SDS 335: Many Ways to Fail & Five Ways to Succeed in Startups

Podcast Guest: Rico Meinl

January 29, 2020

After Rico had a failed AI startup, he reached out to me to come on the podcast and talk about his failure. We talked about B2B vs B2C, doing customer research, when to startup, and Rico’s 5 lessons he learned.

About Rico Meinl
Rico is an enterprising, extroverted Data Scientist with passion for Artificial Intelligence (AI)/Machine Learning (ML) and unique ability to easily forge relationships with colleagues, customers, and other business partners. In the past year, he co-founded a fashion tech startup where he got exposed to the realities of bringing a product to market working in an extremely fast-paced and uncertain environment. Before that, during his work at novomind Inc. he established an AI research lab and learned how to communicate with executive stakeholders and ship ML models from research to production. At the same time, while building up meetup.ai, which is now the biggest AI Meetup group in Hamburg and Berlin, he learned how to manage communities and connect inspiring speakers with a large audience.
Overview
Rico previously was a guest on the podcast in January 2018 after visiting DSGO 2017 in the fall. To attend the conference, Rico raised money from his company and his school and had the goal of meeting just one good professional while he was there. Through the conference he was inspired to start his AI meetup and got to have meaningful chats with myself, Hadelin, and others. 
Rico’s former startup, Dresswell, was designed in the fall of 2018 to be an app to help people find correctly fitting clothes. In the summer of 2019 they pivoted to the business to talk directly to consumers rather than try to sell to online retailers. The algorithm used to take body measurements was a human-in-the-loop machine learning system to catch any mistakes on the AI’s part. The vision for the company was to tackle the monetary inquality that comes from how you dress and how you can afford to dress. The company started with jeans, one of the most difficult pieces of clothes to measure and they figured if they could do that, they could do anything. They worked with consultants and began to get in touch with online retailers to try and get real revenue. Meetings were slow and far between. It wasn’t a good speed for a startup and people weren’t interested in using an app to measure themselves. So they switched from B2B to B2C.
It reminds me a lot of Hadelin and I starting BlueLife. We had the opposite journey, however. We started with B2C ideas and then changed out approach and moved into B2B. But we saw similar things, for example, people were already taking pictures of their food so we wanted some form of computer vision that could tell you the nutritional content of your meal. We moved away from it because it’s a very competitive space and apps get copied immediately if they’re successful. B2B, as Rico points out, is about connections and you’re ability to sell to people. Startups are interesting. Rico points out that starting a startup for the sake of it isn’t smart and it might not be the best way to solve a real problem. Which I agree with, to a point. I think the world will start to care as you gain experience and I thik failure is important for everyone to try out. 
After failing, Rico learned some things. The first thing he learned was to start with probelms in your own life when developing products. It’s easier when you have experience with the problem you want to solve. If you wouldn’t use your solution in your own life, why should anyone else? Next, he learned to stay focused on end results that have the highest impact on the company strategy. What you care about isn’t necessarily what your buyers, partners, or customers care about and make sure your metrics reflect this. The third learning is: it’s not about the money, it’s about the ‘expletive money’. Every tech person wants to build cool stuff. But startups need to monetize in some way. You need a business model that’s sustainable. This is something I pushed back on because not every startup can be Amazon and trying to chase sales can create dillution of focus. But, Rico points out, assembling a startup is like jumping out of a plane and putting that plane together as you freefall. It can result in shifting and pivoting to keep the startup alive. Number four is to avoid hoping into the academic field and implementing learnings from journals without contacting the authors to find out what’s truly behind the research. According to a previous episode with Sam Hinton, 1 out of every 20 research papers is wrong, full stop. His final learning is to reach out to people with specific questions and they may get back to you more than you think.
Rico’s got big things on the horizon and isn’t done learning. He recommended some habits and some books to make use of for anyone out there looking to launch their own startup
In this episode you will learn:
  • Rico at DSGO [8:50]
  • Dresswell [17:10]
  • B2B vs B2C in startups [34:03]
  • Rico’s 5 learnings [53:25]
  • Learning #1 [53:54]
  • Learning #2 [56:33]
  • Learning #3 [1:10:43]
  • Learning #4 [1:24:08]
  • Learning #5 [1:34:02]
  • Rico’s next steps [1:45:35]
Items mentioned in this podcast:
Crossing the Chasm by Geoffrey A. Moore
The Goal
by Eliyahu M. Goldratt
The Power of Habit by Charles Duhigg
Christopher Bell
‘s LinkedIn
– Co-Founder at dresswell
Follow Rico
Episode Transcript

Podcast Transcript

Kirill Eremenko: This is episode number 335 with AI entrepreneur, Rico Meinl.

Kirill Eremenko: Welcome to the SuperDataScience Podcast. My name is Kirill Eremenko, Data Science Coach and Lifestyle Entrepreneur. Each week, we bring you inspiring people and ideas to help you build your successful career in data science. Thanks for being here today and now let’s make the complex, simple.
Kirill Eremenko: This episode is brought to you by BlueLife AI, our very own consulting and now corporate training company. If you’re enjoying our online training programs in Machine Learning, Python, Visualization, Tableau, Artificial Intelligence, Data Mining, and all these other amazing areas of data science, then now you can actually send a request for us to come into your company and train-up your whole team. At BlueLife AI, we provide training services where we can come in and train-up your whole team in a specific area or in a specific tool set so that you are all on the same page and that your company progresses forward.
Kirill Eremenko: Our team of trainers, including myself and Hadelin and numerous other experts cover-off areas from R, to Python, from Machine Learning, to Artificial Intelligence, to Computer Vision, to Tableau, to Power BI, to Amazon Web Services, to Data Mining, SQL and many, many more tools and areas of data science that you might need expertise in, that you might want to train-up your team in. We’re very happy to come in and guide and train-up your team.
Kirill Eremenko: Also, not only do we provide corporate training in the technical space, but we also provide executive training. If you need help understanding how data science can be applied in your business, what artificial intelligence is, and how it’s disrupting your industry, then we’re happy to help with that as well. To connect with us and discuss your training needs and what solutions we could provide for your company, head on over to BlueLife.AI, that’s one word, BlueLife.AI, and send us a request there, and we’ll be happy to connect and take it forward from there. On that note, let’s jump straight back into the podcast.
Kirill Eremenko: Welcome back to the podcast. How is everybody going? It is crazy. It is already quarter to 8:00 in Australia, almost quarter to 8:00, and I just finished recording a podcast with Rico. This went way over. You probably can tell from the length of the audio file that this is a long podcast, one of our longest ones, but it was so well worth it. We just got into so much chatting and we got so carried away talking about AI startups. It just like blew my mind, and I’m sure you’re going to enjoy it.
Kirill Eremenko: Rico, for those of you who don’t know, Rico was on the podcast back in January 2018, so exactly two years ago from when you’re listening to this audio is almost exactly two years ago. Since then, his career has taken him in very interesting places. That was episode number 123 if you’re interested. Rico, we met at DataScienceGO. Then he came on the podcast, and he actually spoke at DataScienceGO in 2018. Since then, he’s started a startup. He’s gone in different countries, from Germany, to Amsterdam. He’s moved, he’s pivoted his strategy or they’ve pivoted their strategy many times, done lots of things and ultimately the startup failed.
Kirill Eremenko: Unlike most people, I guess, who in this situation would try not to make that very public and vocalize it and talk about it, Rico actually reached out to me and said, “Hey, Kirill, can I come on the podcast? Let’s do a podcast together and let’s talk about this failure. Let’s talk about the lessons that I learned from it because I want to help people. I want to help other data scientists out there, who are thinking of starting an AI startup, to avoid those same mistakes and to learn from my experience.” That was so cool. That’s why this chat went on for so long.
Kirill Eremenko: Here is a couple of things that you will hear in this podcast. How to find out your customer needs. Coming up with AI product ideas. B2B versus B2C. To start-up or not to start-up? When is the right time, when is the right decision to impact the world through a startup, and when is it better to do it through another company? Then we spoke about the five lessons, the five top lessons that Rico learned from his startup experience. Just bear in mind that it takes us some time to get into the flow, between maybe the first 10 to 20 minutes, to really get the conversation going. After that mark, I’m sure just like us, you’ll settle into it and enjoy the ride. Without further ado, I bring to you for the second time around on the SuperDataScience Podcast, AI entrepreneur, Rico Meinl.
Kirill Eremenko: Welcome back to the SuperDataScience Podcast, ladies and gentlemen, super pumped to bring you, for the second time around, our returning guest, one of our favorites, Rico Meinl. Rico, how are you doing today?
Rico Meinl: I’m doing very well, thank you.
Kirill Eremenko: Very cool. I’m getting you just woke up and you’re calling, as you said, quote, unquote, from the middle of nowhere in a part of Germany. How’d you end up there?
Rico Meinl: How did I end up back here, you mean.
Kirill Eremenko: Yeah, “Back here.” You were all over the place for the past couple of years.
Rico Meinl: Yeah. I guess the last time was like two years ago that we talked.
Kirill Eremenko: Well, we talked in DataScienceGO like a few months ago, but the podcast was, yeah, man, it was exactly-
Rico Meinl: Was it two years or one year ago?
Kirill Eremenko: I think it was two years because it was before your DataScienceGO 2018 talk. Right?
Rico Meinl: Right, yeah.
Kirill Eremenko: That’s crazy.
Rico Meinl: It’s interesting that time flies.
Kirill Eremenko: That’s crazy.
Rico Meinl: Yeah. I mean, in that same year … Let’s take it back. We talked in January, the first podcast. Then DataScienceGO, was in October or November of that year?
Kirill Eremenko: That year, I think it was October.
Rico Meinl: October, right. When we came back, actually, after DataScienceGO, we were doing this quick project at Richard’s company in Tasmania.
Kirill Eremenko: Oh, yeah. You went to Tasmania.
Rico Meinl: Yeah, which is great because I’d never been to Australia before. It was a super interesting experience. After we came back, I started a startup with one of my friends, which I met in the summer in Hamburg. He was, back then, he was working at a retail store, because he was really passionate about the whole fashion-tech space. He also has a tech background. We had met up for dinner a couple times and chatted about the idea, just generally chat about AI because he was interested in that. Then we joined forces after DataScienceGO and started Dresswell. Visually started in January-
Kirill Eremenko: You started to dress well?
Rico Meinl: No, the name is pun intended. No, the company was called, the startup was called Dresswell.
Kirill Eremenko: To summarize, you came to DataScienceGO in 2017. It was November, October or November. Then you went back to Germany. Then you came on the podcast in January 2018. Then in November, or October 2018, you came back to DataScienceGO as a speaker.
Rico Meinl: That is correct. The thing that really still fascinates me is, so back then I was doing your guys’ Udemy courses because I think I was studying computer science in third semester or something, and I just started to hear about all the things about data science. I think I read somewhere that AI is kind of like the holy grail of computer science, so I was intrigued to learn more about how to implement it and what you can do with it these days.
Rico Meinl: I was doing your Udemy courses, and I think I got an email one day from you saying that you’re going to host this data science conference. I was like, “Oh, man. That would be amazing to go, because it’ll really accelerate me with my career in AI if I meet someone there.” I was looking at the speakers lineup, I was like, “Man, if I would meet one person, that would already be worth the investment.” Because I was a student back then, so it was not really feasible to just fly over to the U.S. for a weekend to go to the conference. Then I, essentially, raised the money from the company I worked at and from my school to be able to attend the conference.
Kirill Eremenko: That’s crazy, man.
Rico Meinl: I would say, if I would have met one person … I think at that conference, I mean, I met you, I met Hadelin, had a pretty long conversation with Ben-
Kirill Eremenko: Ben Taylor.
Rico Meinl: Met Richard. Ben Taylor, yeah. Richard Hawkins. It’s just like amazing how much got initiated through that one conference, like. Ben … Through the conference, I was inspired to start AI Meetup, because when I was going to back to Hamburg, I really wanted to have something like the conference to be able to connect with people who are trying to pursue that same career path, who are also working in data science, working with AI.
Rico Meinl: When I came back to Hamburg and I looked up on this platform called Meetup.com, there was just nothing, there was no Meetup for AI in Hamburg by that time. Also, back then I was thinking about starting an AI lab at the company I was working at, Novomind, so I asked Ben for advice, how to approach the executive team about it, and he gave me a lot of tips how to approach them. That really gave me the confidence to, after the conference, go and approach them, so a lot of stuff that got initiated at that conference. That’s so fascinating for me to see how that was like a rolling- [crosstalk 00:11:05]
Kirill Eremenko: Like it all happened. You put it all in motion just by attending that one event.
Rico Meinl: Exactly.
Kirill Eremenko: I don’t understand, man. How did you raise the money? That’s crazy. How do you convince your company and your university to pay for you to fly from Germany to San Diego for one weekend to attend a data science event? That just doesn’t … It’s very impressive how you did that. I want to know more.
Rico Meinl: The funny thing was, my school, I don’t think, they didn’t used to have PhD programs, now they are starting to have some of them. They had this offer that when you’re a research student and when you attend a research conference, they would essentially pay for it. I kind of split that story and told them it was not a research conference, I’m not a research student, but it’s still a conference.
Rico Meinl: There’s budget for sending people out to these conferences, and it’s obviously not well-used if there’s no PhD programs. I knew there was not a lot of people going to these conferences, so I was able to convince them to take over some of it. Then my company, because I talked to them about the idea for the AI lab before and they were intrigued. I thought going to these conferences would help, help me be able to set it up.
Kirill Eremenko: It did, right? You came back and you set it up.
Rico Meinl: Yeah, I did, so it was a good return on investment.
Kirill Eremenko: That’s epic, man. That’s epic. In a nutshell, it was worth it to fly there for two days and go back.
Rico Meinl: It was worth it, yeah.
Kirill Eremenko: Nice.
Rico Meinl: I would really encourage people to do that, to take these chances, even though the probability might be low that you might meet someone that changed your life. Obviously, it doesn’t really happen … I went to different conference afterwards and it didn’t change my life in the same way, but I mean, there’s such a low risk associated with it, and if just go out and ask people to fund you, for example, there’s no downside to it. The risk, the chance of actually meeting that one person is pretty-
Kirill Eremenko: That’s interesting-
Rico Meinl: I still would-
Kirill Eremenko: Would you say there’s something different about DataScienceGO, or do you think just because it was your first conference it had such a massive impact, or it was just a random thing? Why do you think you felt the difference at our event and not at other events?
Rico Meinl: Now, actually, I think part of it is that being a novice at conferences, I was just really excited to see what it’s like. Now that I’ve been to different industry conferences, I would say that most conferences are not that beneficial, actually. Now, for example, I was at a conference, an AI conference in Berlin last week just to catch up with some people, and it was conference happening. Nico, my friend from Berlin, was organizing one of the workshops there, so he got me a free ticket, but it was just not …
Rico Meinl: Now, when I attend a conference, I don’t expect much to come from there, but back then, I was expecting a lot. I was like, “Oh, man. Conference, they’re like this amazing place where you’ll meet all these people, and there’s so much opportunity.” I guess, maybe that was part of it.
Kirill Eremenko: You were hyped up. You were, like you had-
Rico Meinl: I was hyped up. Exactly.
Kirill Eremenko: Yeah, I see what you mean.
Rico Meinl: I wanted to connect with everyone. Yeah.
Kirill Eremenko: Yeah. Very cool, man. Thank you for that you keep coming back. You’ve been DataScienceGO ’17, ’18 as a speaker, ’19 you came back. Very, very excited to have you every time. It’s really cool.
Rico Meinl: Yeah.
Kirill Eremenko: You’re like one of our veterans.
Rico Meinl: There’s a couple people that were there this year that were there all three years. That’s amazing.
Kirill Eremenko: Yeah, yeah. We have, I think, a dozen people who’ve come back all three years. It’s pretty cool.
Rico Meinl: Yeah.
Kirill Eremenko: It’s pretty cool. Speaking of Berlin, Germany. Rico, it’s so cool. It’s so cute how you guys calls each other Rico and Nico. His full name is Nicholas. What’s your full name? Is it just Rico, or you have a full, a longer name?
Rico Meinl: It is just Rico.
Kirill Eremenko: Just Rico. Okay, well-
Rico Meinl: People ask me that all the time. It’s just Rico.
Kirill Eremenko: Nice, nice. Speaking of Germany, so Nicholas, thanks a lot for the introduction. Nicholas, as you know, is helping us set up DataScienceGO in Europe, and this is going to be epic news. As an exclusive update to podcast listeners, we are working on bringing DataScienceGO to Europe. It’s going to be epic. I can’t say more at this stage because I don’t know by the time this podcast goes out if it’ll be announced yet or not, but look out for it. It will be announced very soon if hasn’t yet. Are you excited?
Rico Meinl: Yeah, I’m super excited. Let me say this, Nico is, in person, when we’re meeting people and talking about it, he was really shy about it, but he was the one that essentially got Meetup AI in Berlin to grow from, essentially from nothing to like over 2,000 members within like less than a year.
Kirill Eremenko: That’s crazy.
Rico Meinl: He was the one that made it super big. Yeah, I’m super excited, when he’s the one that’s helping you guys also set it up.
Kirill Eremenko: Yeah. That’s cool, that’s cool, man. Yeah, that’ll be cool to have you there. Apart from that, so let’s shift gears. We’ve been talking about conferences, DataScienceGO. Last time I spoke to you was in San Diego for DataScienceGO 2019. It was September 2019, so a few months ago. Tell us more about Dresswell. What was going on there?
Rico Meinl: Right. I mean, so the short story, we started in November. I think I said that before, but I locked myself in my room for two months reading research papers to … Because essentially, the high-level pitch is we were, for people who have trouble clothes that fit well, we built an app or an algorithm that you can use on your phone to take your body measurements, and then find perfect fitting clothes. We started only with jeans.
Rico Meinl: That algorithm we developed in November, December, and then January, and then all over the time, we collected data and made it better and stuff. In April 2019, we moved to Amsterdam because we’d gotten in touch with two jeans brands there. In Europe, Amsterdam is the biggest place for fashion, for denim fashion.
Rico Meinl: We then started to reach out to all these big online retailers over the summer, and essentially switched, pivoted to B2C in August
Kirill Eremenko: B2C, meaning for customers, for consumers?
Rico Meinl: Yeah, yes, yes, yes.
Kirill Eremenko: Okay, okay, interesting. Originally, you were going to sell it to companies, like a service.
Rico Meinl: We were going to sell it, exactly, yeah. We were going to sell it as a plug-in to like the Nordstroms, Zappos.
Kirill Eremenko: Interesting.
Rico Meinl: All these online retailers.
Kirill Eremenko: Okay. Tell us more. Why was this an AI startup? Sounds like a fashion thing.
Rico Meinl: The algorithm that we used to take people’s body measurements was, we tried a full-on, like a full deep learning model to implement that, but we ended up with a human in the loop, plus machine learning approach. That was like, because I recently wrote, because we put the startup on hold, and I recently wrote down some of my learnings. That was one of the learnings that when in November and December, when I was reading all these research papers and trying to understand how we could use AI to get people’s body measurements, I mean, that was a lot.
Rico Meinl: That was a steep learning curve to be able to read and understand these papers, and then implement them into production. I guess, also for the usability. Because unless we would have collected all the data for the edge cases that we had for the app, like if you’re turned a little bit too much to the side, or if you have weird objects in your image, we didn’t really want people to deal with having bad measurements, so we had a human in the loop that checked if AI predicted the right key points and the right segments in the image. If it didn’t he would manually correct it.
Kirill Eremenko: Very interesting. Using machine learning or AI to predict measurements, having also a human in the loop. First of all, why that idea? There’s so many things you could be doing with AI. Why did you see that there is a need that you can help people with to find better clothes? Why did you pick that specific idea?
Rico Meinl: The bigger vision that he and me started with, and I really bought into was that there’s a huge inequality in the world that when you were raised by parents that either had the money, or knew how to dress really well and taught you, or just like in different sectors of the world, you just have an advantage in life. Because we all subconsciously judge people by the way they look, even though we might not want to admit it, but we all do. Our, essentially the vision-
Kirill Eremenko: Like first impressions, right?
Rico Meinl: Exactly, yeah. Like if you meet someone and he has sweatpants at a conference, you’re not going to talk to them the same way as with somebody-
Kirill Eremenko: Yeah. I’m terrible at dressing. I always need help with this stuff. I would be one of the first people to buy the thing. I’m very bad at matching colors, or even just picking out proper clothes. Yeah, I can see that. Okay.
Rico Meinl: Then after the first startup didn’t work out, he moved back to Germany because I think they had it in New York. Then he moved back to Germany because he has family in Hamburg. He started working in a retail store to understand the customer need. I thought that was really inspiring because if someone has that much passion for a project, to be able to …
Rico Meinl: Essentially, when you’re a tech person, you have all these career options these days, high paying career options. Then because you have so much passion for your project, you go back working as a salesperson in a retail store, taking a lower salary than you could have got anywhere else. I thought that was really inspiring. He did that just to connect with customers and learn what their needs are in this space.
Kirill Eremenko: He’s in tech. He could, technically, have a really cool, high paid job, but instead, he choosed to go to retail to understand the needs of the customers.
Rico Meinl: Right, right. It’s funny because Uber Eats had that … When Uber started Uber Eats, I don’t know if they took it over, but they had people working in restaurants before they started working on the product. I mean, now, I think that’s really the way to go. If you want to break into an industry, go down to the very lowest level where you can to the customers and really understand what they need.
Rico Meinl: Essentially, he saw that a lot of men, because he was working in the men’s section first, suits and then jeans, had a problem of finding stuff that fits them really well. That was like the, usually the first problem. I mean, one thing we learned later is that style is a huge factor too, and we didn’t really take that into account. Because, so there’s the thing, when you go shopping somewhere, you go shopping there because you already know they will sell a certain style.
Rico Meinl: For example, if someone goes to Nordstrom, of course, he has problems trying to find stuff that fits well, but he already made the decision what kind of style he’s looking for when he goes to Nordstrom, because they sell to a very certain segment. That we also didn’t take into account that much. There was a lot of men that had trouble finding jeans. Like you just said, men usually don’t like to spend their time or deal a lot with finding stuff that fits them well, so they would just come and be like, “Find me something that fits good.” He would give them a couple of jeans and they would be so happy with the service if they found something that they love, and they would come back.
Rico Meinl: We thought, “How can we bring this online?” Because right now online shopping, I think, is fundamentally flawed. There’s these huge returns. For example, on jeans, like 60% return rates, so it’s a big struggle for the online retailers. It’s funny because I think in America, it’s you have to pay, essentially before, you have to pay with a credit card or a bank account before you order something at like a Nordstrom, but in Germany, you don’t even have to pay before. You can just put it on a bill and only pay for what you keep. If I would order at like a Zalando here, I could order 10 pairs of jeans, don’t pay anything and if I don’t like them, I’ll just send them back, so there’s like-
Kirill Eremenko: Then you just pay for what you keep.
Rico Meinl: Yeah, and if I don’t keep anything, I don’t pay anything.
Kirill Eremenko: Very interesting. That’s very cool.
Rico Meinl: That’s bad for the environment, that’s bad for the customer because you have to walk to the post office and send it back. I think now they even actually drive to your house and pick it up, that’s how far it has come.
Rico Meinl: We thought how we could bring this online. The idea was to build this measurement tool that people could use to measure themselves, and then we would, essentially, recommend them the perfect size, the perfect fit, and then they wouldn’t have to send it back.
Kirill Eremenko: Fantastic. That’s the idea. Okay. Did you see that thing I sent you? Or I’m sure you saw it before, the Algorithms Tour by Stitch Fix.
Rico Meinl: Oh, yeah, yeah, yeah. They did a great job.
Kirill Eremenko: Is that similar to what Stich Fix does?
Rico Meinl: I mean, the B2B product we tried to create in the beginning was not similar because Stitch Fix has a massive questionnaire of about, I think, 10 pages, where you just fill out your whole style profile. What we were trying to do is, you don’t have to fill out anything about your style. You would just do the measurements and be able to just do the math between the jeans dimensions and your body dimensions, and then be able to recommend you what fits you best.
Kirill Eremenko: Okay, gotcha.
Rico Meinl: Stich Fix goes more from a style perspective, which we- [crosstalk 00:27:10]
Kirill Eremenko: All right.
Rico Meinl: … B2C product strategy tracks well.
Kirill Eremenko: Let’s keep walking through this. This is very interesting. November 2018, you lock yourself in your apartment in for two months, and you’re going through-
Rico Meinl: That was in Hamburg still.
Kirill Eremenko: Oh, still in Hamburg.
Rico Meinl: Yeah, yeah.
Kirill Eremenko: You’re going through all these research papers on how scientists or machine learning, AI, has been used or proposed to be used to detect people’s measurements from images.
Rico Meinl: Exactly.
Kirill Eremenko: What happens next?
Rico Meinl: Essentially, what I did, I connected to all these authors of the papers before reading them. Because what I found was that, especially also in research, and I might be wrong, because this might be only for our specific field, but a lot of stuff is over-sold too, because I mean, you have to keep research funding going as well, and if you don’t make any process, they’re not going to keep funding you. A lot of the papers, you get really excited when you see them. You get really excited when you see the accuracy, but then when you talk to the authors, it’s like, “Ah, yeah, it doesn’t work in production. It only works in this really specific dataset.”
Rico Meinl: Sometimes, it’s even like the results in the paper are kind of like, let’s say, optimized too. It was really frustrating to bring that into production, and it took a lot of time too. I mean, we spent over three months trying to get to the accuracy we thought we needed to play the game. Then, essentially, we got to a measurement accuracy that we were happy with, so we contacted … That’s what Chris said. He contacted jeans brands in Amsterdam that we could start working with, so we have like a prototype or like a product in production is when we would approach the big online retailers.
Kirill Eremenko: Could we pause for a second?
Rico Meinl: Sure.
Kirill Eremenko: I want to understand, what do you mean? What’s the difference between having something in research and having something in production? What exactly do you mean when you say, “We had something in production”?
Rico Meinl: Oh, like have something that’s already running where you have customers, or that hopefully makes money. Just putting yourself out there, like launching a product. Because there’s also a lot of, I guess, also a lot of emotional side. I mean, you probably know this with your online courses, right?
Kirill Eremenko: Mm-hmm (affirmative).
Rico Meinl: You spend months and months developing it, and then when you put it out there, you already have an end result in mind, like what kind of metrics you want to reach. If you publish a course and … If no one would like it, it would probably be pretty frustrating.
Kirill Eremenko: So production, you mean like you had an app or something like that that people could already use?
Rico Meinl: Right. Like a plug-in on their website or an app that people … Back then, there was an app that people could download for the- [crosstalk 00:30:15]
Kirill Eremenko: Okay, so you already had this app, and so then you contacted, you got in touch with this company in Amsterdam.
Rico Meinl: Yeah. We got in touch with some jeans brands in Amsterdam, and they were interested in working with us. We chose this one guy, he had his own brand, which is pretty big in Amsterdam actually. He was local, he was not internationally. We wanted to work with him because he was also doing custom jeans, so we thought he had a lot of insider knowledge on how to select … Like what needs to be done to make your jeans fit well. That was interesting because we essentially thought we needed …
Rico Meinl: In order to fit jeans well, you need to have the hip, waist, upper-thigh, lower-thigh, calf and ankle to fit well. Because otherwise, people are going to be like, “Oh, no. It’s too loose at the ankle.” There’s all these details you have to get right. That’s why jeans are like the most hard a thing. We thought, that’s also why we started with jeans, because we thought if we solve jeans, everything else will be a walk in the park. We thought, we need to get like 1% accuracy for all these points, essentially, and we started working-
Kirill Eremenko: You mean 99% accuracy.
Rico Meinl: Oh, no. Sorry, like plus, minus, 1% measurement accuracy.
Kirill Eremenko: Ah, gotcha.
Rico Meinl: If I have a hip of 90 centimeters, I want the algorithm to predict something in-between 91 and 89. Actually, a little bit more room because jeans are pretty stretchy these days. When we started working with him, we realized that there is a relative degree of importance for all of these points. For example, the hip and the waist, essentially run 90% of the game.
Rico Meinl: If the hip and the waist doesn’t fit, you can already say, “Okay, these jeans’ not going to cut it.” If we were to just focus on these two essential measurements, we would have saved a lot of time and would have been much quicker into production, essentially. Then we were working with him-
Kirill Eremenko: Is that what you did? Did you focus on those two points?
Rico Meinl: Yeah. What we did, so we essentially took the app, we went to all the universities in Amsterdam. I think we did user testing with like over 700 girls-
Kirill Eremenko: Wow.
Rico Meinl: … because we were only doing it for women at that point. Because we had to collect data, like measurement data. Because we had like a, I guess you could say like a pre-trained model, but then of course, we needed our own data because we had specific measurements. We thought like, “How do we get this data?” Because we’re not Amazon, who could just … Because Amazon was also kind of working on this project. We’re not Amazon, we can’t spend millions of dollars trying to people to get measured, and then train a model, so we had to hustle, and go to the universities in Amsterdam, and pick people one-by-one, and ask them if they wanted to try this app.
Rico Meinl: We collected these data points, and that’s at what point we felt comfortable to contact the bigger online retailers to … Because at that point we thought, okay, we need … Because we actually, we didn’t launch the app with the jeans brand because we thought it would be good to have something out there, but we also need money. Developing the app and launching it would take at least probably a month to do a good job and-
Kirill Eremenko: That’s very optimistic.
Rico Meinl: Right.
Kirill Eremenko: “A month.” More like 12 months.
Rico Meinl: Yeah, I guess, [inaudible 00:33:42]. Yeah, so then we thought, “Look, we have to move faster and get in touch with these online retailers, because if they like what we do, they would probably fund us ahead.” Then we could get it in production and also get paid for it because I mean, we’re a startup, we need to make money. We had friends and family funding at that point, so we needed to make revenue.
Rico Meinl: Then we contacted like ASOS, Levi’s, Zappos, Nordstrom, Shopbop, like all these big online retailers. We got a couple of meetings, but they were slow, and we didn’t anticipate that, because we’d never done B2B sales before. Nordstrom, for example, we had a talk in July and the next meeting would have been in October. There was a slump going on in the whole retail space, so ASOS told us that, that everything is going super slow right now and that we shouldn’t expect anything to move until the end of the year.
Rico Meinl: That was kind of like, we couldn’t deal with that cycle as a startup, plus we also saw that with this measurement technology, there’s stuff out there right now, but there’s nothing that works and that people actually like to use. Because it turns out that people don’t like to measure themselves with their phone. They’re not excited about the technology. They don’t care about this technology. They just want to get the jeans that makes them look good, and the easiest and fastest way to get there. They don’t want to download your app, they don’t want to deal with your startup, buggy software.
Rico Meinl: We thought we needed to iterate faster to go to market. Also, some of these retailers told us that they’re worried about customer adoptions. Like Revolve, this company in LA, who just recently got pretty big through the whole … Because they were the first to do influencer Instagram marketing in the fashion space, I think. They said, like, “Hey, we’re worried about if customers will actually use the app. We see that it provides values, but we’re not sure if people would use it and like it.”
Rico Meinl: We thought, okay, the best way would be to switch into B2C, have our own prototype running where we just buy jeans from Nordstrom, have people use the app and then send it to them. We wouldn’t make any money off of it in the beginning but, essentially, we could prove the concept.
Kirill Eremenko: Let’s pause there. Very interesting, very interesting story. It reminds me a lot of how Hadelin and I were starting BlueLife back in, I think that was 2017. We came up with the idea in June, and we were brainstorming for a very long time. I remember, we were in Portugal. We spent like a whole month brainstorming. We’d go out for dinner every evening, have a bottle of wine between the two of us, and just keep coming up with ideas, “How, what are we going to do?” and so on.
Kirill Eremenko: We had kind of the opposite journey in our approach to everything here. That we started with ideas in the space of B2C, then by iterating for like … We were almost set on a few ideas in B2C, and then we changed our approach and changed to B2B. It’s interesting that even though our approach was opposite, we observed very similar things. In B2C, we … While brainstorming, we were coming up with these ideas. For instance, one of them was, “Let’s take photos …” Because we were at dinner all the time, brainstorming.
Kirill Eremenko: We’re like, it’d be really cool to have an app where you can take a photo of your meal, and it will tell you right away how healthy that meal is, and moreover, do computer vision and recognize the different types of food and provide you, “Okay, this many calories in this meal” or “This many vitamins, these amino acids are present.” If you do that every time, you take one photo … People take photos of their food anyway.
Kirill Eremenko: You take one photo of your food every meal, that’s like 21 photos per week, and you get a whole summary of all of the vitamins, and minerals, and amino acids, and macronutrients, and everything, fats, all these sorts of different types of fats that you ate. That can really help you understand where you’re missing out. Because you might think you’re eating a healthy diet, but ultimately, you might be missing out.
Kirill Eremenko: Or moreover, for instance, if the app doesn’t recognize an image, we could crowdsource the whole image recognition. Because if you took a photo and, I don’t know, the fish on your plate wasn’t identified properly, or the carrots were identified as broccoli or whatever, you can just click and change the label. That way, we could crowdsource the data as well.
Kirill Eremenko: We had really cool ideas like that, but ultimately, we moved and … I had cool ideas in my view. I don’t know how other people think about that. Ultimately, we moved away from B2C because we found that very competitive and as soon as you create an app, it’s going to be copied very fast if it’s successful. Also, it’s just like that whole revenue thing that you said. You were experiencing that issue at the start. Any startup experiences that issue.
Kirill Eremenko: We found that people are probably not going to pay money for an app like that. Very hard to monetize that type of thing. You need to put ads, we don’t want to put ads, you need to sell it to a company. We didn’t see a very clear path on how to monetize that. We decided we’ll switch to B2B because it’s much … You have less customers, much less customers. Instead of millions or hundreds of thousands, you’ll have dozens. At the same time, the deals are bigger, and therefore the revenue is less of a concern.
Kirill Eremenko: However, little did we know, same problem as you came across, that it’s very slow. B2B, the sales cycle is very slow. The bigger the deal, the slower the sales cycle. Interesting that we went B2C, changed to B2B. You went B2B, changed to B2C and still, we experienced very similar issues along the way.
Rico Meinl: I mean, B2B is really all about … Maybe not all about. If you’re a great salesperson and you have experience, but it’s all about connections, right?
Kirill Eremenko: Yeah.
Rico Meinl: Like knowing people. At in Nordstrom, for example, I mean, we had to, essentially, cold email them over LinkedIn. I think I reached out to like almost, I think, 500 people on LinkedIn.
Kirill Eremenko: 500 people.
Rico Meinl: I don’t know if that was the best strategy. I mean, you get to talk with these people but then, like Levi’s, for example, they have a lot of these offers coming in. I mean, also, we didn’t … Now in retrospective, I would have loved to known that before. I’ll go more into books later, but there’s this one book called Crossing the Chasm, which is about, essentially, technology marketing, pretty focused on B2B. Drew Houston, the founder of Dropbox recommended it and it’s in Ferriss’ podcast. If I would have read that before, I would have done things differently in terms of selling the product to these businesses. I mean, that was definitely a big learning.
Rico Meinl: I mean, one of the things I’ll go into now, because I think it fits, before I go into the B2C part. I wrote recently an article, maybe you can link this to the show notes. It was like the whole takeaway was, essentially, the world doesn’t care. It’s funny, because I was intrigued to ask you too if you felt the same. Because you guys essentially went from …
Rico Meinl: Let me explain first. I think the whole startup, like starting a startup at a young age right now, or even generally, starting a startup is pretty over-hyped. I think there’s a lot of people, I think I was partly one of them, that just start a startup to start a startup. To be part of that whole founder story, to do something while you’re young. Probably a good idea, but you shouldn’t do it just because you’re young, if you don’t have a good idea.
Rico Meinl: It would even go that far that I would have people tell us like, “Oh, wow. You guys are actually solving a real problem.” That’s ridiculous. If you’re not solving a problem, why would you start something? Even if you’re solving a problem, maybe the right way is not to start a company, maybe it’s to join an existing company because using their resources would solve that problem faster. I don’t if that’s usually the case, but it might be the case. All this stuff, because I mean, starting a startup … Ben talks a lot about that too, but starting a startup is brutal.
Rico Meinl: You work crazy hours and essentially, if you’re successful, the world loves these stories of like the college dropout who started his own startup, put together a team of brilliant engineers and they worked super hard and slept on the floors under the table, because getting a place in San Francisco is too expensive. Everyone loves these stories when they get really big, but the other side of the story that people maybe, or that usually people don’t talk about is that you do all these things, but it doesn’t mean that you’re going to be successful because the world doesn’t care about really what you put in, it only cares about what you put out.
Rico Meinl: The world cares about great services for business, I say, great products for consumers, but actually making impact, that’s what the world cares about, not what you put into it. If you’re not successful, no one cares if you work 100 hour weeks. No one cared, essentially, New Year’s last year, I was working all night, and my birthday, I think I was working until 4:00 a.m. We slept two hours and went to a customer interview in Amsterdam.
Rico Meinl: It’s like, essentially also, the world doesn’t care how old you are. If you’re thinking, “Oh, it’s going to be great to start something when I’m 21.” No it’s not, because the world doesn’t care if you’re 21. If your product sucks, no one will talk about it. I think you could probably share that too, because you guys were massive, hugely successful with your online courses, but then going into B2B, my guess would be that none of these companies really cared that much about your online courses, right?
Kirill Eremenko: Very interesting.
Rico Meinl: It’s a different game.
Kirill Eremenko: Yeah. I would say for us, it’s true. Yeah, none of the companies really cared about the online courses, but mostly because we went into a different niche, we went from … Or a different industry even. We went from education into consulting, into building AI. Because there were quite a few students that were asking us, “You’re teaching this stuff. Can you help us build this thing at our business?” and so on. There was some initial interest, but then other than that, you’re right. You have to start all over again.
Kirill Eremenko: However, now what we’re doing is we’ve added a new branch to BlueLife, where we provide corporate training services to companies. Rather than just consulting, we can engage a business with, “All right, we’ll come in and we’ll train-up your team the same way we train online, but in person training with very rigorous curriculum and exactly what you need.” Like tailored education for your team to bring up to speed, or executive training as well in artificial intelligence strategy.
Rico Meinl: Right, right.
Kirill Eremenko: We just started that, so we’re testing that approach out. I have a feeling that now we can transfer those credentials. We can transfer, “Hey, we’ve taught a million students online. We have a ton of experience on how to create curriculums and things like that. We can train-up your team as well.” Then that will open up doors for consulting as well.
Rico Meinl: That’s awesome, yeah.
Kirill Eremenko: I guess, yeah, it’s interesting, it’s very interesting. I was listening to a podcast from the founder of Superhuman, there’s like an emailing client that they started, which is like a competitor to Gmail, but not for everybody. It’s just for people who get like a ton of emails all the time. Very different, very personalized. It’s like 30 bucks a month. You get like personalized onboarding. You have to join a wait list to join them, and so on.
Kirill Eremenko: The guy is very knowledgeable and has interesting ideas about product-market fit and things like that. I was listening to this podcast with him. One of the things he mentioned was that because his first startup was super successful, it was very easy for him to get funding for the second one. It was in a context where he was saying that he didn’t actually want to get too much funding too early on, but I think the world doesn’t care to a point, after a certain point of experience.
Kirill Eremenko: This is another thing that, you’re right. The world doesn’t care how much time you put in. If your product is crap, then nobody’s going to care at the end of the day. At the same time, if through … Why I think it is a good idea to try these things, and regardless of your age. It’s a good idea to try projects to, like you say, fail. This was in your DataScienceGO 2018 talk. It’s good to fail because the learnings you get in the process, that is an indicator to people that you are tried and tested with time. You’re a seasoned entrepreneur.
Kirill Eremenko: The word “seasoned” has the meaning. You’ve been through seasons, through happy and bad, through ups and owns, through amazing successes and through terrible failures. You’ve seen the good and the bad. You won’t be surprised by anything that comes your way, that is thrown at you. It’s very hard to throw you off your feet. While, at the end of the day, your B2B or B2C clients won’t care about that, the people that are going to be in your corner, whether it’s investors, your team, your advisors, mentors, your early adopters, your fans, they’re going to care.
Kirill Eremenko: They’re going to see, like, “Hey, Rico’s already failed at this thing, and he’s tried this, and this, and this. He’s growing. He’s got so much wisdom that the chances of him even starting something that he doesn’t believe in or he hasn’t calculated through well are quite low.” In that sense, I think it’s very valuable to have this experience. As you would probably agree, that from failure you learn 10 times more than from a success.
Rico Meinl: That is really interesting. I totally agree with the point that the people in your corner, they do care. Especially if you’re a serial entrepreneur who had a successful business before, investors will majorly care about if you approach them with a second idea for a business. Almost like something, they just throw money at you because you’ve done it before, successfully. Then, so yeah, I guess, that is absolutely true.
Rico Meinl: Also, your team. If you’ve been successful before, it will be easier to hire people, rather than being like a 21-year-old, first time entrepreneur who hasn’t done anything. What I was more trying to get to was that in the end, the customers, I guess most customers … I guess, and that point, you’re right. The early adopters do care about …
Rico Meinl: For example, Evan Williams, the founder of Twitter. When he founded Medium, I think people did care in Silicone Valley at least, when they were like, “Oh, Evan Williams started something again. I’ll probably check it out.” Like the mass market customer, I’d say, who doesn’t even know who is the founder of Medium right now, they don’t care who founded the company. They care about the product they see, and they don’t care if Evan Williams is the guy behind it, but that was more like the …
Rico Meinl: The other thing you also touched on is I think, yes, I agree that failure is a good way to learn, but I think you have to take it with a grain of salt. Because just because you fail, doesn’t necessarily mean that you learn from it. It’s not a given, and it’s a really hard and painful process to go through and reflect on what you did wrong.
Rico Meinl: Also, there’s a huge gap between what I saw too, is between reading or listening to something, or even watching a video on something and then applying that stuff. For example, maybe we maybe can touch on that later when we go into the B2C of our product. Like Steve Jobs always said, they succeeded, one of the reasons they succeeded was because they built something where they, themselves were the customers for it. I’ve listened to that video, I think, maybe like 10 times.
Rico Meinl: Still, when we launched our product, we started by building something for women. Because one reason was that we thought the women’s market, the market of women buying online is way bigger than men, and they’ll be, naturally, more interested in the product. That was maybe a wrong assumption. It was super hard to build something for someone else because there’s so many details in a great consumer product you have to get right. Like-
Kirill Eremenko: Or you could have had a woman on your team to help with that part.
Rico Meinl: Yeah, right. That’s true, but-
Kirill Eremenko: That’s a very, very good point. If it’s not a learning … A failure is not a learning if you didn’t actually learn from your mistakes.
Rico Meinl: Exactly. Sorry. Did you want to say anything?
Kirill Eremenko: Yeah, yeah, sure.
Rico Meinl: I will say that I’m not good at this. I usually like to move fast, and move ahead, move forward. I’m usually not that reflective on my actions, but I was really lucky. We’re still in great contact, me and my co-founder, Chris, because he’s the exact opposite. He would take more time for his decisions. He’s very strategic, and he would take the time. Sometimes, even overly reflect on things to get a learning out of them.
Rico Meinl: For me, that was really helpful to sit down with him and go through that thinking process of every month, “What did we do wrong?” Also, it was really important, like, “What were we thinking back then?” Because it’s always so easy to look back and say, “Oh, that’s where you went wrong.” Being in that situation with all the knowledge you had back then, the decision is not that clear-cut, usually. It’s easy to look back and think, “Oh, that’s what we did wrong,” but being in a situation, it’s not that easy. Doing that whole process was really painful, but also that’s what we learned from.
Kirill Eremenko: Let’s recap, just in order to clarify, because you told me a bit about this before the podcast. You sat down with your co-founder after you put the startup on hold, which if you don’t mind me saying, is another way of just saying, like, “We failed. Let’s move on to something else.”
Rico Meinl: Exactly. Yeah, yeah.
Kirill Eremenko: After all that happened, you said, “All right. We’re going to make sure we learn from this, and we’re going to rewind and spend some time to analyze month-by-month, what did we do wrong? And what we were thinking back then.” You told me you got like five learnings from that. Let’s go through them. I think this will be very useful to somebody who might be considering to start a startup in the space of AI.
Rico Meinl: Right. Oh, so we’ll say that he was the one proposing that, to be fair. I was really down with it too, but he was the one that proposed to do that. What we came up with was a longer list, but I wanted to cut it down to five things that I thought would also provide value for other people. The first thing was that whole, start a problem in your own life when you’re building a consumer product.
Rico Meinl: It also doesn’t mean that you only have to build for problems in your own life. If you talk to a friend and he tells you about a problem he has in his own life, but maybe he doesn’t have a technical background or maybe he doesn’t want to start a company, but he’ll give you really accurate information about that problem and you can go ahead and solve it. That will be valuable too but in general, if you’re seeing a problem in someone else’s life, it’s much harder to solve that problem and build a great product that that person will use.
Rico Meinl: For example, when we switched then, when we pivoted to men, that’s when the first time we realized like, “Wow, the solution, the app we had built, we don’t see how this would fit in our lives, and frankly, we wouldn’t use it,” and that was crazy. Because before, I would call some of the girls I’m friends with and ask them about the product and stuff, and ask these, now, unrelated questions. Once I really thought through, “Is this something I would use in my own life?” And the answer was no, that was a pretty revealing moment for us, a thing like wow-
Kirill Eremenko: Wow.
Rico Meinl: Realizing that 10 months into the business is pretty painful, trust me. If we would have just done that thinking process 10 months earlier, that would have saved us a lot of hiatus-
Kirill Eremenko: Okay. [crosstalk 00:55:46] Very useful.
Rico Meinl: I’m not the only one. Steve Jobs, as I said, he said that too. Because they were the ones that were building, they were building computers because they wanted to use it themselves. I guess, that was really, really cool. Wait, let me … The second one, oh yeah. This one was really interesting also, I think because I said like, “Start with a problem in your own life.” This is in no means like a “how-to” manual or a “how-to” blah, blah. This is just like, “Let me try it differently,” like what I learned and not to advise anyone to follow that advice. I think, if it’s useful, try to apply it but it’s not that people should definitely follow these rules.
Rico Meinl: The second thing was to really stay focused on the end results for your company that has the highest impact on the business strategy. This goes on actually two layers. I told you, I think I said earlier that we thought, for the different body measurements, we needed 1% in accuracy. Then we found that if we just focused on the hip and the waist that is, like with the 80-20 rule, that cuts it already for on a very high level. If were just focused on these two measurements, that would have been smarter.
Rico Meinl: Then at the same time, going another level upwards, even the 1% measurement in accuracy, or the measurements in itself was not the right task to focus on for the high-level business strategy. Because the online retailers don’t care about the body measurements. They care about the conversion they get, what drives conversion in their online job and what reduces cost associated with returns.
Rico Meinl: Then later in B2C, the customers, they don’t care about your measurement accuracy. They care about if your product helps them to get there faster, easier, and also in the end, if the jeans makes them look and feel good. Focusing on these measurements so rigorously, spending like three months optimizing for this 1% measurement accuracy, instead of going back and thinking like, “Are there may be easier ways to test our assumptions?”
Rico Meinl: One thing we thought was to find people jeans that fit well, we could just ask them to send us one of their pairs that fits them well. Because usually, people have one pair that fits them well and it’s either broken or, I don’t know, they’re just looking for a different color. Having them give us the specific model or send it in via mail so we can measure the jeans and then find them something just based on the jeans measurements maybe would have been the better product. It would definitely be easier to test it out, because we didn’t need to build any machines or any tools to get the measurements.
Rico Meinl: I mean, for the B2B, we could have essentially built this test environment where we had the product and the human in the loop, and then be very upfront about it with the businesses, so testing environment, it would take about three months to build it and gage if they would be interested in it. I mean, do that before you start working, like heavily go into the measurements and sell them a ready product.
Kirill Eremenko: Basically, kind of understand, come up with a product that has a better product-market fit. I think that’s the bottom line there.
Rico Meinl: Right, right. Yeah, and choose your metrics. Don’t focus on this 1% accuracy if it doesn’t have any impact on your business strategy. Because like I said, the B2B customers and also the B2C customers didn’t care about the measurement accuracy, so focus on the stuff that … Like the high-level stuff that matters for your business, and then, yeah because-
Kirill Eremenko: Very interesting.
Rico Meinl: … Because we got lost in the trenches there, I would say.
Kirill Eremenko: I agree, focusing on the metrics, but there’s one specific metric I’ve been reading about recently that is quite a new metric. Again, it was from a blog written by the founder of Superhuman. They came up with a way to actually measure product-market fit, quantify it. The way to do that is you basically ask your users, who … You’re slowly getting people.
Kirill Eremenko: Like you said, you build a test environment or you start with something small where they send you in measurements and you help them pick the right outfits. You have a smaller user group that is slowly maybe getting new users, losing some users. Ultimately, you ask them a question after about two or three weeks of them using your product. You ask them a question, “How disappointed would you be if you were no longer to use Dresswell?” in your case, or whatever else it is, Dropbox, Airbnb. “How disappointed would you be?”
Kirill Eremenko: Then you measure what percentage of people reply. You give them options, “Very disappointed, somewhat disappointed, not disappointed.” You measure the percentage of people that reply, “Very disappointed.” It’s actually not his original idea. This was already done by another venture capitalist, I think, who tested this out with many companies including Slack and others. They found that for your product to go viral, you need at least 40% of your customers to say that they would be very disappointed if they were no longer able to use your product and so, ultimately, for a startup-
Rico Meinl: That is so interesting.
Kirill Eremenko: I know. I’ll send you the article. Ultimately for a startup, you just need to measure that one metic. All you need to measure is, “What percentage of our customers are saying they’d be very disappointed if they’re not able to use your product?” You need to navigate your product in such a way, like make changes to it, adjustments, rethink your strategy, rethink the user experience, whatever else, submission, promise to customers, whatever. Whatever components you have in your product, you need to keep rethinking it.
Kirill Eremenko: You’ll start at 10%, 20%, 30, but once you hit that magic number of 40%, you’ve got product-market fit, and then you start measuring things like accuracy, revenue, profitability. One thing you mentioned at the start was that, “We’re a startup. We got to make revenue.” Well, I don’t agree with that. I don’t think that a startup needs to make revenue.
Kirill Eremenko: No, a startup, the revenue has to be figured out some other way. Whether you’re sleeping on the floor or you have your parents’ or friends’ money, or you’ve got investors investing in you. A startup can even for two, three, five years not make revenue. The goal of a startup is to find that product-market fit. Once you found it, once you have that 40% down, all sorted and ongoing consistently, every month-over-month, you’re getting 40% or above of your product-market fit according to this quantified metric, then you start thinking about whatever else you want.
Kirill Eremenko: You want revenue? Go for revenue. You want accuracy? Go for accuracy. There’s lots of things to go from there, but I think that would, based on what you’ve said, and I’m no expert in startups, but sounds to me like that would have been a bit of a game changer for you, if you had been measuring product-market fit from the start.
Rico Meinl: I would partly disagree. I think it’s very what you said, “Startups don’t need to make revenue.” That’s why I partly disagree. Because I think also, the metric, it depends heavily on what kind of startup you run. For example, our product, how many jeans do you buy a year? For the first part, we had just like measurements for jeans. How many jeans do you buy a year? Probably around two, girls maybe more like three. The whole three month rule would be … It would kind of not work, because if you only used a product once, there’s not a strong …
Rico Meinl: I mean, it depends much more on the first interaction with the product, other than like a Slack that you would use every day. I think the metric would still work if you find a way to maybe adjust the cycle, or I don’t know, frame it a different way. For us, it would have been very hard to test it over a three month period, because over a three month period, you only buy one pair of jeans.
Rico Meinl: Like Spotify, for example, I would literally cry if you would take it away from me tomorrow. I would be super disappointed, but I use it every day. It’s different than like a … I guess, maybe, it’s also maybe an underlying fault of our app. Then regarding the revenue. Again, I don’t if- [crosstalk 01:04:45]
Kirill Eremenko: Wait, wait, wait. Hold on with that. It’s a very interesting thing. Can you think of an app that you use that you really love, but you only use twice a year? Is there any other app that you have an attachment to that is very useful to your life, but you use it so rarely?
Rico Meinl: Well, Apple Podcast when I can’t find a podcast on Spotify.
Kirill Eremenko: Yeah, so but you wouldn’t care if Apple … Apple Podcast, they’re closing down Apple iTunes anyway.
Rico Meinl: I would actually care. They’re closing it down?
Kirill Eremenko: Yeah. They announced it like six months ago.
Rico Meinl: There is a lot of podcasts that are on Apple Podcast, but not on Spotify, so that would actually be a pretty big deal.
Kirill Eremenko: I don’t know. Well, iTunes they’re closing down. I don’t know if Apple Podcast-
Rico Meinl: Oh, yeah. That would be ridiculous.
Kirill Eremenko: Basically, it’s like it’s a backup up alternate, but what’s a life changing app that you use? My point is that if it’s really important to you, you use it at least weekly. Probably daily or several times a week. I just don’t imagine how a startup that is used, an app that is used twice a year is going to really ingrain [crosstalk 01:05:59] itself in the lives [crosstalk 01:06:02]. Because attention spans are so low.
Rico Meinl: I guess, yeah. I guess that’s why … For the online retailers, I don’t have any personal connection to them. When I ever, which is like never, but when I, a couple times, I ordered a jacket online, for example. The way I shop was I ordered it, and then I found it was cheaper on Hagebau than on Zalando, two German online retailers. It was cheaper on Hagebau, so I got it from Hagebau, even though I found it on Zalando.
Rico Meinl: There’s no connection, and I don’t feel like I’m betraying Zalando because I’m buying off their competitor, which I mean, kind of like Spotify and Apple Podcast. I have a personal connection, so I feel like if I’m using Apple Podcast all the time, I’m letting Spotify down, so I need to use Spotify more, because I like the product, so I want to use it. I guess, with these online retailers …
Rico Meinl: What I’m trying to say is there is no brand loyalty because you use them so rarely, but the people that use them heavily, they have, I guess, they have a brand loyalty towards one of the online retailers they use, they shop every week. They also wouldn’t have to use the measurement tool, because I guess, they only have to use it like once a month, unless you have a weird diet.
Rico Meinl: It’s an interesting point you’re raising, yeah. I guess that was definitely part of what we also saw as a problem. That it’s not a product that you would use every day, and it is very hard to build up, just with the measurement system itself, build up a brand loyalty, which is what-
Kirill Eremenko: Yeah, I see. On the other hand, if you had sold it to B2B as a service, because they have like thousands of clients, they would have users using it every day, but not the same users. You know what I mean? Every day, there would be like thousands of people using it, but just different users. Then you could measure satisfaction based on the B2B … I know, I guess, for an AI app, you also need that constant influx of data.
Rico Meinl: Right, right. Yeah.
Kirill Eremenko: Interesting. Okay, interesting.
Rico Meinl: Yes-
Kirill Eremenko: All right, so [crosstalk 01:08:09] What was the other thing you were going to say?
Rico Meinl: The founder of … I always forget their name. It’s an American-
Kirill Eremenko: Zappos?
Rico Meinl: No. It’s an American insurance startup. Zenefits, Zenefits. The founder of Zenefits, during YC Startup School, and when he was on state with Mark and Jason and Ron Conway, he was saying that with his first startup, he had to hustle and go through all these investors. He talked to like hundreds of investors, and they were all like …
Rico Meinl: This one investor essentially told him, “Hey, you guys’ presentation is really shallow. There’s no numbers on revenue and users and stuff. I mean, if you guys were the Twitter guys, we wouldn’t care abut that, but you’re not, so we need all these details about your business and the numbers.” What he took away from that was not to ramp-up the pitch deck and have all these numbers, but to become the Twitter guys, where you just have that much product growth that people, investors don’t care about your numbers.
Kirill Eremenko: Okay, so that was learning number two, to stay focused on results that have the biggest impact on your high-level business strategy.
Rico Meinl: Yeah, yeah. The third one is actually like for the title of an article. I pulled the quote from what Bill Campbell said to Ben Horowitz when he was trying to take his company public, like, “It’s not about the money, it’s about the … Money.”
Kirill Eremenko: It’s about the “beep,” money. Yeah, by the way, this article, I don’t think we’ve mentioned it. This article is available on Medium. I just gave it 50 claps because I like Rico. I haven’t read it yet, but it looks amazing. If somebody wants to recap on all the things we’re talking about here, we’ll link to it on the show notes, and follow Rico, Rico Meinl. The article is called The World Doesn’t Care-Lessons from Startup Failure.
Rico Meinl: We can also link it in the show notes. It’ll be a more brief version of what we’re talking about here. A little-
Kirill Eremenko: Yeah. You should link the podcast in the article because this is-
Rico Meinl: Oh, yeah.
Kirill Eremenko: … the full version.
Rico Meinl: I’ll do that.
Kirill Eremenko: Okay, yeah. It’s not about the money, it’s about the … It’s not the money, it’s the “beep,” money.
Rico Meinl: Exactly, yeah. Being a tech person, it’s we want to build stuff that’s cool. Essentially, I feel like that’s with every tech person. They want to build cool stuff. No one wants to build boring, old fashioned products that just drive revenue. We all want to build something cool. We all want to build products that we can show off to our mom, be like, “Hey, I built this.”
Rico Meinl: Essentially, what we learned is because we didn’t … Back to the revenue part. Because we didn’t focus on the money, we thought like, “Oh, we’re going to build this product, and because it’s so useful for customers and it’ll be so useful for these businesses, we’ll find out a way to monetize it.” In the plan we had, monetizing it for the B2B market, the B2B customers would have worked at scale, at a very, very large sale. Because then we only charged like a little bit, a very small fee for every customer that measures themself, or that buys the jeans after measuring herself. That would only work if you had millions of customers, aka, like a lot of these retail partners. Otherwise, you’re not going to become profitable in the long-term.
Rico Meinl: Because we didn’t focus on the business model from the start, that’s essentially what hurt us in the end. When investors asked us about the business model, we didn’t have any really good answers. That’s when we ended up pivoting many times, actually. That’s when it kind of got out of control and we had to put it on … We had to make a cut.
Rico Meinl: Because being an online retailer, like the idea when we went into B2C then was to, like I said, buy the stuff from Nordstrom, sell it over our own website and app, and then in the longer … Actually, pretty short-term, because we already talked to some of these brands, buy the stuff from the brands for retail price … Or for wholesale price, and then sell it online for retail price, and that will be our business model.
Rico Meinl: If you only do jeans in the beginning, that’s really hard. It would not make a lot of money, which is then why we decided to have our own jeans essentially, make our own jeans. Because with that, if we have our own jeans and a pretty strong brand behind it, we would be able to get them produced for like $20, like premium jeans, sell them for like $100. In that segment, that’s pretty cheap because for that quality, you usually pay like $200. Then sell that using our app as like a new technology. That would have been the business model that would have made good profits, essentially, would have worked.
Rico Meinl: That’s what really hurt us, that we didn’t focus on the money and we thought like, “Oh, we’re going to build a product. We’re going to build a great product. We’re going to build a great technology, and the revenue model will figure out itself.”
Kirill Eremenko: Interesting, interesting. I will partially disagree with you on that one. Because, yeah, I see the importance of money and making the revenue, but man, you’re not Amazon. You started by, “Let’s create an AI app that helps people get measurements.” Then with this whole, “Let’s set up a business. We will buy, now we will buy jeans at wholesale, sell them at retail, and our app will help people find the measurements. Oh, no. Wait. That’s not profitable. We’re actually going to create our own jeans brand and sell that way.”
Kirill Eremenko: It sounds like you’re going off on a huge tangent, very far away from your original goal and the app, you don’t even need the app. Just start a cheap brand of jeans and sell that. Why do you need the app? At the end of the day, I feel that dilution of focus, it would also kill your business as well. If you’re not finding product-market fit with just the measurements app, it doesn’t mean that adding a brand of jeans on top of it is going to help the situation. You’re just like, “All right. Just do the jeans then. Forget about the app.” You need to figure out one thing at a time. That’s what I would say here.
Rico Meinl: I agree with that. The problem was that people don’t care about the app.
Kirill Eremenko: Then ditch the app.
Rico Meinl: People didn’t … I mean, yeah.
Kirill Eremenko: Either change the app to one that people do care about, or ditch the app and do a brand of jeans. It’s like, I don’t know, “I’m going to sell you a super fast bicycle that gets you from LA to San Diego in two hours, but it’s super dangerous, so I’m actually going to sell you that bicycle and attach a car to it. You know what? Forget about the bicycle, just buy the car.” That type of thing.
Rico Meinl: The one thing I will say, because this all sounds really like we were just losing it, and just randomly iterating. At the same time, you always have to remember, and this is like Reid Hoffman says, “When you build a startup, it’s essentially like jumping out of a plane, assembling your plane on the way down.”
Kirill Eremenko: Yeah, I heard that one. That’s a really cool one.
Rico Meinl: If you’re assembling a plane on the way down and your app is not working, and you have to make a decision, it’s not like you can just go on vacation for a week and make a decision … I mean, maybe you could if you have enough money in the bank. We didn’t have enough money in the bank so we had to figure out something pretty fast, because we had to raise money at that point or we would die. I mean, there’s a lot of pressure on you to move, keep moving fast, and be nimble about your changes. I guess, that also hurt us in the end.
Kirill Eremenko: Interesting learning. Basically, I guess, we can agree on that. Have a way that you’re going to monetize this in mind. It’s noble to do things for the sake of doing them, but that’s research. If you want to build a business, you got to have monetizing in mind from the start.
Rico Meinl: Right. I think that one, I put that on because it’s really big for me, and I think other technical people as well, because you usually don’t think about the business side that much, but it’s so important.
Kirill Eremenko: Yeah, it’s true. I was actually listening to a podcast with a panel, with three venture capital investors, so it’s like people who work at venture capital firms. I think one of the actually owns their own venture capital firm. Their job is to find lucrative startups to invest into.
Kirill Eremenko: One of the comments they all three agreed on was, if your business model is, “We got to get 100,000 users or a million users, and then we’ll think about monetization,” those days are long gone. Now, they don’t invest in that type of stuff. You got to have, if you want investments, you got to have not just the plan on how you’re going to monetize, but you actually got to show that you are having monthly recurring revenue or ARR, annual recurring revenue. Or people are actually buying your product. You need to have a proper business plan around this stuff.
Rico Meinl: That is so interesting, yeah. I actually wanted to put that in … I don’t know if I mentioned this in the article. Reid Hoffman said that too during … I get a lot of influence from him. He’s great, he’s a pretty strategical thinker, very analytical. I like to listen to his stuff. He had a talk at … I can link the podcast. It’s called Entrepreneurial Thought Leaders. It’s like Stanford invited all these successful Silicone Valley entrepreneurs to give a talk, and he was one of them.
Rico Meinl: Essentially, one guy in the end asked him, like, “How do you monetize product in the mobile age?” Because everything moved mobile now. He actually said that it’s not as easy as just building something with a lot of users and just putting ads on it. It doesn’t work anymore. It’s not a very good approach for mobile. I guess that’s maybe also what you’re referring to. Back in the days when everything was web, it was a decent strategy to build something, gets users and put ads on it, but with mobile, it’s different.
Kirill Eremenko: Yeah, yeah. I totally agree. I was also listening to a podcast with the CEO of, I think it’s called The Athlete, the company. I got to check the name of the podcast, or the name of the company. Yeah, same thing. They were talking about ads. One thing, so they came up with this one … The company business idea is basically to, rather than have click bait articles about sports, where people click on them on Google, go read them and the monetization of that is through advertising. Instead, they’re like, “How about we do a closed website, where you pay for a membership?”
Kirill Eremenko: You pay like, I think, 30 bucks, 60 bucks a year. Your first year is like $30 a year. There’s no ads at all, and we have non-click baity, very curated content that is actually very valuable to our users, because it’s created by a dedicated team of people who are paid to do this, rather than paid to create click bait content. I found, the website is called The Athletic, the CEO of Athletic-
Rico Meinl: Sounds similar to Medium.
Kirill Eremenko: Yeah, yeah, yeah, but Medium is free. There’s no paywall in Medium.
Rico Meinl: There is, there is. I think you can only read like one article a day. Some of the articles are behind their paywall, which means you can only read them if you have an account.
Kirill Eremenko: Then that’s cool. It’s just to your point. It’s maybe not exactly the business model you’d use in the case of your startup or your idea, but to your point that advertising is dying off, like people … There’s better ways to serve your customers with good content, and people are willing to pay money to you, rather than to the advertising in exchange for a higher, more premium, tailored service.
Rico Meinl: I’ve been listening to, there’s this one company in LA that we also briefly talked to. They’re called Brand & Entertainment Network. They do some really amazing stuff with AI. They’re talking a lot, their CEO, who’s a friend of Ben, who I met that way. They talk a lot about how product placement is taking over the advertising industry. I think that’s really …
Rico Meinl: I don’t think advertisement is dying off in itself, but I think just banner ads are dying off, because people hate them. What you get now is like this native ad … I don’t know if “native ad” is the right word. I’m not an expert in this field, by no means. For example, paying influencers to wear your t-shirts, or have them advertise your beauty products. Or, I don’t know. If someone would pay me, in my Medium article to use Google Cloud for my machine learning projects, and I would write about using Google Cloud, essentially. I don’t know. That kind of stuff, where it’s not that obvious. People know it’s ads, but it’s actually integrated into the content.
Kirill Eremenko: It’s kind of like, a classic example is the Truman Show. That movie. Remember that?
Rico Meinl: Yeah, yeah, yeah, yeah. We watched it at school.
Kirill Eremenko: You watched it at school.
Rico Meinl: Great movie, great movie.
Kirill Eremenko: Amazing movie. I re-watched it recently. It’s so good. I love Truman Show with Jim Carrey. Oh, it’s just so good. [crosstalk 01:23:56] It’s like that’s the definition, or I don’t know, the extreme version of product placement in people’s lives basically.
Rico Meinl: Right, yeah.
Kirill Eremenko: Okay.
Rico Meinl: Okay, so the fourth one I briefly mentioned before, just to really recap. Unless you’re a PhD and you already know what’s possible in your field, be very careful with trying to hop on the train and just implement the latest and greatest research papers. Because they usually look super nice, but then it’s really hard to get actual business results from them. If you do that, make sure to connect with the authors, because they will make that process, like of uncovering what’s really behind that paper- [crosstalk 01:24:40]
Kirill Eremenko: Telling you the truth.
Rico Meinl: Man, they’re open about it. That’s the thing, I’m not trying-
Kirill Eremenko: That is so surprising. I was surprised-
Rico Meinl: … I’m not trying to spit on research here. Research is great. There’s great potential in these papers. I mean, frankly, one of the guys that really helped me out is a really good friend of mine from Hamburg. He now works at a Facebook AI research in Pittsburgh. He’s doing great. It’s awesome, but I talked to a couple of them, and I would be super excited about the paper. I would ask them and they were like, “Yeah, it doesn’t work. Don’t try to …” [crosstalk 01:25:15]
Kirill Eremenko: What is this? This is crazy? Oh my gosh. This is even worse than, like I had a guest on the podcast, Sam Hinton. Really cool guy, astrophysicist, he’s teaching Python now. He was actually on the Survivor show, random fact. Didn’t win, but had a great time. Anyway, he mentioned this obvious thing that we … It’s so obvious, but nobody actually thinks about it. In order for your research to go through, you need a P value of 0. what? 0.05, right?
Rico Meinl: Uh-huh (affirmative).
Kirill Eremenko: You need 95% confidence that your findings are correct. What that actually means is that one out of 20 research papers out there is wrong, full stop.
Rico Meinl: Wow.
Kirill Eremenko: Just think about it. One out of 20. You pick up a stack of 20 research papers, one of them is most likely going to be wrong. What you’re saying is that, heck, this is probably true in many industries as well, that not just one out of 20, probably like one out of five or one out of four is wrong, or is not reproducible.
Kirill Eremenko: There was an article about it a few years ago in Nature, in the Nature Magazine that it’s a big issue, this whole frequency statistics that we’re following, that is taught at school with the P values, and so on. It’s extremely, a lot of the time, you can get cause and effect the wrong way around. It’s very hard often … It’s easy to run the experiment but hard to reproduce. It’s quite a low standard for research. A lot of people abuse it because that’s, indeed, how you get the research funding.
Rico Meinl: Yeah, I mean, they’re not wrong. They’re working in theory, and sometimes also in practice with massive clusters of GPUs. It’s just for the proposed use case, like in the beginning when they say, “Oh, this research could be used for online shopping,” blah, blah, blah. That usually doesn’t work. I mean, the research works, the papers work, but maybe-
Kirill Eremenko: The use case is like, “It could be used. No guarantees.”
Rico Meinl: Right. The friend I mentioned, for example. His first paper, he had a series of papers on virtual avatars, which is also how he got to work at the Facebook research lab, because they’re working on that right now, like the virtual representation. His first paper, I was really excited. I contacted him about it, and he was frank with me that it does work. The results are not as good as we would have needed them, but it would have been good to have the starting model, but the wait, the time to get the measurements is not a gradient descent. I forgot what it’s called. It was not like a deep learning optimization. It was like … Forget the name. Maybe I’ll remember it later, but it was a different type of optimization- [crosstalk 01:28:18]
Kirill Eremenko: Like a genetic algorithm, or something like that?
Rico Meinl: No. Something that’s iterative.
Kirill Eremenko: Okay. Gradient descent, stochastic gradient descent. Anyway, we can put that in the show notes.
Rico Meinl: Yeah, anyway, so it just took six hours. If the customer would have taken a picture, it would have taken six hours to get the measurement.
Kirill Eremenko: Oh, you take it before going to bed and in the morning you have a result. “Don’t forget to measure up before going to bed.”
Rico Meinl: Yeah, but like I said, that might be very specific to the field of … I’m not saying that as truth. For example, Google released DeepLab, this image segmentation algorithm. That was a research paper, and that actually works pretty well in production. I guess, there’s a difference between also industry research and academic research.
Kirill Eremenko: Yeah, yeah, that’s true. You take the DeepMind papers, they live by them. I watched that video, how DeepMind … What is it? AlphaStar, AlphaGo beat humans in Go, like 2016, April. Since then, it’s gone through AlphaGo Zero, which can learn from scratch. Now they have AlphaStar, which beats humans in StarCraft, completely wipes them out.
Rico Meinl: Wow.
Kirill Eremenko: It’s so beautiful to watch. I used to play StarCraft, love the game. It’s amazing to watch, these humans have no chances. They’re just like, these are the top players in the world in this super complex, strategic game with building units, fighting, mining minerals, having three different races, and so on, micro-managing your units. Like no chance at all. It’s just destroying in all the Protoss or Terrans, or whatever. I watched the Protoss games. Amazing results. If you see those results, then you know the research paper works. “Hey, the results are right here.”
Rico Meinl: That’s brutal, man. The guy who was the top player in Alpha, in Go, he quit.
Kirill Eremenko: Lee Sedol, I think.
Rico Meinl: Yeah. That’s like a pretty traditional game in China, I believe.
Kirill Eremenko: Yeah. All of Asia.
Rico Meinl: He was doing that most of his life and now he just quit because of freaking AI.
Kirill Eremenko: Oh my god. That was like, that day or that weekend or week when they were playing, the whole world, online stores worldwide sold out of the game, Go. You couldn’t buy it anymore, because that game was watched by tens of millions of people because it’s that popular. Imagine losing to a computer-
Rico Meinl: Wow.
Kirill Eremenko: … in front of tens of millions of people. Of course, you’d quit. Poor guy, poor guy.
Rico Meinl: Right.
Kirill Eremenko: Aw, man. Yeah, okay. Well, good tip, very good tip. Be careful of the latest research, indeed. I think that’s a big learning. You locked yourself in your room for two months, you read all those research papers. It would have been disappointing to find out that half of them are not really useful because they’re not that accurate or reproducible.
Rico Meinl: Yeah. I mean, it was not like some of these others things, like an aha moment. It was more like a learning over time. You still try to stay optimistic about like, “Oh, maybe it still works.” Or part of it works, and you could implement just a part of it, because that’s always, you could take a part of it and just use it as inspiration on what not to do, or what to do.
Kirill Eremenko: Then, what do people do? Do they do their own research? What would you suggest? How would you have done this differently?
Rico Meinl: Well, I guess, like the way, what we did is we chose the simplest solution, which ended up working best. It wasn’t glamorous. Essentially, it used machine learning to … Like image segmentation, to do image segmentation. Like cut the person from the background, and then detect certain key points. I also do edge detection to make sure that it really cuts at the … Because image segmentation is not always really accurate, at least, say, the art that were using. It was like a Chinese paper. It actually worked really well. That Chinese paper, on humans, cutting them out from the background works amazingly well.
Rico Meinl: Then having the key points, for example, on the hip. Because it’s not pixel accurate, we used edge detection to then be able to correct the key points, also using a deep learning algorithm. That was not the most glamorous solution, and it was kind of like patched together, but it worked really well.
Kirill Eremenko: Interesting.
Rico Meinl: I guess, you could go with what’s not glamorous, but what works.
Kirill Eremenko: Yeah. That’s the key, right.
Rico Meinl: The funny thing is, look, that’s the first algorithm we came up with, and then we still iterated over all these automated deep learning solutions. Then we actually went back to the original algorithm and used that.
Kirill Eremenko: Yeah. It’s like in Russia we say, what do we say? “Best is the worst enemy of good.” You had a good thing … The 80-20 rule, basically. Right?
Rico Meinl: Yeah, yeah. Exactly.
Kirill Eremenko: You want to stop, at some point. Okay, cool. Send me a link to the research paper, please, so we can put it in the show notes. That Chinese one that you said that worked really well, maybe people would be interested to read it.
Rico Meinl: Oh yeah, sure.
Kirill Eremenko: Okay, cool. Learning number five. What was that one?
Rico Meinl: That one, we can go over that really quick. There’s this management methodology called Objectives and Key Results. I’ll also put the book in the show notes, you can put the book in the show notes.
Rico Meinl: It was basically like the management methodology that Andy Grove brought to Intel back in the days. Intel, obviously super successful. John Doerr was working there back then, venture capitalist. He brought it to Google, and it’s how Google ran their operations since the very beginning. We were, naturally, pretty intrigued to try it out. Because also, it’s pretty big now. LinkedIn apparently uses it. Well, all of the big Silicon Valley companies do OKRs. We wanted to implement it, and we did from the beginning.
Rico Meinl: It always seemed like a little bit of too much overhead, but it was essentially, like you said, an objective. For example for us it would have been, like one was getting to 1% accuracy. That’s an objective for the next month, and then setting key results essentially like milestones, how to get there, so like measure 50 people and get your accuracy so we know where we stand right now as a baseline. I don’t know. Whatever you think will get you to that 1%.
Rico Meinl: Then you would have like a set of OKRs is for your company. Maybe also “raise funding” as an objective. Then you would narrow that down to the individual level, so the company key results, like measuring 50 people would be, for example, my objective. I’m in charge for measuring these 50 people, and then that aligns with the overall company goal. That was a rough explanation.
Kirill Eremenko: The learning was like, use OKRs? Or what was it?
Rico Meinl: No. The learning was that we had a lot of trouble implementing it, so after, I think, eight months or something, I was so tired of it, I just shot out an email to some of the people that shared their stories in the book. The one that was like Atticus Tysen from Intuit. He’s the CIO, and Brett Kopf, who runs an education startup called Remind. They got back to me like the next day and said-
Kirill Eremenko: No way.
Rico Meinl: Yeah, it’s crazy. Sometimes, when you just shoot out these emails with very specific questions, people help you. They’re so willing to help-
Kirill Eremenko: Where did you get their emails in the first place?
Rico Meinl: Oh, a lot of Googling.
Kirill Eremenko: Very, very … What’s it called? Confident approach. You don’t know these people, you just found their emails online. Just like, “Ah, you know what? I’m just going to send them an email.”
Rico Meinl: It’s also, it’s sometimes just like guessing. If I would know that Kirill is CEO of SuperDataScience, I want to get to him. Their Support email is Help@SuperDataScience.com. How many ways are there? Your email is not going to be Kirillbabygirl234. It’s going to be like Kirill, KirillEremenko, KEremenko, or EremenkoK.
Kirill Eremenko: My god, you’re going to ruin me. I’m going to get all these emails now. You’re right, you’re so right. I’ve thought of it so many times. It’s so easy to find people if you know their name and you know the company they work at. Oh my gosh. Just five variations. Totally right, man.
Rico Meinl: Yeah. It’s very low-risk to try it out. It literally takes like five minutes. Be specific about your questions. That’s one thing I learned. Because sometimes, I would just like shoot out emails to get in touch with people, and that’s just not the right way. If someone just asks you out of the blue, “Hey can we chat?” Obviously not, but if you’re asking very specific questions like … Also, funny thing, like John Doerr, he’s a very successful venture capitalist. He was the first to invest in Google. I think he has a net worth of 3.8 billion or something, or even more.
Rico Meinl: I saw this video of him online, and he was talking to, he was teaching a startup class. In the end, he said, like, “Oh, if you send me your five favorite books, I’ll send you mine.” That video was in like in 2016, and then he gave his email address. I sent him an email saying like, “Hey, John. How’s it going? Here are my five favorite books. I’d be super intrigued to hear about yours.” He got back to me, I think, a month later saying like, “Hey, Rico. Sorry for the late reply. These are my top five books. Hope you like them.”
Kirill Eremenko: That’s amazing, man. Amazing.
Rico Meinl: Yeah. Learning here is definitely … Because OKRs is a lot of overhead and a lot of process for a small startup, like an early-stage startup, so they wouldn’t recommend using it. My learning was, if you have believable people, because they were believable because they both implement OKRs in their company pretty successfully, so I guess they’re believable in OKRs. If you contact them and ask them for advice, it might potentially save you a lot of time.
Kirill Eremenko: Fantastic, yeah. Reach out. Reach out to people, don’t be shy. Find their emails or ask Rico. He’ll find them for you. Amazing. Amazing, man. Well, what a journey. I love it how you limited it to five learnings. It’s a good number to … More than that, you’d probably be hard to keep them in mind and remember, but this is good. This is really good. Let’s recap them quickly.
Kirill Eremenko: Number one was, start with a problem in your own life. Number two. What is number two?
Rico Meinl: Stay relentlessly focused on the end results that have the highest impact on your business strategy.
Kirill Eremenko: Gotcha.
Rico Meinl: Then, it’s not the money, it’s the “beep,” money. Make sure you’re worrying about the money a lot, because it’s what keeps you alive.
Kirill Eremenko: That’s true, yeah.
Rico Meinl: Careful with the latest research.
Kirill Eremenko: And, reach out to experts, in the end. [crosstalk 01:40:19] Very cool, very cool. I love our discussion. The Greeks had a saying that, “Truth is born in argument,” so if we agreed on everything, I think it would be … I think more insightful the way that we had disagreements along the way about your points, learnings number two and three, and to some extent. Yeah, man. I really enjoyed this chat. It’s probably one of the longest podcasts, but heck, it was really good.
Rico Meinl: Yeah.
Kirill Eremenko: Yeah. [crosstalk 01:40:52] Oh, you also wanted to mention some books, how you pick books.
Rico Meinl: Exactly, yeah. We had this rule during running the startup to essentially … Because as a startup founder, you know, you have almost like very limited time, especially in the beginning when you’re just out of the gate starting something. How many books can you read per year? Maybe up to 20. I don’t know-
Kirill Eremenko: I don’t know. 365?
Rico Meinl: I mean, audio books definitely made it easier for me to consume more content when I’m commuting somewhere. If you just look at books, there’s so much noise out there. I think there’s like 12 million books on Amazon. I checked when I did my own research.
Kirill Eremenko: Wow.
Rico Meinl: You just got to have a way to filter that. First of all, I wouldn’t say, this doesn’t apply to niches. For example, let’s say data science. You’ll have maybe a hard time to find something that was recommended by someone, with the rule I’m about to mention. On like high-level business stuff and like startup strategy, and so on, it’s a pretty, it’s a good filter. We filtered, we only read books that were recommended by billionaire founders. People who not only built a unicorn, but also became, had a net worth over one billion on the way.
Rico Meinl: It seems shallow to only put that down on the money level, but essentially, what we believe is that when you were able to build a company, when you’re able to become a billionaire with your company, if you look at the list of people, they all built products that made a huge, massively, majorly improved the world. I’m talking about, because I’m more from the tech side, I’m talking about more like tech people here.
Rico Meinl: I’m not that worried about the investors, but more like, look at the Mark Zuckerbergs, Larry Page. Like Bill Gates, Mark and Jason, Jack Dorsey, Kevin Systrom, this whole list of super inspiring people. We limited it down to only read books that were recommended by them, and it made it so much harder to find books that way. I recently put together the list on, it’s called Alicorn.blog. I put all books and podcasts that only feature these billionaire founders, and it’s still, I think it’s like almost 200, 300 books.
Kirill Eremenko: Wow.
Rico Meinl: It’s still like a vast amount of books. That was only within the weekend that I curated these books, so there’s going to be more to come. It’s ridiculous how much content there still there is, but it’s very … For example, you try to find a book on marketing and you have no idea about marketing and you want to find one. Drew Houston, the founder of Dropbox, and Joe Gebbia, the founder of Airbnb, recommended two books. It’ll probably be a good idea to read them if they really helped them succeed with their startup.
Rico Meinl: That’s how we selected the books. I have a list of 10 here that I read over the last year, and I think it will be really, really useful for people. The one I recently read was called The Start-Up of You, by Reid Hoffman. It’s also recommended by Jack Dorsey and Marc Andreessen. It’s about treating your career like a startup. I’m not going to describe what it’s about, but really cool book.
Rico Meinl: There’s book called The Goal. I don’t know if you know it. It’s like a management novel, basically. Jeff Bezos required everyone in Amazon’s senior level management team to read that book whenever they reached that point. It’s a really, really good book about the concept of critical path.
Kirill Eremenko: Give me one more. [crosstalk 01:44:45]
Rico Meinl: The Power of Habit.
Kirill Eremenko: Oh, I like that book, by Charles Duhigg.
Rico Meinl: Yeah, it’s amazing. Recommended by Ray Dalio. Really, really good book.
Kirill Eremenko: Very cool book, yeah. Man, that’s a really cool list. Where is it again?
Rico Meinl: I’ll put it in the show notes. I haven’t curated it yet, but we can put it in the show notes.
Kirill Eremenko: Yeah, please. Send it to me. I want to get a few books out of there just for myself to read as well. Sounds really exciting.
Rico Meinl: I will.
Kirill Eremenko: Man, very cool. We’re not going to go through all 10 or 50 right now.
Rico Meinl: Let’s not do that.
Kirill Eremenko: Let’s put them in the show notes. Our team will put them in there. Yeah, but how are you feeling overall? Started something, spent 10 months in it, failed. Like I asked you at the start of the podcast, are you feeling like world is over? This is the end of your startup life? What’s the next step for you?
Rico Meinl: Obviously, when we first put it on hold, I was kind of like a little depressed for a couple of days. Actually, on the weekend when we did the reflection and really looked back at it, it was great to see how it’s not a personal failure. It’s not like, “I suck, that’s why we failed.” It’s a lot of factors coming together, so that really gave me confidence. I mean, I’m back at work at Novomind for now. Definitely want to start the new year, I want to get more involved, maybe get a different opportunity because I’ve been there for a while now. I kind of want to move away from Germany. My next step would probably be London, because I’ve been saying I want to go to London for so long, and now I’m just going to do it and move there.
Rico Meinl: Just be more involved with actual AI teams that bring products into production. I’ve been talking to Ben, Ben Taylor a lot, who’s been doing great with his company, so that’s been a great influence definitely to be in touch with where the space is right now.
Kirill Eremenko: Cool, very cool.
Rico Meinl: I’ve been working on my project a little bit.
Kirill Eremenko: That’s exciting. If anyone listening is in London, make sure to hit up Rico. When are you going to be there?
Rico Meinl: I’m going there, probably early January just to connect with some people and then we’ll see.
Kirill Eremenko: By the time this podcast’s out, you’re already there. If anybody’s in London listening to this, hit Rico up and catch up for a coffee, lunch, dinner, whatever, and exchange some information. Always great to grow your network. What’s the best way to reach you?
Rico Meinl: LinkedIn, definitely. Actually, I have a website at RicoMeinl.com, so you can reach me. Otherwise, LinkedIn.
Kirill Eremenko: Oh, well, I’m going to spend some time guessing [crosstalk 01:47:42] your email. I don’t know. Might be Rico@RicoMeinl.com. Might be Me@RicoMeinl.com. Not that many-
Rico Meinl: There’s not many options.
Kirill Eremenko: Not that many options. Oh, man. Cool, cool hack. All right. Well, amazing. Glad to hear that. It’s interesting. I agree. Sometimes … This is what I’ve been telling my brother. He’s like, “Ooh, I want to start a business.” He’s only like 23. He’s like, “I’m going to rock the world. I’m going to build a big startup,” and things like that. I’m like, “Dude, you just finished uni. Go and work for a company. Get some experience. Learn some stuff. Learn from people who are doing it. See what it’s like to be in a business.”
Kirill Eremenko: It’s a whole different thing. You get this … I don’t know. You just kind of like get this sense of groundedness once you’ve seen it. Not from a, like as you say, fast-paced, startup world, but you’ve seen it done gradually, monotonously, but correctly and progressively in a running business. Preferably like a mid to end, or late type of startup, or even a big enterprise. That gives you the sense of you know what to aim for down the line.
Kirill Eremenko: Because if you’re running so fast and you don’t have a target, it’s really hard to run towards a target that’s always moving in your head. Once you’ve seen it, once you’ve felt it, like you say, you want to get some experience in a company that’s already productized, that’s using AI, that’s growing. Of course, you’re going to bring value to them but in exchange, you’re also going to see what it is you’re aiming for, and that will always act … I think, for me, having worked at Deloitte for two years, always acts as a North Star. I know what a proper, organized, serious business looks like inside. As much as I’m having chaos with my startups, I know where I’m going.
Rico Meinl: I mean, it’s also, I’m not going to start another big point here, but there’s this whole thing of, have you heard this before that, “Some people don’t have 10 years of experience, they have 10 times that one year of experience”?
Kirill Eremenko: No. I haven’t, but it sounds pretty good.
Rico Meinl: I think that’s something to be aware of. That 10 years of experience doesn’t mean that that person has 10 times more skill than someone with one year of experience. I think, again, it really depends on the person and what kind of learnings you take from that.
Kirill Eremenko: Yeah, yeah. [crosstalk 01:50:11] That’s true, that’s true.
Rico Meinl: I would agree with your point, generally, but it doesn’t apply to everyone.
Kirill Eremenko: Yeah, yeah. I agree as well. Everyone’s journey is different, but I’m glad to hear you have ideas for you where you want to go next. You never know how it’ll work out. Hopefully, you’ll move to London and see from there what happens. It’s always good to have like a compass, some general direction in which you’re going. There’s a saying that, “Entrepreneurs are people who are adamant about their beliefs, but they’re very easy and willing to change their beliefs in a heartbeat.”
Rico Meinl: Right, right. I mean, you know me. I’m super open. I’ve been talking to your partner, Hadelin, recently. He’s really told me great things about Dubai, so that’s definitely something. I was at the startup, the AI conference last week in Berlin. I heard great things about Tel Aviv. I mean, I think China would be an interesting experience. I’m open. I really like to go places and experience different cultures too.
Kirill Eremenko: Gotcha.
Rico Meinl: I’m going to go where stuff is happening, which I think is London or something like Dubai or Tel Aviv.
Kirill Eremenko: Nice, man. Nice. Well, keep us updated and as I said earlier before the podcast, would love to have you as a speaker at DataScienceGO, Berlin, I think.
Rico Meinl: That would be amazing, yeah.
Kirill Eremenko: Yeah, man. That’s where people can catch you if anybody’s interested in learning more about your story and meeting you in person, man. So cool. You’re so tall. You’re taller than Hadelin. It’s insane, it’s insane, man. All right, my friend. It’s been a pleasure. Amazing podcast.
Rico Meinl: Thank you.
Kirill Eremenko: The world’s going to love this. Thank you so much, once again, for coming on the show. It’s always great.
Rico Meinl: Thank you. Appreciate it.
Kirill Eremenko: Thank you very much, ladies and gentlemen, for listening to this super long podcast. I really appreciate you staying until the end. We want to share stuff with you, some really amazing things, so head on over to the show notes at SuperDataScience.com/335. There you will find not only how to contact Rico., but also all of the materials mentioned on this podcast, including the list of books that Rico mentioned, and including a link to the blog post on Medium with Rico’s learnings. If you want to recap on them, you can definitely find them there.
Kirill Eremenko: What a conversation. I totally, totally enjoyed chatting to Rico, and I hope you enjoyed our chat too. If you are thinking of starting an AI startup, then surely, 100%, there’s very valuable insights here, very valuable lessons. Things that you can already take away and apply in your own journey. If you know somebody who is interested in starting up a company in the space of artificial intelligence, send them this podcast. It might save them a lot of headache down the way.
Kirill Eremenko: Very easy to send, very easy to share. SuperDataScience.com/335. Of course, connect with Rico. You heard he’s going to be in London in early 2020. Connect with him there. If you live in London, I know we have a lot of people listening to this podcast from there. I was there, I met a lot of you guys. Connect with Rico as well, and also he’s coming to DataScienceGO in Berlin, which is happening in quarter two, 2020.
Kirill Eremenko: I can’t give away much more information than that for now, but if you’re going to be in Berlin in that time, make sure to connect with him as well. If you see our announcements about DataScienceGO, it will be very limited. It’s the first time we’re doing it in Europe, very limited. Make sure to grab your seat because that’s a great way to connect with Rico and people like Rico as well.
Kirill Eremenko: On that note, thanks so much for being here today. I am super grateful that you are part of the SuperDataScience community. Can’t wait to see you back here next time. Until then, happy analyzing.
Show All

Share on

Related Podcasts