58. beCommerce with Julien Lemoine

Julien Lemoine



Markus Lorenz  0:48  

Welcome to a new episode of the eCommerce podcast. We're talking about search technology today. What is the first company you think if you hear search technology? Are you thinking of big players in the US? For sure. I'm thinking about Algolia, the search and recommendation engine founded by Julien Lemoine in France. Algolia raised 150 millions in the series D funding at a post money valuation of 2.25 billion. That's totally impressive. I'm really thrilled to announce that we have Julien today with us as a guest on our podcast. We will talk about how to scale an engineering team trends route search technology. Julian as CTO of Algolia, one of the world's leading search software's company, scaled his team to over 250 employees in seven different offices. But Julien, please introduce yourself.

Julien Lemoine  1:47  

Hello, yes, I'm Julien, I started as a software engineer. I actually started in the Research Center working on text mining, and natural language processing technologies. And after working in research, I wanted to have something more to do with real customers instead of a buttons or a publication. So I joined a startup very quickly when I started to work. And this startup was working on search engine technology. So I work during all of my career in this field. And after working for a few startups, I decided to launch Algolia with one of my former coworkers. And it's basically because we decided and we looked at different companies that were struggling to have a decent search. So what we wanted to be was a newest solution for developers to help them have a better search in the applications.

Markus Lorenz  2:57  

Perfect. So thanks for the introduction. Maybe we can start with the question of how it all started? And how was it growing? I mean, you introduced a little bit how you found it that you have a co-founder alongside but why did you decide to go with a search engine technology, was it more done by accident, or what was the root cause here.

Julien Lemoine  3:19  

So at the beginning of my career, it was more by accidents. As I said, in my first position I worked in the text mining/data mining field, which is very close to search engines. So this one was more by accident. But then I really discovered a field I deeply love. After working on search engines and actually launching a startup, working on a search engine was based on all the problems that exist. And the complexity of this field, like search engines, is quite complex. And there are a lot of sub problems to solve. So it's pretty much a domain where you can spend your whole life solving problems, you still have a lot of problems to solve. So it's more like, by working on this field and discovering over those problems or those solutions that we had to build that I just spend all my time in this field.

Markus Lorenz  4:31  

Nice. Yeah, we will come later on to that point, which you have to tackle by using search engines probably and the requirements you have to solve as a company or also as a tech vendor to that, but basically I would be interested in the team. So how was the team back in the days and how do you manage the team because actually, we're running a quite big company with over 250 employees. But you started with two so how was the development and how did you manage to scale the team back in the days?

Julien Lemoine  5:03  

We are more than 500 today. But for the first year, we were only the two founders. In the first year, like you have to do the product and market fit like, and it's better to do that in a very small team. So it was 2013, end of 14, we were 10 people, when we reached the first million of revenue for the company, and we grew very quickly from those 10 people, we were 60 to two years after in 2016. 300 in 2018. And today, we are just a bit above 500. 

Markus Lorenz  4:31 

That's totally amazing. 

Julien Lemoine  5:03  

Yeah, it's it's quite challenging to keep the culture on keep a good momentum as you scale. One of the best advice we got from one of our seed investor very early was to very quickly get an internal recruiting team. And it was kind of counterintuitive, where we are so small, like, as I mentioned, when we were like before 10 people, when we get our first advice, where you're someone that it can be justified very quickly, as soon as you're more than 10 people in a year, it's justified. And it's a game changer because you have someone that is inside the company, and they can already talk about the company way better than any like agency that helps you hiring when they are not after your team.

Markus Lorenz  6:42  

Yeah, absolutely, I can totally underline that. I mean, even from the financial figures, if you're paying headhunters or something like that, you're just paying, in Germany, at least a fee from 25 to 30% of the year salary. So if you hired 10 people a year, basically, you have more or less a pay off, and you have somebody who's can can transport the culture directly to the to the employee, and those guys who were, yeah, the further employees of your company. Maybe you can tell us a little bit, what challenges have you face over that time? And how do you tackle it because scaling such a big team, as you mentioned, is quite challenging. And you have different angles, how you can take on what was the most crucial for you.

Julien Lemoine  7:29  

Indeed, it's difficult, there are a lot of challenges. The biggest one is culture and values. And keeping it while scaling so fast. So it's very important to really like the team under one key objective one mission. And it's especially out when you are people with so many different backgrounds, they have different experiences and different ideas. And even if you define your culture with words, which everyone is doing, people will put a default meaning behind those words. So it's really important to share context, share examples, share exactly what you mean with this culture, and constantly work on it and constantly refine and update. So to give you an idea we were and we are still doing like those one, one session on values for all new hires, and explaining exactly what we mean with a culture. And what we want is a culture, which is not words on a wall, but something which is used every single day in every decision. So making sure we stay true. And we detect when, but it's not the case, and we can fix it. So it's a very deep investment. And it's something where you need to work on it like pretty much every single day.

Markus Lorenz  8:56  

Yeah, you're absolutely right. Culture is a totally important topic. And we discussed that also with Benoit from Akeneo, when he was also mentioning that culture is the most important thing that you have to shape for a company like that. Pretty interesting. Now you have over 500 employees, like you mentioned, seven offices worldwide, and how was your role changing? Basically, I guess you have absolutely different roles and challenges nowadays than back in the days.

Julien Lemoine  9:24  

Oh, yes, of course. In a fast growing company, like my role changes every six months, pretty much when I say change, I mean it's completely different. And one of the big challenges is that you need to constantly check, whereas this kills you in need for this new role in the next six months. So in the early days, I was one of the many engineers working on the search engine itself. So working on the code base, and I progressively moved to a whole of software architects were I was more like, looking over at the technology's direction, and what has to be done to stay relevant on the markets. Of course, today, I'm not coding anymore. And I'm still spending a lot of time on the architecture and technical direction of the company. But I stopped to code or to work on the code base in 2018, where we were on real bad people.

Markus Lorenz  10:31  

Yeah, interesting. I can totally underline that as well, that you have to pivot your role. And just looking in front of yourself, what's coming up there and be ahead of the game, but at the same time, be supportive as a leader to the team. And in your opinion, regarding the leadership, what is your, in your opinion, as the CTO the most crucial thing by scaling an engineering team.

Julien Lemoine  10:57  

So I think it's not only the CTO, it's the CTO and the VP of engineering. So every company has a different partition of the hole between CTO and VP engineering. In our case, I was one working on the technology on the way. What a good direction in terms of technology to stay highly differentiated, was wide Silva, our VP of engineering was working on the processes and making sure that all the teams stay highly motivated. But some people do too, as a team and of different profiles.

Markus Lorenz  11:36  

So you're constantly changing the company and by scaling it, which is absolutely valid. What in the end needs a permanent transition into a company? And from your perspective, what should transition look like if you're just getting an engineering team from one level to the next within the growth phase of a company or even on a technology level?

Julien Lemoine  12:00  

Yes, a lot of transitions indeed. And I think key conditions for CTO is to move away from coding were keeping, while being close to the technology itself on show, we are still early rounds. And for me, the big responsibility of the CTO is to make sure the technology is relevant for the market need. And than it can scale with the company. So I was having, although these CTO and CPO, Chief Product Officer ads until very recently, so I think splitting those all is also one of the very key conditions for a company you cannot keep booth hats for an infinite amount of time. And the moment you split really product and engineering is a key moment in the company where you need to split the responsibilities between two people. So there's going to be transitions.

Markus Lorenz  13:03  

Now, I guess a CPO or somebody who's taking really deeply care about the product is absolutely key. And then you have a clear split in the responsibilities. And there's one person who said what are we going to build? What are the market demands? And the other person who's saying how we are going to build with our tech team and the engineers, that makes total sense for me. Yeah, after the insights about the company and the team, I'm curious to speak over search engine technology itself, because everyone is today aware of search engine technology like Google, but I guess even then the income there are crucial for the user journey. So let's jump into search technology. 

Search technologies are in my opinion about converting traffic into revenue, you know, you get traffic on the platform, you try to convert it into sales by fulfilling the exact match from the search or the perfect match from the search. And in most cases, which we are actually observing, the search is underperforming. So merchants are losing money here, because well, leading brands are actually investing a lot of money in the search capabilities. They built maybe their own search sec, which is not affordable for small emergencies. And they may also have teams to research on data with specialists. And most of the smaller merchants can't afford the tech stack like that. So what's your opinion about building search capabilities and an emergent world about people tech and process and build versus buy approach? What would you advise a merchant who would like to set up his search capabilities in the future?

Julien Lemoine  14:43  

That's a big topic. And that's actually the reason why we started Algolia. Yeah, like we observed by doing consulting, but companies are spending a huge amount of time to iterate on their search. And most of the time, it's just to catch up with competition. So search is so cool to a lot of businesses that they need to control. And this is why I think we are not anymore in this build vs buy mindset. And we are more now in a mindset of buy and build. Like people will buy an API like Algolia, or the value service and develop on top of it. And the key elements for all the companies worldwide is to iterate faster. So you want to improve their search with a smaller cycle. So they can constantly improve it and make sure they have the best conversion rate possible. And this API is really playing a key role, because we package a lot of functionality, a lot of tools like debugging tools to IDN L, developers to be more productive and have a smaller cycle.

Markus Lorenz  16:06  

Yeah, and in the end, increase the outcome of the user journey, because this is any end where the revenue is generated. And then you're able to consume those API's from your product in the end and combine it with the user experience in the front end, which helps the client a lot. Not focusing on things like hosting or deployment or something like that, on the technology itself, totally got that point. From my perspective, merchants with the highest set of SK use, maybe like Amazon, have to understand the user intent and everyone wants to be user centric today. So standard search engines are good at finding nearly the exact match. Like, I don't know, red lamps or something like that. I today makes the world better and understands user intent, like red desk lamp with ambient lightning, you know. But this means in the end, that you're semantical and able to parse data. So the search engine has to understand the intent, slice the search term and weigh the different terms. In my opinion, in our example, that means our searching for a hat or searching for a lamp and what makes the difference here. And therefore, it's crucial to have enough data in your PIM system or our product experience management system that you can utilize later on. And how do you actually dealing with that and other requirements? And why is Algolia the most sexy technology here?

Julien Lemoine  17:36  

Yes, indeed, understanding user intentions is the world purpose of search. The goal is to understand the user intent, and it can be more or less complex, depending on the query. And keywords, or AI based machines have very different pros and cons. It's a very different balance. And QA keyword search today is not enough anymore. And we have to leverage user behavior via a machine learning algorithm to pretty much fix keyword search. You mentioned Amazon, that big challenge of marketplaces is that they have content which is directly written by the customer. So the data is quite different from one customer to another. Some even try to play with the engine. And to be first on every query. So you have some keywords that I did on that product to try to be first on pretty much every every query. And to do that what we do is that we use a V of user clicks conversion of users are related to the query to have a direct impact on ranking so it's not anymore I don't think today there is any search engine which are only relying on keyword search, all of them are an honest with some figures. And there are a lot of default approaches. But most of them leverage over signals, you have like signals from the business, you have like the products that are sold the most, but you have a lot of behavior from the users from the peaks from the products that are both finally customer and all of this is now used in the banking. More recently there is a new type of machine learning that intends to remove all the natural language processing NLP layer of keyword search. And this layer is often called semantic search even if the term is in use for 20 years. This is like an end to end deep learning model to completely replace a search engine. So you don't have any keyword inverted list you have directly this Victor search. And those techniques provide some more effects on some queries where the result is just amazing compared to a pure traditional keyword search. But they are over queries where the results are very poor compared to a keyword search. That's why I was mentioning legs and the pros and cons of food. I deeply believe we will soon see, I breed systems that will mix both to try to get the supposed benefit of both systems and get the positive effects to improve again, the conversion rates for the e-commerce website because that ultimate goal of predicting into customers is super interesting.

58. beCommerce with Julien Lemoine

Continue Reading