human being that supports you, that motivates you, the personalization, somebody to understand what is exactly your specific need feel that is going to be very different from mine. So creating that human connections that you know, is how many people do you know that tells you, oh, I took French in high school for four years and I went to France and I couldn't order a Cafe. You know, it's, it's all about that. It's all about learning, especially language. It's all about these human connections that you have. And, and this is the biggest opportunity because this is where real progress happens. And at the end of the day, people learn a language to make progress. And so when you put all of that together, it's the biggest opportunity of a human based marketplace when it comes to language learning. You progress faster, you create deep relationships with your tutor and with other humans. Yeah. Speaker 2 I think that point you made around motivation is so important because a lot of ink has been spilled on the efficacy or lack of efficacy of an app like Duolingo, just when it comes to the pure pedagogy of how quickly and effectively you're learning the language. But there's a separate component to it, which is staying motivated. Learning a language is really hard and it takes a long time, and having a human coach who can be there for you, not just to teach you the language, but to keep you motivated is so important. How Preply matches language learners with the right tutors And so this brings us to one of the core competencies of a platform like Prepley, which is matching supply and demand. And in this case, that's matching the tutors with the students. So what are some of the most important criteria that Prepley has identified for matching the right student to the right tutor at the right time? Speaker 1 That is the $1,000,000 question. A few weeks ago we were talking about everything that we do as a team, team and where we are going and and somebody. So, you know, part, obviously a big part of our strategy and growth is that yeah, matching students with the right tutor for them. And somebody raised their hand and say, hey guys, I've been at Preppy for four years now and I've seen this very same problem come up again and again for the last four years. When are we going to solve it? And my answer was, well, you know, there's companies like Booking or Amazon that have been added for the last 20 years and they still have not solved it completely. So we're probably still going to have this as a problem to solve for the next many, many, many years. One of the core challenge is what we were talking about just earlier. It's very humid. Frankly, it's extremely human. So if you think about travel or e-commerce, you know, it's very tangible. You have some very easy comparable parts. What is the star rating of the hotel? What is the size of the room? Does it have a swimming pool? Does it have a view, Yes or no? So it's fairly binary when you think about humans, it's a lot about the vibe, which is very hard to quantify. And it's not something that is that is binary on whether you're going to vibe with with a person or not. And every learner is very, very different. So what's very important to consider when you think about that matching is you're at a point in a journey of many, many months where you are helping the learner make the decision of their lifelong partner, you could say to, to learn a language. So how do you do that? How do you take all of that complexity and are able to create a great match? I, I think of it as something that I refer to as the Holy Trinity. It starts with the learner. You need to learn as much as possible about the learner. It cannot just be a learner that wants to learn English, that comes from France. It needs to be much, much deeper than that. You're French wanting to learn English because you have an exam in three weeks and you need to get better at the reading part because you've already studied with another tutor to work on that other part. So you need to go very deep in understanding what why is that learner here, But then you also need to have the same type of information from your tutors and the same understanding about what the tutors can provide to the student. So if you have all of that information from the student and you don't know if the tutor can help with that specific exam, it's not going to work. So there's a lot of work to be done in understanding which tutor is going to be able to help you at that specific thing that you're looking for. And then in the middle of that Trinity, you have the intelligence layer. That is everything that we do with our machine learning algorithms and AI to really match these two sides of the marketplace together towards the needs and ensure that we're going to create that long lasting relationship. Speaker 2 So what I'm hearing is, and this makes sense, you need to deeply understand the learner and not at a surface level like what country they're from and what language they're learning, but why they're learning the language, what level of proficiency they're already at, what level of proficiency they're trying to get to. Are they learning it for academic or career reasons? Are they learning it for travel? There's a long list of questions to understand the learner. Similarly, on the tutor side, you want to understand how many years of experience do they have tutoring? Are they a native speaker? Did they learn this language later in life? Do they have any sort of certifications? Their long list on the tutor side. And then it sounds like the hardest part is I love your analogy to travel. Since you spent eight years working at a Gota, you know when you're matching somebody to a hotel or to a particular activity as part of a trip, there are a set of structured fields that you can just sort of fill in and compare apples to apples. But with a human to human interaction, so much of it is difficult to quantify. So I'll push you a little bit on. Are there any specific metadata or characteristics of learners and tutors that you have found to be most important in driving successful matches? Experiments that significantly improved matching quality Or maybe another way of asking the question is can you give me a specific example of a matching improvement that Prepley has rolled out during your time there over the last four years and and sort of what sort of impact that drove? Speaker 1 Yeah. So impact, let's start with the impact because I think it's it's good to work backwards from from that in growth abruptly, we strive for impact that represents what we are really trying to achieve. And like we mentioned earlier, learning is not something that happens in a day. It's many, many months and sometimes years depending on where you start from. Yes. And so again, unlike travel or e-commerce where you book a hotel for one weekend, when you're trying to make an improvement on the matching, we try to measure it as how many long lasting connections have we created. So it's not about getting the user to, to pay for that first lesson and then go then this is not a success because we're not helping the learner to hit progress and to hit the value, right. So what we're really trying to achieve, what we define as success in growth is how many long lasting connections we we generate that we measure as a subscription. I would love to go even beyond subscription plus one month, 2 month, etcetera. But for now, we're, we're at the subscription that is already a great representation of commitment because you wouldn't commit to a relatively large ticket size to somebody that you don't feel is going to help you achieve your goals, right? So that's success. How we measure everything that we do in growth and we measure all of that via experimentation for please, extremely experimentation driven. So every change that we roll out is going to be experimented with and checked against our success metric. So now to to some of the specific improvements that we've made. We've run hundreds of experiments. So you're kind of asking me to pick my, my favorite child here. And I think there's, there's a lot of obvious wins. So I, I mentioned a couple, there's some obvious wins because I think obvious is good. You know, there's you've worked in growth for a long time as well. And you know that there's very often the most obvious things, all the things that generate the most value and maybe some about less obvious. One that I'm particularly excited about is how do you play back the information to the learners on the matching? So you can have the best algorithm, the best If we go back to the Holy Trinity, you can understand the learner super well, have the great data from the tutor, have the best intelligence in the middle. If the user goes through your funnel and does not understand that that intelligence is there and that the treater is there, it's all from nothing. So really explaining to the user, why is that treater good for you has been some of the best wins that we've had in in that matching, in that matching funnel. Everything that we do on the intelligence layer with the machine learning improvement. So as soon as we get more data from the learner, how do we integrate that in the way that we show the treaters? So a lot of that, a lot of wins we have gotten from this machine learning's improvements. And then there's also a lot more about UGC, specifically about reviews. So like we said, it's intrinsically human. We are matching two different people. And so for learners to see that other learners, other humans have had a good experience with that other human is something that is extremely valuable and gives a lot of confidence in the match, especially if that other learner is similar to you. And so going much deeper into the UGC, understanding what student like about the tutor and why. And again, playing that back to the learner based on what we know of their needs are some of the things that where we've seen some some great success in in improving matching. But like I said, there's hundreds of them. So it's very hard to pick my favorite shot. Speaker 2 But that makes a lot of sense and let me you said a lot there. So I want to try to break down some of the most important insights I heard and then let's make sure I got them right. So first you measure success based off of long term subscriber relationships with a tutor. So as a new learner on Preppley, you have what's called a trial lesson where you're able to do 1 lesson with a tutor to make sure that they're the right fit. But for you, that's just a leading indicator of success. Ultimately success is measured based off of a long term retained subscriber with that tutor. That's the success metric. And then what I've heard you say was, I'm oversimplifying a little bit, but there are three primarily primary sources of value you've found in driving up successful learner tutor match rates. The 1st is the metadata you're showing in the actual user interface to the learner about that tutor. So this could be things like the average five star rating of the tutor, how many years they've been teaching this language, what types of certifications they have. The second component is the machine learning layer that's happening on the back end. So it's stuff that the learner doesn't necessarily see, but it's driving which recommendations you're making to the learner. It's driving the order of tutors that appear in search, search results when they search for a tutor for a particular language. And then the third layer, if I'm understanding correctly, is user generated content. And, and I understood that to mean content that the tutor is creating like a video or a short bio about themselves that explains to the learner who they are. And it goes beyond the sort of ones and zeros of why they're a good match towards the vibe of like, am I actually going to be a good fit with this person in all three of those pillars, The metadata that you surface and the user interface, the machine learning algorithm that's driving recommendations and search results behind the scenes, and the UGC created by the tutor are all driving the perception of on the learner side that they're going to be a good match. Because even if they're the perfect match, even if you've done the math and you've determined they're the perfect match, if the learner doesn't perceive it that way, it's all for naughty. Did I get that right? Speaker 1 That's correct. The, the only thing I would maybe comment on is I, I don't know that I would necessarily qualify these three as pillars as opposed to some tactical example of, of what has happened. I think the pillars slightly higher level goes back a little bit more to the Holy Trinity that I was talking about earlier. And there's another very important component that that we can deep dive into if you're, if you're keen, that is the, the actual trial lesson experience where you know, at the end of the day, you can do a lot of things around the platform. So you can make prep play as great as possible. If the delivery of the lesson, if the experience, that human experience is not great, it's again all going to be for naughty because it would be the equivalent of you going to a gym that looks great, but all of the machines are broken, you know, so you're not going to go back to the gym. So yeah, I think they're good examples, but I wouldn't necessarily qualify them as the key pillars that we after as opposed to tactical examples. Speaker 2 Yeah, that makes sense. You described the Holy Trinity, your 3 pillars as the learner, the tutor and the machine learning algorithm that connects the two of them. These are the tactical areas of opportunity focused on. And I think your point around the trial lesson is a good segue into the second major topic I wanted to cover today. How Preply’s new user onboarding experience has evolved And so that is the news are onboarding experience which for any consumer application is important and it's only becoming increasingly important over time because human attention spans are getting shorter and shorter. So probably is now a fairly large company. As you've scaled and gotten more complex and you have hundreds of thousands of users and tutors on the platform and you're covering all of these different languages across all these different countries. I imagine that the on boarding flow has needed to evolve and adapt to support all that additional complexity. So what did the on boarding experience originally look like at Preply and how has that evolved over time? Speaker 1 Yeah. I mean, the initial on boarding experience was very, very basic. It was mostly based upon filtering basically and it's definitely been an area where we've invested a lot over over the last few years. And you know, we initially introduced a much deeper questionnaire. And just as a fun fact, you know me coming from travel where friction is bad, you just got to get to people, you got to get people to the hotel as quickly as you can and, and that's how you win. I was very sceptical about this ever providing an uplift and hold and behold the the win in terms of conversion was like nothing I'd seen before by introducing friction into into the funnel. Yeah. And I think it's been since then we've had dozens and dozens of improvement on this going deeper into the needs of the learner, like like I was mentioning. And what we've learnt at this point is there is such a thing as, as good friction for, for the learners. And it helps identify who are the high intent learners because probably is a paid subscription. Therefore, you need a certain level of commitment to to get into the funnel and, and, and become A and become a user. And so the good friction has helped us in better qualifying these higher intent users that are ready to put in the work to find really the best match for them, while potentially tiring out the low intent users that probably would never have converted anyway. And so it's this very fine line that we're constantly walking that is to find what is the good friction that we can introduce that generates genuine utility to the learners to find a better match and creates that long lasting relationship. Speaker 2 That makes a lot of sense. And this is something that's come up on a lot of my podcast episodes, this idea of short on boarding flow, long on boarding flow, good friction, bad friction. Where do you strike the balance? And I find that it's so dependent on the nature of the product and the target customer. And in your case, it makes sense, right? You're making a significant investment in finding and a tutor, a coach who can help you on this journey of learning a language. And in some cases, this is a very high stakes journey, right? It will determine where you go to college or what sort of job you're able to get or how how high you're able to climb in your career. So it's very important and it sounds like in your case a little bit more good friction in the form of asking the right questions and that onboarding experience has has been successful. Are there any specific experiments that you can again point to that have been big needle movers in terms of boosting new user registration rate, trial less and start rate and most importantly long term subscriber conversion or retention rate? Preply’s most successful onboarding experiments Yeah. So you know, I'll go back to what we were talking about earlier, right? We focus on subscriptions. Everything in between is a proxy, but there is not a single team in the company optimizing for visits, registrations, trials and and so on and so forth. Especially because usually what ends up happening is you focus on one thing top of the funnel, the other thing at the bottom of the funnel drops and you're flat at the end, right? And you touch yourself on the back because you've increased the metric when what you really want to achieve has not been achieved. So we obviously look at them from a funnel perspective that is OK. Is there unnecessary friction or something that can be improved here? But we would not claim success on something that increases registrations but not subscriptions at the end. And I think that's very, very important to the way that we think about product in, in growth. And, and I think I practically in general that is we optimize for the success of of the business and the learner, not for a vanity metric. Let's let's put it that way. Speaker 2 I agree. I'll just quickly say I think that's so important. It's part of why I often coach our clients to, if they don't already, have a very well defined activation metric. Make sure you understand what your activation metric is because unless and until you understand what that aha moment is that is causing you to understand why your product is valuable and why they should not only stick around but pay for the product. Then you can optimize for the wrong things in the onboarding flow, in the life cycle, marketing communications, and end up driving up vanity metrics that don't actually drive value for the business. Speaker 1 Exactly. And so, you know, thinking about it from from this line, some of the experiments that have been the biggest thing removers again, it's a little bit hard to to choose what I would talk about because if you think about the whole funnel product is very deeply embedded across every single stage of that form, starting from marketing automation going to all of the matching that we were talking about to the check out. But I think I'll talk a little bit about because that's the part that I'm the most excited about and also the probably the harder one to solve. You know, we were talking just earlier about the trial experience itself, like that human experience. And usually what you find in a lot of the platforms or the marketplace, you know, Booking optimizes for their platform. They don't optimize for what happens when you're in the hotel. Same for Airbnb. Amazon doesn't optimize for what happens when you get your product. And but for us, because it's so intrinsically human, everything that we do for a very long time, I felt we were optimizing what happens before. Great match, great experience, what was optimizing after, How do we help you with the next steps of your subscription and your learning, etc. But we were not really tackling that moment, that 25 minutes or 15 minutes that you spend with your tutor in that trial lesson that arguably is going to be the make or break decision for you. If you don't have a great experience, you're not going to subscribe. If you have a great experience, no matter what the platform is around you, you're probably going to subscribe to public. So it's kind of the harder because then you are working with humans, you're not working with a platform, right? But I think that has been a major unlock for us really tackling this part of the trial is an experience. Whether it is, how do we prepare the tutor better for the class? How do we help in the class, both the students and the tutor, to build a great rapport, to talk about what we know from the pedagogical perspective, plus talking with many hundreds of tutors, what matters for a great trial lesson? So how do you drive that experience in the class when you don't have control over it? It's been a very challenging and inspiring problem to solve and and we're still at it and I think it's going to continue being a major unlock in in what we do in that, in that acquisition file. Speaker 2 I mean, that trial lesson really is, I love how you called it the make or break moment, right? Because that really is the first time the learner is experiencing the core value promise of Preply. And in your case, that core value promise is only as good as the tutors on your platform. How Preply has maintained tutor quality as it scales And so that brings me back to another question, which is, you know, so many startups face challenges scaling just the consumer side of their business. But as A2 sided marketplace, you also have the supply side of your business. So how have you been able to maintain very high quality standards among your tutors as you scale from thousands to 10s of thousands to however many thousands of tutors you have on the platform now? Speaker 1 Yes. So maybe I'll start by saying that the tutors are at the core of everything that we do at properly without tutors, there's there's no properly tutors are the reason that students progress. They're the ones generating amazing lesson plans and providing progress to to all of the learners. And you know, a lot of them are full time and properly making a full on living with with the platform. So tutors are a really critical part of everything that we do appropriately, but indeed maintaining a high quality experience that it's, it's something that that we think a lot about. And when I think about one of the biggest challenge that we have to solve is, is going to seem very functional, but is actually really important is availability. At the end of the day, a tutor is selling their time and their time is very finite. There's only 24 hours in a day and seven days in a week. And when you think about most of our tutorings being cross-border, which is also what I find amazing with correctly, right, the vast majority of all the relationships we create are from people that are in 2 completely different countries. That adds a lot to the availability concern to the availability challenge that we have to solve. So from an acquisition standpoint, if you find the perfect tutor for you that you believe you're going to vibe with, but that tutor is not available at the time that you're available to learn, then it's as if this tutor didn't exist. Yes and similarly as the tutors as they do very often get more and more popular on the platform, their agenda start to fill up. And so maybe now your slot is starting to get taken by another student and you layer on top of that the very human components of tutors being human beings. So they get sick, they also have holidays, they also have things to take care of in real life or things that happen. So you mix all of that together. It makes availability one of the core challenges for us to to solve and ensuring that learners can't find the right tutor for them, but also maintain a the same pace of learning over time. Speaker 2 Yeah, that makes a lot of sense. I'm a big American football fan. I'm from Minnesota, so I'm a Vikings fan. And I listen to a number of podcasts, and they often like to say the best ability is availability because if you're injured and you can't play, then it doesn't matter how talented you are, right? You've got to be on the field. And so the equivalent here is making sure that the tutors are available when they need to be available for the learners to get value. And so that brings me to the last question in this section, which is just as you have scaled both the demand and supply sides of your platform, what investments has Prepley and your team in particular made to try to accelerate this idea of time to value, meaning getting the learner not just matched to the right tutor, but getting the student matched to the right tutor as quickly as possible? Tactics Preply uses to accelerate “time to value” Something I spent a lot of time thinking about, and maybe we should start by defining value because we're talking about it just earlier, right? The, the core value for somebody learning a language is progress in that language and communication with other, other people, which can take many, many months for it to happen if you think about like the real value of properly. So you kind of have to work backwards from that. OK, What is the previous stage of that? It's about learning something, feeling some level of progress, right. If you work backwards from that is having your trial lesson with the tutor where you have to feel like you're learning and this tutor is going to help you to achieve your goal at the end, that value that you're here to find. And if you move backwards from that, again, it's about getting that match earlier, as early as possible with with the tutor. And there's a few different ways that we're thinking about it. On the one hand, focusing a little bit more on that first part of that of that funnel that we just talked about reducing the functional blockers, you know, finding the tutors, you getting to find the best match based on all of the information that you gave us as quickly as possible Availability. So there's a difference if the best tutor for you is available a week from now than if it's available this afternoon at 2:00 PM, right? We're going to be accelerating that time. How do we make scheduling as easy as possible for you? And so those are very functional, let's say. And then there's maybe a little bit more of the emotional layer because we know that this eventual value can take more time. So that's how do you feel about the value that you are getting in the meantime. So how do we create that feeling of the learner that this is the right place for me to get that value? For example, by showing what a class looks like. We know that learners are very anxious. You know, when you think about it, it's crazy. You're telling people you're going to meet a stranger online in a language that you don't speak, like, speak about a nightmare for people. Yeah. It's extremely anxiety inducing. So how do we make people feel comfortable about what they're about to experience and not only the shooter, but also on on the platform and going back to everything we talked about on the matching and the personalization, how do you feel really that yes, that tutor, if I am trying to take an exam 2, two months from now and I see on the platform this tutor has helped 10 of the students just like you take their exam. I can't really taste the value that I am about to get from prep. So all of that together, there's not a single answer on how you accelerate the time to value and you kind of have to think through all of these, all of these different components. Speaker 2 I love how you started with defining value and it reminds me of how you emphasize the importance of on the metric side, driving long term subscriber relationships, not just try a lesson starts. Because similarly, if you define value as well, they have one tutoring session with a tutor. OK, they've gotten some value that will lead to one set of decisions. But if you define value as building a lasting, trusting relationship with a tutor and ultimately achieving their language learning goals, then that's more of a marathon versus a Sprint. And you will make different decisions that even if it requires a little bit longer to get there, it's still worth it if they're ultimately getting a lot more value. So I love that that's how the the prism through which you're looking, looking at it. How AI is changing how people learn languages Let's transition to talking a little bit about AI. Obviously, everybody's talking about AI right now, and having been somebody who's been a large part of my career in Ed tech, I think it's particularly interesting for products that help people learn. So I'll just start with, how do you think AI is going to change how people learn over the next three to five years, both broadly, but then more specifically within language learning? Speaker 1 Right. I think the true answer is nobody knows because things are evolving so quickly at unprecedented pace that it's it's very, very hard to tell. I think AI is absolutely changing the paradigm and making learning more accessible. And that's extremely, extremely exciting. On on the other hand, you know, I hear a lot of things about the, the doomsday scenario where we're not going to need to learn languages anymore because AI is going to translate everything for us real time. You know, we're going to have this thing in our ears or AI is going to replace the tutors. And, and I don't believe that. I don't believe that at all because I think like we were discussing in, you know, the opportunities of a human based marketplace, I think human learning is inherently human and you need that motivation. And especially when you talk about languages, you're talking about cultural nuances, you're talking about human connections. Like you learn a language to connect with other humans. So you're not just going to learn this with with an AI that is not going to have all of these. All of these nuances and on the other hand, I do believe that AI is gonna be an extremely powerful accelerant to learning, both for the student and what they can achieve in between lessons with the humans. But also for tutors. Tutors, the same way as we gain efficiency in our work through AI, tutors can do the same. They can create better exercises for the student because they understand them deeply and they can use AI to make their lives easier. So I think there's there's a lot to be excited about when it comes to the combinations of how human and AI work together in, in learning both in non languages, but especially in in languages. I guess I'm biased, obviously we're working in Triply. So I think yeah, there's a lot to be excited about in that combination of of human plus AI. Speaker 2 Yeah, I tend to agree with you. And I love how you pointed out that when you really boil it down, language is all about connection, right? One of the keys to, I'm getting a little philosophical here, One of the keys to human success as a species is that we are so good at working in groups. And one of the keys to us being so good at working in groups is the sophistication of human language. And so even if artificial intelligence makes it easier and easier from a purely transactional standpoint to learn a language or not even have to learn a language because you'll just have technology auto translating between humans, I actually think the demand for learning a language in order to connect at a deeper level with another human being will only go up because the world is becoming an increasingly lonely place and people are seeking more ways to find genuine human connection. So I really like that argument. I guess to get a little bit more practical. How Simon’s team is using AI to accelerate Preply’s growth So as a product growth leader, how are you and your team using AI both to accelerate growth within existing channels, but also unlock new channels like generative engine optimization or Geo? Speaker 1 When you operate on the scale that we operate, you have opportunities at every single step of your funnel to leverage AI 4. And you know, I think AI now has, has this big hype with the LLMS and let's call it the generative AI, but let's not forget about the good old machine learning part of the AI family, right? Let's, let's put it that way. And I think about the entire funnel of our, our growth funnel and AI is going to be extremely useful in marketing for us to create text, images, videos, scale marketing across many different geos. If you think about the matching process, now that you have this AI capabilities with much more context, you are able to essentially personalize at the NTH level to every single user based on what they told you. You can, you know, going to give a very tactical example, but the headline of the page can be very much specifically for that user. The learning plan. We can generate a learning plan with the help of the tutor very specifically for that learner when probably a year or two ago we might not have been able to go to go that much in in depth, right? And you keep going down the funnel, we can help the tutors to provide a better trial experience again, to communicate with the student. AI is great at translation. We can help students and tutors communicate. So I see so many applications all across the funnel, but it's also the case of those may not be all of the features that are AI branded, you know, and I don't know that I'm such a believer of putting a chat bot everywhere for you to leverage AI or across your funnel. I think there's a lot of smarter, way smarter ways that you can use AI to subtly but helpfully customize your funnel to the learner and based on what you know about them. So, you know, I think there's, there's a lot that we are doing and, and we will continue doing there on the growth funnel. And that's without talking about everything that is happening on, on the new frontiers that that we're pushing when it comes to learning with AI and tracking your progress with AI and all of the feedback that that we can provide you as a learner to help you progress faster. Speaker 2 You are looking for ways to use AI to solve real customer problems rather than the making the mistake that I think a lot of companies are making right now, which is just building the sexy, sexiest new AI tools they can. And then it becomes technology in search of a problem. And often times that can at the very least be a distraction and at worst, actually be counterproductive to the user in terms of accomplishing what they're using the product to try to accomplish. And then the next thing you said there was sometimes those applications of AI that are most incrementally valuable, at least in the short term, are not the really pie in the sky, you know, like generative AI chat bot. It's, it's just using AI to make the matching algorithm a little bit better or using the AI to make the experience for a new user feel slightly more personalized. And yes, there will be more and more interesting applications as time goes on, But right now you're just you're looking for value, and value doesn't always have to be some big new generative AI application. Speaker 1 Or at least not to the user it might. It might be in the background, but to the user it's simply value. Speaker 2 Yep, Yep. I think that makes a lot of sense. There's one other flavor that I'd love to ask about in terms of how you're thinking about AI, and that's in terms of hiring and training new team members. Why AI is changing Preply’s hiring process So I guess I'd love to step back and understand a little bit better, how is the current growth team structured at properly? How many people are on your team? What do those teams look like? And then as that team is expanding and evolving, how are you using AI to both ramp new new employees up to speed faster and help them be more productive? Speaker 1 So there's currently about 10 people on the growth team, so 10 different product squads and we've been hiring a lot not only on the growth squad, but across all of product. You know, I think initially when we started hiring and we AI was there, we started looking a little bit differently at the interviews and the case studies that we were having. You know, we were like, oh wait, is that person actually smart or did they just get this information from AI? So I think initially kind of like everyone, it was a little bit more of a risk of a risk averse kind of kind of behaviour. But then that quickly changed and it became right now my expectations are only higher. I know what you can get with AI. So if you just meet that ball, then you're not meeting the bar, right? Similarly, if you we now know that it takes 2 minutes in one of the many prototyping tool that exists to write a prompt to illustrate your idea, then why didn't you do it? You know, or why not? So I think essentially what's happened is it moved from more of being sceptical to now increasing the bar of, of what you expect in, in the, in the hiring process. Because the reality is you expect people to be using AI in the day-to-day work. So they should be using it during the interview. And that means that the quality should only be that much, that much higher, right. And I also found that I had to rethink my questions a little bit because, you know, I've been interviewing now for for many, many years and you tend to to use the same questions over and over again. You have a very good benchmark. And so I kind of had to rethink my questions because there's critical thinking and answering to the challenges on the spot is still not something that AI can can really achieve. It's not to really be involved to to assess the candidates or whether they would be able to demonstrate the strength in in role. I also had to rethink these these these questions. Speaker 2 What I heard you say was the bar has gotten higher because what you're able to do in a certain amount of time has shifted radically with all of these new AI tools. And so in the hiring process, you're needing to adjust the interview guide and the specific questions you're asking in each interview to account for the fact that people can now do a lot more with AI than they, they were able to do before. And you want to specifically test for somebody who is using AI on a regular basis, because if they're not doing that in the interview, then they're not going to be doing it on the job. And they, they need to be doing it on the job in order to be maximally productive. Absolutely. Lightning round Well, let's jump into our lightning round. So I always start with the first question, same first question. And that is what is the best part about working it correctly And what's one thing that you would change? Speaker 1 It's going to sound very cheesy, but I generally believe it. The best part to me about working appropriately is is it's people for two reasons. Properly is very mission driven. You know, you're helping people change their lives and I feel it attracts a certain breed of people that are generally passionate about what it is that they do. And when you layer that on top of exceptional talent, which I was very surprised when I had drawn preply 4 years ago, the company was much smaller than it is today. And yet the level of talent was already incredible. And that's only gotten better. So you mix the passion with the smarts and the talent. It's, it's just amazing. It's a pleasure to to come to work every day and and work with with these people. Speaker 2 Yeah, I mean, often that is the answer I get, and I've certainly worked in a lot of those environments myself. Quizlet was similar. I feel like education Ed tech in general attracts missionaries, not mercenaries. Speaker 1 It's a good way to put it. Speaker 2 Wonderful to work with talented people of any kind, but it's even more wonderful to work for talented people who really believe in the mission that they're getting up every day and working hard for. Second one is a language specific question. So you speak multiple languages at the very least. You've mentioned French, Thai, English. What is your favorite foreign language, idiom or phrase, and what do you love so much about it? Speaker 1 I gotta say I may be biased because I'm French, but I think the French win in the idioms. I think like I. Speaker 2 Agree. Speaker 1 With you. Speaker 2 Somewhat sarcastic, like a little bit biting humor idioms, I think. I think you do. Speaker 1 Exactly. And I don't know for some reason, but I think about idioms, I think a lot about like these things that my mom or my dad would say. And, and there's one that, you know, when you were a kid and you forget to turn off the light in in your bedroom, then in France, your dad would actually come in and tell you Sepa Versailles, you see, which means, hey, this is not Versailles where the light is always on, which basically tells you turn the damn light off when you go out. Speaker 2 I'm laughing because I was also raised by a father who, you know, grew up in a blue collar household and very much believed in saving money whenever you could. And the equivalent for me was when we were at dinner. And my brother or I would say, like, we don't like this. We want something else. You'd say we're not a restaurant. So if you eat what's put in front of you. OK. So switching back into more of a professional question. So you've mentioned that your Chief product officer at Preppley once told you there's more than one way to do product. So what impact did that have on you and how has that influenced your growth as a product leader? Speaker 1 Yeah, I think that happened maybe a few months after I joined prepply. And you know I just came fresh of agora. Like I said, I spent eight years there. So I was very much agora, Agora brained, let's put it that way. And Agora was had a lot of amazing things and was extremely quantitative in it's mindset while I arrived in prepply that was quantitative and also very qualitative. And so I, I had to sort of like rewire my brand and and when I heard this, I was like, OK, yeah, that's right. So I started, I guess reading much more, listening much more, whether it is podcasts, videos on Youtubes and etcetera, and also really digging much deeper into the other viewpoints. And you know, there's always going to be extreme viewpoints on on both sides of the spectrum of any spectrum really. And so instead of being on one side of the spectrum and sticking there, it was really more about, OK, let me listen to all of the size of the spectrums, understand what everybody has to say and why they're saying it, especially why. And then make make up your mind about it because there's definitely something out of what everyone is saying that you're going to agree with or that can supplement the way that you're currently thinking about things. And I think that's helped me a lot because it's evolved and added much more layers to how I thought about things. And, and then once you break that barrier, you just keep it going. So now I do that all the time. And I find this to be extremely, extremely helpful and really helps with my own personal growth. Speaker 2 I'm so glad you said that because I think a lot of the best, not even just product leaders, a lot of the best leaders at any tech company or or really in any organization, as you get more senior, more and more of your job becomes being in info for and like absorbing all of the best information, all of the best ideas, all of the new cutting edge technologies so that you can then distribute that to your team. And it can be hard to do because especially as a product growth leader, you're so busy, you're pulled in so many different directions. You're responsible for so many metrics. And so the temptation can be to just keep your head down and work harder, but that ends up leading you to a local maximum. And so carving out the time every day, every week to be learning and getting 1% better is so important because it leads to this compounding effect of being more, more productive with every hour of time you are spending on the actual work. A related question is how has your leadership style evolved as you've built a larger and larger team? Speaker 1 I think it's very much related to the previous question also. I reflect on on that a lot you know, about how do you how do you keep on growing? And when I look back at when I started leading teams maybe 10 or so odd years years ago, I think I was I was not really leading teams. I was leading an output. I was leading an outcome and I was doing it with the people with with arguably is not is not leading and especially it doesn't scale. You can do that with 123 teams maybe, but at some point it's just going to get out of hand. You cannot be in the detail of every single thing that that comes out. So I think the way that that everything has evolved for me is how do I make sure that the right decisions are being made when I'm not there and that everyone in the team is alive. So if you if I ask every single person on the team, where are we going and why? Very importantly, why then everybody should be able to give the same answer and take it in separate rooms. They should be giving the right answer, not the right answer, the same answer. Sorry. And if that happens, then I think it means that the right decisions get made when when you're not in the room and the outputs get that much better because eventually people working on their teams are the experts of what's happening and you're never going to be as deep in everything that's happening than than they are. And so by mixing this understanding of where we are going and why with their expertise of what is happening, we enter a very virtuous loop of of impact essentially. And I found that this is so much more gratifying for the team, for the people, for myself, and eventually leads to leads to better results. Speaker 2 Yeah, I love that. And there were shades of what you said there that relate to a lot of similar answers I've heard around. Well, you stop working on the product and you start working on the team and the team becomes your product. But I think you went further and said something I haven't heard before, which is creating the right conditions so that the the right or the best decisions are being made even when you're not in the room. Which of course makes sense because at some point your team is too large, you can't be in the room all the time. And then the next question is, well, what does that require? And I think what it requires more than anything is hiring the right people and creating the right culture, but then also providing a North star that everybody can be excited about and agree on. And so people can make their own decisions and know that they're moving together in the right direction. Speaker 1 Yeah, and I've had something to that that is the north tower is important. But then everyone being part of the conversation because I think and I, I can relate back to this to, to when I was working on a team, I absolutely disliked when my boss just came to me where you need to do that or this is the vision. I may not agree with the vision and I may have very valid reasons why I disagree with the vision. And, and so I think bringing everyone along in the conversation and hearing the words of the experts like we were just talking about earlier to shape that vision and the North Star make the whole difference in the world. That's the difference between you receiving a message telling you this is the vision, or you having built the vision alongside the rest of the two. Speaker 2 Yeah, it becomes sort of a collective decision making process around what that North Star is and IT and I would imagine at a tech company in particular, it's also a bit of a moving target because the right answer is changing as the market evolves. Last question, so to borrow from Lenny Richinski because he always likes to ask this question, what is one life motto or guiding principle that you would buy? Speaker 1 One of the things that I repeat both myself and my team and in my personal life as well, is statistically you're going to be wrong 50% of the time. So don't sweat it like it's OK to be wrong. It's OK to fail. It's not OK to not know why you're doing something in the 1st place or not know why it failed. So that part is not OK. But statistically you're going to fail 50% of the time. So just chill. Just focus on having the right hypothesis, the right analysis afterwards and and keep going. You're going to be wrong. You're going to keep being wrong for the rest of your life, so just keep going. Speaker 2 I love that, I love that, and there's so much liberation in that statement, right? Don't be afraid of failure. Focus on learning and growth. And as a growth leader, it's very apropos to have such a growth mindset. Well, this has been a pleasure, Simone. Thank you so much for coming on. I really enjoyed the conversation. Anything you want to say to the audience before we wrap up? Speaker 1 Here. Yeah. Thanks, Phil. It's been great. It's been great talking I think, but what I'd like to leave the audience with is Triply is hiring. We are we're doing really well. We are hiring like crazy around all possible departments of tech and non tech. So whether it is product, data engineering, design, research, marketing, you name it, we are hiring across the board. So feel free to reach out, look out for the roles that are out there and, and and come and join us. Speaker 2 I love it. It sounds like a great opportunity and having worked with Simone and his team myself for six months, I can say it's a great organization and a great culture. Well, thanks again, Simone. This this was a great conversation and I appreciate having you on. Speaker 1 Great. Thanks a lot, Phil.