During COVID-19, we can’t really go to the gym. So many of us are working out in our (new?) home gyms. Is that making us fitter? And, can an AI coach motivate us?
In this episode of TechFirst with John Koetsier, we chat with Freeletics CEO Daniel Sobhani about what we’re doing to stay fit during Coronavirus, what’s working, and how men and women are reacting differently … including with regard to mental health.
We chat about:
- the biggest ways fitness changed during COVID-19
- what are people doing differently
- how Sobhani’s AI coach for fitness works
- how the AI coach compares to a human coach?
- how the AI works, at a high level
- what’s different about working out at home?
- how meditation is up, and how men vs women deal with mental health differently
- what people are doing with regard to investment in new equipment
Listen: AI, fitness, & home gyms
Don’t forget to subscribe to TechFirst with John Koetsier wherever podcasts are published:
- Apple podcasts
- Google podcasts
- Spotify podcasts
- And multiple other platforms … see them all on Anchor
Watch: AI, fitness, & home gyms
Subscribe to my YouTube channel so you’ll get notified when I go live with future guests, or see the videos later.
Read: AI, fitness, & home gyms
John Koetsier: Can AI keep us fit even when we can’t go to the gym?
Welcome to TechFirst with John Koetsier. COVID has completely changed our fitness routines of course. I used to be in the gym 6 times a week. I’m not super, super fit, but I was at least there. Now of course, we’re all working out at home, not going to the gym, but it’s also changed our mental health.
To dive in and explain more, and to see if an AI coach can be as good or maybe even better than a human coach … we’re joined by Daniel Sobhani, CEO of Freeletics. Welcome, Daniel!
Daniel Sobhani: Hi John. Thanks for having me on the show.
John Koetsier: Excellent, looking forward to it. Let’s start here. We hear about the “covid 19” right? The extra weight that people are putting on during coronavirus. What’s the data say? Are we getting fitter or are we getting fatter?
Daniel Sobhani: Well, that was really one of the interesting questions that we also wanted to figure out. And what we really have figured out, that actually — and that was really surprising, that 70% of the people really said that it was rather easier for them to stick to a fitness routine, start with working out now during quarantine. So the data says on average it has become better.
John Koetsier: Wow, that’s counterintuitive. I mean like we can’t go to the gym. I mean, I really missed my routine that I had. Every lunchtime I would go to the gym and I’d do some workouts and stuff like that. And now I don’t have that, I’ve had to invent a new routine. How’s it working for other people?
Daniel Sobhani: So what we found out, and maybe one step back for those that don’t know Freeletics. So what Freeletics does is, we’re the leading European fitness app, we have now about 50 million users and we use AI to provide hyper-personalized fitness coaching and mental wellness support. And what we have really figured out, the issue here was that, you know, since 30 years we have been told that if you want to become fit you have to go to the gym. If you want to become fit, you have to do a grueling workout for 50 minutes, 6 times per week, I mean, as you did.
The reality is you don’t necessarily have to do so, and the real benefit of working out at home is really the lower hurdles that you have. You don’t have to spend the time commuting there, you can just get up and start. You don’t have to get yourself in any type of shape to go there or anything like that. So super low hurdles. And if you use tech, if you use an AI trainer, it can make the workout so much more efficient for you. They’re typically shorter, mostly even more efficient, and that really in the combination turned out to bring the data where the data is, that it was easier for most people.
John Koetsier: Interesting, interesting. I mean, the one thing that I really enjoyed about going to the gym is that when I was there, I was at the gym, right? I wasn’t at home. I was in a different environment. Eventually I sort of recreated a place here for that as well, but I guess the AI coach can help you in some ways to, I don’t know, have some external reinforcement of what you already know you need to do.
Daniel Sobhani: I think what you said is absolutely right. And really the reality of the industry is there’s no one size fits all. So there will always be that person who thrives better in a gym environment, who doesn’t want the comfort of ‘I can just start at home,’ but really needs to have that barrier ‘so now I’m at the gym, so I’m really gonna do it’ and, you know, ‘I’m meeting up with my friends or something like that, and we’re going through this together.’
So there’s really not this one size fits all.
And I think that also for the future, despite this incredible surge in interest and acceptance now for digital fitness solutions, the stationary gym will always be a part of that experience.
John Koetsier: Yeah.
Daniel Sobhani: It will just be enhanced by technology.
John Koetsier: So I want to get into the AI and how it works, and what technologies you’re using and how they help. But one of the things that you surveyed was in terms of mental health, but also in terms of the intimidation factor of going into a gym if you’re not super fit.
How’s that changed? Is that a good thing that people who wouldn’t go to a gym before, now they’re working out at home because they couldn’t go to the gym they didn’t feel fit, they didn’t want to go there until they were fit?
Daniel Sobhani: Well, what you described is actually really one of the things that we’re somehow screwed about, like this gym culture before. Because, I mean, what’s the point in having to — at least like your perception — having to be fit before you go to the place where you are supposed to become fit.
John Koetsier: Yes.
Daniel Sobhani: So that was one of the real challenges before. So that’s definitely an advantage of being able to do this at home, you can turn on the camera, turn off the camera, you can do all of these things. But what we’ve also seen next to that advantage, the accessibility advantage, is really a trend towards taking better care of your mental wellbeing.
nearly 80% of the people said that now they are more interested in mental wellbeing
And almost 80% of the people — so a higher percentage than the actual fitness part, where 70% of people said it’s easier for them to do fitness — nearly 80% of the people said that now they are more interested in mental wellbeing and they are more interested in how you can, well, for lack of a better word, “manage” your mental wellbeing.
And I think that that’s really a very beneficial trend for the people per se.
John Koetsier: It is super interesting, right. I won’t go to the gym until I’m fit. It’s like, I won’t go to the doctor until I’m healthy. It doesn’t make a lot of sense.
Let’s talk about the AI pieces. You’ve got an AI coach for fitness. How’s that work?
Daniel Sobhani: Well, I mean, let me start with what’s it supposed to do. And really in the simplest way, it’s supposed to give the users a workout that is absolutely tailored to them. So, to physically how they reach the goal, but also, you know, psychologically what they would enjoy, or to their equipment.
And so really the AI coach does three things. First of all, it enriches data, because really the basis for all good workout programming is we know you. The issue is most of the people don’t really want to go through an extreme intense questionnaire, and even if they do, some of the questions they can’t answer. Like if we ask them, ‘What’s your one-rep max on some exercise?’ they’re like, ‘What’s a one-rep max?’ So what the AI does is really like from the base data that we get from you, we combine this with the knowledge and the experience that we have from our almost 50 million other users. And so we can enrich your profile and we can, with a very high accuracy, predict what type of exercises you will be able to do, or not, with what weight and what repetition range and all of these things. So that’s number one, we know you better than probably you know yourself in that situation, at least for most of the people.
The second thing is then this data gets taken and the AI starts to prescribe your workouts. And what it does there, it gets boundaries from a training perspective, you know? So there are certain programming principles that are in there, and within these boundaries it tries to give you a workout that accomplishes one, it has a certain efficiency if you do it, like from a physical standpoint.
John Koetsier: Mm-hmm.
Daniel Sobhani: You enjoy it. You know, there are people who like to really sweat and train very, very hard. There are people who don’t. Where are you on that scale, for example? And [two], to maximize your retention. So, you know, what are the things that make you come back?
And then really the third thing is, and this is where it really becomes interesting, is it learns and it learns based on two main things. One is you can give it direct feedback. So feedback from, ‘I would have wished the workout to be shorter,’ to ‘It was great!’ to ‘This is the equipment that I have,’ or ‘Today I feel…’ in such a sense. But it also learns from your performance and your behavior. So depending how you perform in the certain exercises, depending what you do and what you don’t do, it enriches this data set that I mentioned before. And because of the data set, again, the workout gets better. And so really our goal is to get people out of every workout session with a positive feeling. And so we ask people afterwards — and to give you an example here, the AI can pull out of about 3.5 million options, and by now we have …
the AI can pull out of about 3.5 million [training days]
John Koetsier: 3.5 million options of what?
Daniel Sobhani: Of training days. So you get in every day, the AI gives you a new — so every day it’s adapted, your training plan. It’s not monthly, it’s not weekly, you get the daily training.
John Koetsier: Wow.
Daniel Sobhani: Every day there are 3.5 million options — meaningful options, I mean, we could do random stuff — but 3.5 million meaningful options the AI can pull from, and by now we are at about 85% affection rate. That means the users rate the workout as “perfect” afterwards.
… by now we are at about 85% affection rate. That means the users rate the workout as “perfect” afterwards.
John Koetsier: Is this best for people who are new to fitness or people who are already deep in fitness? And there’s a couple things around there, because in the gym, I had a couple different routines that I would [do]. I didn’t have 3.5 million, but I had literally dozens of pieces of equipment that I could use and everything like that.
When COVID happened, you know, February, whenever it was, I started building out my home gym, I had a few pieces here and there. I bought a few pieces. I went and did a lot of bodyweight exercises. I started doing quite a bit more cardio stuff because I didn’t need a lot of equipment there, but eventually I got some pieces of equipment. How do you tell the system what you have and how does it tell you what you can do with those pieces?
Daniel Sobhani: Yeah. So the last part of your question is basically pretty simple. You can tell us what equipment you have and on every day you can even adjust that. So, you know, if you told us that generally you have access to a gym, but now it’s COVID so you take out the gym option. So that’s more often an easy part.
Also, you can tell the AI coach generally what type of movement forms you want to have included in your training plan. Do you want to have bodyweight exercises? Do you want to have weight-based exercises? Do you want to have, for example, running-based exercises and so on. So we can cover that more in a kind of simple way. Really the more challenging question is who’s the exact target for what you mentioned at the beginning.
And I would describe the target group as one, you should probably be somewhere between really a beginner, you can do this as an absolute beginner.
John Koetsier: Yes.
Daniel Sobhani: The AI will adapt to you, so from a beginner to an experienced. If you are really a pro, then reality is that you are, you have so specific wants and wishes, you are already so knowledgeable, that the AI at least right now where we are where the technology cannot yet hold up to this level of you know your own body. You know what you’re like, you know how you want to do it, you can feel into yourself.
So the answer really is from an absolute beginner to an advanced. If you want to achieve a certain goal, this is really a very good option for you, and the only thing we ask from you is to have a certain willingness to invest work.
John Koetsier: Yes.
we don’t advertise … “five minutes per day and you will see incredible results within three weeks” … it’s not true if you hear anything like that
Daniel Sobhani: So very explicitly, we don’t advertise — and it’s also not true — all of these things where, you know, five minutes per day and you will see incredible results within three weeks, no sweat, nothing. That’s … a nice word for this is it’s not true if you hear anything like that.
John Koetsier: Okay, cool. So we had a question here from YouTube, and it’s the question that I was going to go to next anyways, so that’s why I’m showing it on the screen here. Talk about the AI technology. What have you built? How does it work? What technologies are you using? Give us some insight.
Daniel Sobhani: Yeah. So the machine learning algorithm is based on a human-augmented AI approach. So that means that you have human set parameters, especially around the training physiology, the programming within, and within these parameters the AI works. And the AI itself is because we want to coach, so we want to give a recommendation and not backwards looking at how data matches in the industry.
John Koetsier: Yes.
… we combine causal inference modeling with data mining techniques, to generate this personalized training experience based on user source data
Daniel Sobhani: So what we do is we combine causal inference modeling with data mining techniques, to generate this personalized training experience based on user source data. So the data that we get from the user and capability estimation … that’s really the absolute gist of what we’re doing.
John Koetsier: You may be interested to learn that Helm.ai, a self driving car startup in Silicon Valley uses something quite similar. I interviewed them like a week and a half ago or so, and they use what they call “priors,” which is things that are known about the world. Things like object permanence that you’d learn as the child, right. And they insert those in and so you’re talking about some human knowledge that’s going into the system and guiding it, and then the AI working there. What was the most surprising suggestion maybe that you’ve personally gotten from your AI coach?
Daniel Sobhani: Well, I don’t really think I got anything really surprising. What I really hated was that, very correctly, the AI figured out that my leg work and my leg muscles are like, let’s say a little bit underdeveloped versus the other ones. You know this is what sometimes happens.
John Koetsier: It was all upper body, huh? You were a glamour model in the gym. Wow, haha.
Daniel Sobhani: And it made me really work hard to counter this imbalance. So I think that was really one thing that I noticed.
John Koetsier: Excellent, excellent. And talk about the comparison to a human coach. You know, how would you compare the AI coach with somebody who’s actually there physically?
… the great thing about using machine learning techniques to personalize a workout is that within given parameters, you can provide very accurate workouts prescription
Daniel Sobhani: Yeah, I think this is really a fantastic question also because it leads into discussions around the future. So the great thing about using machine learning techniques to personalize a workout is that within given parameters, you can provide very accurate workouts prescription. And on top of that, and you can do this at the cost of a cup of coffee per week, so at a very low price point.
And you can have this change adapted every day, every given minute, you as a user decide that today for some reason I want to brief my coach somewhat differently, you know, more time, less intensity, different equipment, different type of modality. So really this anywhere, anytime approach of good personalization within given parameters.
John Koetsier: Mm-hmm.
Daniel Sobhani: The challenges of this model really come towards when the starting point of that prescription gets murky. So if you have today, you know, certain injuries or something aches and you really don’t like this type of exercise, and for some reasons today you just don’t feel like doing X, all of these things, they are very, I mean, very difficult to impossible to predict for an AI coach.
… the one big benefit of a human coach is, yeah, the human connection, you feel accountable towards that person
And so then really the question goes, how many input parameters do you want to take from an AI coach? You know, how big should that questionnaire be before you can start working out? Or do you at some point of time really venture out into, okay, for these edge cases, I let the user select his own workout for himself, but that’s really one thing. And the other thing is that the one big benefit of a human coach is, yeah, the human connection, you feel accountable towards that person. That person can more help you really to keep your routine than an AI-based coach.
Even if the AI-based coach gives you personalized messages, it explains the workouts for you, and all of these things that human connection misses. And so, yeah, really one of the things that we at Freeletics are thinking about is like, what’s the winning proposition, the winning model for the future, where you can combine the best of these two worlds?
John Koetsier: Yeah.
Daniel Sobhani: And, you know, something like provide 80% of the time AI-based machine learning, so personalized coaching, but for the 20% of the time where you really have a challenge or you are going to struggle to motivate yourself, or it’s really a more complex question, get a real person in. We can help you out in these educators. And so the idea really would be that at still a very competitive price point you can have this 360° service that really help you not only to have effective workouts, but stick to a routine.
John Koetsier: It almost sounds like you’re announcing a product here.
COVID is a tragedy … but one of the systematic things that it did was it really brought a lot of people and a lot of attention to the … digital fitness industry
Daniel Sobhani: Well, it’s something that we think about. Like, I think just generally speaking, now with everything that happened around COVID — and I mean, overall, COVID is a tragedy for people, you know, their health and the economy, and so many other things — but one of the systematic things that it did was it really brought a lot of people and a lot of attention to the fitness industry, into the digital fitness industry. And so, I’m certain that the rate of innovation and quality of service will go up very steep, at least for the leading companies in that area. And so, yeah, also we are thinking what are the real, big next steps for Freeletics to give users the best experience around this.
John Koetsier: Very interesting. I mean, I can assume that it will get better over time as well. And I can assume that some of the things that are maybe challenging, you know, like entering what you want to do, or what equipment you have, might get easier as well as maybe you get more access to voice assistance and stuff like that on the various platforms, over time we just talk to our AI assistant at some point. But let’s talk about, you mentioned that you don’t make ads like ‘5 minutes a day and you’ll be super fit, no stress,’ you know, ‘3 weeks and you’ll be at your desired weight,’ but do you have some data on how people perform over the long term with Freeletics?
Daniel Sobhani: We do. So what I can share is that our average retention rate is better than what you would see if you would go to the gym. That’s a given fact.
John Koetsier: Cool, interesting. I don’t know what those numbers are. What are the average numbers for people who go to the gym? Do 30% stick with it? 50%?
Daniel Sobhani: So we really look at who’s still active a year afterwards. And like, if you really want to be in a great position that should be a minimum of 30%, if not higher.
John Koetsier: Okay.
Daniel Sobhani: Most of the people in a gym it’s much lower. Like unfortunately the gym model very much works on this, you go there, subscribe, but you know …
John Koetsier: And you don’t actually go. Yes, that’s how it works.
Daniel Sobhani: If all of your subscribers would use the product, the gym would be packed. No one could go to the gym anymore. So in that model, you already incorporate that a certain percent of — like a high percent of people don’t show up. And really the difference with digital solutions is that the major part of your company is just focused on getting you and keeping you active, which is very different to the stationary industry.
John Koetsier: Well, if you’ve got year-long retention rates that are around or above 30%, it’s a very unusual mobile app, because of course retention rates, Day 7 retention rates for most mobile apps are in the 10% range or something like that. So it’s pretty low. Let’s get back to some of the COVID and fitness stuff for the survey that you ran. What have you found are some of the biggest differences when people are working out at home versus in the gym?
… almost 65% of people who were working out at home were using some digital tool in order to support them
Daniel Sobhani: I think really the biggest, the by far biggest difference is the adoption of digital technologies within this. So we found out that I think it was almost 65% of people who were working out at home were using some digital tool in order to support them, and really a big 40% of these did this for the first time ever. And that was really the main headline that we found out.
And I personally believe, and also based on the data that we internally currently see in terms of usage, despite some of the reopenings of the gyms in some parts of the world, that this trend towards at least using digital technologies to like round up your overall experience in the gym, that that will definitely stick. So that’s definitely one thing. The other thing we really found out that this combination of fitness and mental wellbeing is something that is on a trend, if you want to call it like that. And I personally couldn’t be more happy about this, because really if you want to feel good, if you want to see results in the long term, these are really the two things that you should look about and invest it.
John Koetsier: Yeah, well, that’s interesting. I mean, one of the things you’ve found is that meditation was up, right, and mental health. I mean, we all need a little bit more during COVID, right? But there were some interesting men versus women statistics there as well.
Daniel Sobhani: Yeah. So what we found out, I mean, it was — so I think roundabout, and that’s really contrary to what I would’ve predicted, because typically what you see is that women have a higher engagement and interest in fitness and mental wellbeing and meditation … throughout all of these statistics. But for now, what we figured out it was exactly the opposite.
… typically what you see is that women have a higher engagement and interest in fitness and mental wellbeing and meditation … throughout all of these statistics. But for now, what we figured out it was exactly the opposite.
So you had about 20% higher engagement, interest rates, people starting with this, on male than with female. I mean obviously there’s a little bit of, okay, because less male in the beginning did already something about this, but that was really very, very new to us. And this is also a part yet where we still have to make sense of the data. Why is that? So what changed?
John Koetsier: Interesting. I mean, I bought a few things to work out at home. It was tough, by the way, everything was out of stock, but you saw some new investment in actual tools and equipment and stuff like that. What was the average level of investment?
Daniel Sobhani: So what we surveyed was that the average spend was I think somewhere $120 for the male and around $70 or $80 for female.
So there is a difference between male and female, but this also comes a lot because male are more like excitable by technology and, you know, they want to have weights and all that stuff which costs money. I think one interesting fact you should mention, or one should mention about these average figures is that while the average is, I mean, pretty low, but if you segment or you like deaggregate this data, what you see is that a lot of people spend much less than the $120, and then you have some of these really very expensive purchases that you have, you know, the Pelotons or your dumbbell set that cost $1,000 or something like that.
John Koetsier: Yes, yes.
Daniel Sobhani: So the willingness to invest into your health and fitness went up dramatically. And I think also that will be a factor that will increase innovation in that industry because especially for app-based models, in the past you always had the challenge of, you know, we come out of an industry where in the beginning apps were for free, you just downloaded them. And then you paid $1 for them, and then you paid $5 for them. Now we are at a level where it is okay if you charge $9 or $15 per month or something like that. But if you really want to provide great service, fantastic content, always new, personalized, a real person coaching — you can’t offer that at $15. So this higher willingness to pay will come back to users with greater service and greater innovation.
John Koetsier: Thank you, Daniel. Thank you for your time.
Daniel Sobhani: Thanks for having me on the show. Really appreciate this. I love the conversation.
John Koetsier: Excellent, had a lot of fun as well. For everybody else, thank you for joining us on TechFirst. My name is John Koetsier. I appreciate you being along for the ride. Whatever platform you’re on … please like, subscribe, share, comment, all the above. If you’re on the podcast later on and you like it, please rate it, review it. That’d be a massive help. Thank you!
Until next time, this is John Koetsier with TechFirst.