Make Your App As VIRAL As TikTok With This Gamechanger - Ben Gutkovich (Superlinked) | Startup Scout

Ben, thank you so much for joining me on the show today. It's such a delight to have you here, especially after we actually met quite some time ago, all the way back in November in Lisbon at the Web Summit, which was a fantastic event. And we met there when you were presenting at the stand about Superlinked. And I was really intrigued by what you're doing. I think you're onto something quite significant. It's a very special product and I wanted to have you on the show to talk to you about the product, about your background and about where everything is going. So first of all, thank you so much for coming on the show.

My pleasure. Nick, good to see you again. And I'm happy that we finally got to do it.

Before we get into about Superlinked itself, let's talk a little bit about you and about your background. Can you tell us a little bit about yourself, Ben?

Sure, absolutely. So I'm software engineering by training, kind of. First half of my career was in software technology, product management. Right. Worked for quite large corporations and also had my first startup back in 2005. So quite some time ago where I really fell in love with the whole idea of, you know, delivering impact, building something from the ground up and then seeing users using it and join it and getting value. Right. Came to the UK about 11 years ago to do my MBA and then went down the more commercial route. Joined McKins and company and strategy and had another startup in the middle which I joined as an employee. In this case, the startup was acquired by a US player. And then finally about two and a half years ago, just before the pandemic, I decided that's it, time to go back to startups and to building my own business.

What was that startup that you're a part of that then got acquired?

So it was EasyCar Club. I don't know if you're familiar, it's a peer to peer car rental marketplace. We were UK focused and did some really good things in the uk. Quite innovative. For instance, we had a deal with a leasing company where you can lease a car and actually pay for it by renting it out five, six days a month, which was first in the UK and one of the first in the world. But then big American players came into the market, get around in Turo with a lot of money and obviously global footprint and ability to scale. And then at that point we decided that it would be better for us to sell.

What are some of the lessons from your experience with that process that you'd give as advice to somebody else going through that process now?

Yeah, absolutely, I think. Well, I mean, first part is obviously seeing that there is some strategic fit and cultural fit between the companies. But in our case, I would say the acquisition wasn't very typical. Mostly they acquired the users and moved them to their infrastructure. Because in the end the propositions were very similar. There was no point of maintaining both brands. So the idea was there was a short transition period between that, but in the end it was basically moving our customers and mainly the car owners, which were quite difficult to acquire. And then Turo entered the UK working mostly with boutique car rental companies rather than individual owners, while we built a database and a platform with 12,000 owners. And at that point of time across the UK, mainly in London, but across the UK, and they really wanted that part. So it was mostly integrating our users into their platform and migrating them. But yeah, I think there were a lot of industry events which was great within this whole Sharing Economy space. And our CEO was also the chairman of Sharing Economy uk, so we brought together a lot of those companies that created the relationship which eventually led to the acquisition.

Working in larger companies, it seems like you always had a bit of a desire to work in smaller, more innovative, more exciting startups. Why is that?

I think it's all about impact and ability to deliver impact. I mean, I wouldn't say working in large corporates is bad or not satisfying. Definitely was when I worked for Orange in the kind of golden age of telecoms. We build our own Spotify version, our own Netflix version, we had our own app store. Of course, everything was controlled by the operator and we innovated a lot. And having millions of people seeing your product, using it, providing feedback, is very, very satisfied. On the other hand, getting things done in a big corporate is hard. There's a lot of politics, there is a lot of competing interests which are not necessarily all aligned to the success of the company. And that's why I prefer small organizations. If you work in a startup of 10, 15, 20, 30, 100 people, everyone has to run in the same direction, otherwise the company has no chance. This idea of people being committed, people being driven, passionate about what they're doing and kind of running forward is really appealing for me. Right. And, you know, at McKinsey, it was something I've really enjoyed. For instance, you work in very small teams, right. On a project. Yes. You work with large corporates. Yes. There's a lot of politics within the corporate, but you are surrounded by a group of people who are dedicated, super smart, motivated and just kind of Riding in the same direction. And for me, this is the kind of environment where I thrive. Right, right.

How did you come up with super links? What's the origin story of the company?

Absolutely. So we started during the pandemic that was kind of end of 2020, and the whole social media online community space was just booming and exploding. Of course, people had nowhere to go. And as we know, humans are intrinsically social. They want to connect with other people, exchange experiences. And that happened more and more online during that period of time. What we realized is those communities are scaling and growing, and there is a lot of value in those online communities. Right, which is not being leveraged. And what we started from building was actually a marketplace for communities. So you could deploy a marketplace within your community, let's say community of developers or community of entrepreneurs, and then ask for things within that community through the marketplace, but while offering a reward. Right. So if you think about the community of developers or technical people, recruitment is one obvious use case. Right. There are people in that community who are recruiting and there are people who are looking for jobs. How do we now bring them together and actually help them find each other? And we thought that this marketplace in incentives would help. And we were able to generate a lot of demand. So people were interested in hiring, others were interested in finding business development opportunities, investors, et cetera, and willing to pay for it. I think we had at some point about 200,000 pounds of opportunities on the platform. The challenge was, well, there was twofold challenge. One is knowing who to collaborate with, right? You go into community of 30, 40, 50,000 people. How do you find the right person? How do you make sure that that opportunity or request that I posted actually is presented to that user who can help fill it? Right. And the second one is that the engagement, Right. So during the pandemic, the average was about six to eight communities per person, online communities per person. This is excluding Facebook, Twitter, et cetera, the big social media platforms. So people joined a bunch of communities, but actually staying engaged with each of them, making sure you find things which are relevant for you, which was very difficult. Right. And so we built this, we had this marketplace, we deployed it in multiple communities, started building the demand, but with the supply side, we got some opportunities, actually. Opportunities or request field, but not enough to actually justify the, you know, the VC scale for this, for this business. And we started trying to understand around those two things. Okay, so how do those communities actually present the right stuff to the right people and connect between people who should be collaborating together? And my co Founder who is an ex Googler and an amazing developer, super smart, brilliant, just kind of in a couple of days build a few scripts that basically Python scripts of course, that basically match people around those opportunities, right? So looking at the profile of a person and looking at the profile of the other people in the community and looking at the opportunity, who are the people that should be collaborating? And we started actually quite manually with this, right? So we would send those recommendations to community manager and ask them to make the connections. And it worked very well, right? So we saw that actually there was value there because people sign up to the communities, they have the best intentions, they want to collaborate, they want to engage, but then they come in and oh my God, they missed like three days of content. Now it's all a huge, huge thread. How do you figure out what is relevant for you? And this is what we started doing and what we build on top of that is, well, that was kind of initial product is kind of a network introduction app for communities, which we still run with about 25 communities, entrepreneurial, business, networking, et cetera, which basically connects people around the topics they care about, but not just kind of randomly or not just based on the topic, but also based on their profile and their information and their activity in the community. So we have very, very high quality matches within that community. We launched that in a bunch of communities and started getting feedback. So the two things that were very encouraging, first of all, people really used it, right? So we still had about three months retention was around 40%. When you compare it to other kind of social media apps, this is more around 20, maybe even less percent after three months. So we saw that there is engagement with the app, so that that gave us confidence. And second, the feedback, right. For me, generally it's very satisfying to see customers using what I built and given feedback and getting value out of it. And people were just raving about it, great connection or we got to stay in touch or we found the business opportunity, et cetera. And that was amazing, right? So that gave us the confidence that this is a problem worth solving, right? And that our solution has a potential to be relevant and valuable one. So what we did next, and that's Superlink now, is basically generalizing the problem, right? We had this infrastructure which understands people, or as my co founder would call it, user modeling infrastructure, because he's been doing it actually at YouTube for seven years, to really understanding what the users want, what are they interested in, what you present to them, both in terms of content and ads. So user modeling infrastructure and we had this app which is an introduction app for communities to build engagement and connection. So we said, well, we have the infrastructure, let's generalize the problem and let anyone who wants to build an engaging social experience actually use our infrastructure to do that within their product. Most of the communities didn't have their own development capabilities. They either run on Slack or maybe a Facebook group or maybe a LinkedIn group or maybe some of the other community platforms like Disciple or Mighty Network, Circle, et cetera. There are a few of those, but they can't really develop. So our app was a bit of a standalone product besides the community. Right. So we said, well there are products which will value social interaction, where social interaction gives value. So let's help those products deliver it without the need to hire data scientists, machine learning engineers, data engineers figuring out this whole landscape of machine learning, data science propositions there in the market, but rather just connecting through APIs, sending user data and I can show a chart later on sending user data and then basically receiving recommendation whether of relevant users, relevant content, similar users who might be interested in the same thing, et cetera, just to make sure that their experience is relevant. So that person that goes back to the community or to the product and gets overwhelmed with a bunch of new information doesn't have that experience. Right. Actually they see only the things which are relevant for them.

This personalization engine, the idea is you take this, you, you could even have customers the likes of which could be Circle or Disciple or whomever and saying, hey, we've got this personalization engine, we can plug it into your back end. And therefore what your users are going to see is going to be much more tailored to their needs, to their desires. They're going to make much stronger connections as a result of having Super Linked embedded into, into their product.

Yeah, absolutely. And we are in conversations with a bunch of those community platforms, event platforms and also standalone products which want to leverage the power of this social plus experience. Right, Social plus the functionality which the products deliver experience. So yeah, absolutely. And then, you know, we did a few integrations with event platforms already where again, you know, you come to an event, there is a bunch of content, there's a bunch of people, attendees, there are some rule based systems that help you identify the relevant users or you can go and search. But what we saw from the data is that people don't actually do it right. They want a recommendation which is super relevant for them, that they don't need to invest anything in order to get it. And that's what we power Right. We get those more relevant recommendation actually based on user profiles and user behavior on the platform to understand what they're actually interested in.

Are you currently working with customers right now or proto or beta customers who are trialing the Super Linked product or how is that working currently?

Yes, yes. So we have the communities which basically run on the same infrastructure, which have introduction and we have now roughly 3,000 users there. And then we have what we would call design partners with whom we are building different use cases. So for example, we are working with a US based social network which is kind of LinkedIn for Web3, right? So one of those LinkedIn unbundlers who trying to take a vertical out of LinkedIn and provide a more specific, more specialized experience in that vertical. So of course they have crypto tokens, upvotes and things like that, which actually benefits and compensates creators to an extent. And they had a rule based recommendation engine just says, oh, I'll show you the content which has the most likes, etc. But what happens as a result is they lose the new creators or if I'm a new creator, I come in, I don't get my likes very quickly. What is my incentive to go and create content? Right. So we actually launched with them just, just a few weeks ago, right? To show feed, which is specifically personalized per user, not this one feed that everyone sees, which is just based on recency and popularity of content, but rather things which are relevant for the person and taken into account how they engaged with the product or the app. Right? And there are a few more of those kind of proof of concepts and pilots, et cetera, at this point of time. Because our goal at this point of time is to develop those use cases, right? The event platform, the community platform, the social network, the job board, right? Where we are now designing a solution with a job board for executives in the US with 300,000 users, which gets very little click through and engagement on the jobs they post and they send, like many job boards do, send a bunch of emails and get almost zero engagement. They want to send much fewer emails, but then much more targeted, much more relevant to the users to increase the engagement. So we are building those kind of, let's say vertical use case because our platform is horizontal, getting the results and the KPIs and then of course we'll see where we want to double down or maybe continue working, working with multiple clients. I mean, in the end we are still at a very early stage and small company and don't have a huge capacity to work with A bunch of different customers or different verticals.

Is the model which is going to be given to me the same as the model which another social network has, or is it tailored to my particular needs and to the strategic direction that I want to go with my network?

That, that's an excellent question. So one of the benefits of what we do, because we've been working with all of those communities and we have all the data and we did all those connections, etc. We have pre trained models, right? They are trained specifically in the professional context because most of the communities we work with are in the professional space. So we have pre trained model that we can deploy. And basically one of the challenges a lot of companies raise when we talk about data science and machine learning, hey, we don't have enough data. And for smaller social networks that is true. Therefore we have a pre trained model on tens of thousands of profiles that will work out of the box. So that's one. But we also build a configuration capability which at the moment we support the client with doing it, just also because we want to learn, but also because we were very much working in tandem with the clients to deliver the KPI they're interested in or what they care about. But there is a configuration file which is the place where you can actually configure the model to your specific use case. And this is where kind of the IP and the understanding of the clients will go into. And that's where the difference between the models will, will materialize. Right. And those configurations, if you think about it, for jobs, right? I want to show jobs which are at the relevant level for the person. Right. So if the person is some kind of a director or senior person, you don't want to show them intern jobs, no matter how well the description matches their profile or their cv. Right. So this kind of seniority feature, if you like, would be a special feature for those, for this case, for this specific job board. For a social network, recency is important, right? For jobs, maybe not as much. Okay, well, the job was posted two weeks ago. Yeah, still open. Why won't I show you? But if this is a social network where people post all the time and they talk about current events, you don't want to show something which is two weeks old. Again, this is a feature which will be relevant for a social network. And this is all configuration, right? In the end, yes, we definitely want the clients to be able to do those configurations themselves. And then once we set up those features, once we put the right, well, weights, let's say weights, on those different features, then the model keeps training on the customer data and actually gets closer to the customer use case learners, the users, et cetera. Right. Hope that makes sense.

Understood. And the idea would be that in a company which maybe is a bit further along, has lots of data, probably will have its own data science team. The idea would be that superlinked would be almost kind of handed over to that data science team and they would work with your in house, you know, who's ever taking, is taking care of customer success and there would be a collaboration between the two. Is that sort of the direction that you see yourself going?

Yeah, it's a very good question. So we currently we mostly support the use case of someone who doesn't have machine learning data science expertise in house. But the next step for us is definitely generalizing this, this infrastructure even further and letting clients deploy their own model. Right. Benefiting from our evaluation framework. So one thing that we do already with that social network, we actually deploy multiple models or models versus algorithms, right. And we can measure the results so we can a b test that. And this is actually not trivial because obviously you want the same users to get the same kind of experience no matter if when they come back. So it's not just per session. You need to be consistent for the whole user lifetime or that kind of the testing period. And the second thing is the real timeness of it. So we build infrastructure really kind of pushing the envelope of what's possible currently in the space. Build a model which, well, that one part is easy but worth mentioning. Return results with less than 100, 200 milliseconds. So this is very much in real time, but it also updates the user properties and the user model within less than two seconds. So whenever the user already gets to the next page, they will see something that will be influenced by what they did on the previous page. So this is what you know, some might call tiktokization of user experience, right? So when you're on TikTok and you watch a video, right, the next one will already be influenced by how you interacted with the video you're currently watching. And this is what we support, right? So if we talk about the social feed, right, Once you scroll the next page, your feed will already be updated based on what you interacted with on the first page. If you talk about jobs, same, right? You clicked on a job, you like the job you applied to a job, your whole feed is different, right? The jobs you will see are completely different.

When you're talking about personalization, if you're able to pick up on certain data points that then are allowing customization of the experience to surface more relevant information and content to a user. It seems to me that there's so many different use cases for Superlinked and maybe I'm being crazy here, but I'm thinking news sites, I'm thinking E commerce even. Have you considered those different paths as well?

Yeah, absolutely. And news is definitely an area or a vertical where we're looking at. I mean in the end news are not very different from social media. The one big difference there is that it's not a user generated content. Right. So the news are produced by a writer. Writer can use labels, right. Or tagline or tags. Right. In order to classify the content and that actually determines where this content is presented. Right now there is actually a new news website that familiar with which is using very similar kind of techniques for personalization. So I do think it will go there. But at the moment with the tags it's okay. Ish, right? It kind of works, right? Interestingly enough, we did speak with an advertising network, you know, with, with the third party cookies going away hopefully soon because you know, Google been postponing for a bit, but now in 2024, advertisers, especially on publisher websites, that they struggle a bit because most of the traffic is not locked in. So how do you actually provide a relevant ad to the user? And a lot of that depends on context. So what is actually in the article? Right, but with our tech, even within the session, we can already learn the user path and give them more relevant ads. We decided to at the moment not to go down that route. And the reason was that it's a completely different use case from the others we deal with. Right, but as you say, there are a lot of opportunities, a lot of different use cases. E commerce is an interesting one and there we consciously decided that it's probably not the right space for us because the real. Well, although guys like Shane and a few others are going towards this kind of real time commerce where you just kind of swipe products and you buy on the spot. So, so that, that, that is an area where it could make sense. But your typical standard E commerce would have a relatively limited catalog. Right? And you will, you will also be able to. And this data is structured. So products is a structured data. You have the category, you have the price range, you have the color, you have the men, women, etc. So, so there's a lot of data already embedded within, similar to this news article where you have tags and you can use those tags for matching and recommendations. And second, the real timeness is less critical, right? Because yeah, it's not like on TikTok where you buy one product, buy second product, buy third product, right? So you like YouTube with TikTok videos, but rather you go into the website maybe looking for something, maybe not looking for anything, but I can batch process the data maybe overnight, maybe once every four or five, six hours and still give you pretty good recommendation based on what other users like you bought or the other users that have the same kind of transaction history as you bought. Our real value and differentiator is in the real timeness of it, right? And that's where real time, this is one part and second is kind of unstructured user generated content, right? Where you don't have those tags or labels to be able to classify it easily. And that's where we excel, right. And we don't see a lot of other solutions in the space. Most of the solutions are actually homegrown, right? And the homegrown solutions are obviously hard to build, require this kind of data science, machine learning, expertise, hard to maintain, et cetera. And this is all of those things that we take on ourselves to make the product better. There are some solutions in the space like Algolia Recommend or AWS Personalize would be some of the bigger solutions in the E commerce space. But they have two challenges that we think are important. Well, what we hear from the clients are important for them. One is the black box, right? You feed in the data, you get the recommendation, you have no control, you can put those configuration specific labels, you cannot say oh this is more important for me than this. So this is one thing which we hear from customers and we solve it through the configuration file and the ability to actually for you to customize the model for your use case and for your product. And the second part is the real timeness. Both of them watch work and batch processing. I mean AWS Personalize the minimum training frame is two hours, right? If you go to a social media website, you can't wait two hours for it to give you the relevant content or recommend relevant users to follow because they come in, they don't like what they see, they don't find it useful, they're gone. Right? I mean we see churn rates 95+% on social networks, right? Within the seven days. Because people come in, they want to see something relevant. If they don't, they don't bother to do it anymore.

Before the call we were running through a deck that you had put together and there were some really interesting slides there. I don't know if you would like to show some of the slides from that and we can touch upon some of the more interesting points or we can, we can kind of skip through whatever makes sense for you.

But so maybe some examples. First of all, so what is the Social plus or Strava is a good example of social plus fitness, right? So this is basically a fitness app, a fitness tracker, but it has a bunch of social features which make it better. People can compete with each other, can collaborate with each other, can organize social, social fitness activities together as a cycling ride, run or anything like that. And Strava in this case actually owns the users. It's not Facebook, right? If you build your product on Facebook, Facebook owns the users and they can decide what to show your users. Just not great. In this case, Strava is the one that is able to show and decide what the user should see. Another one is of course Duolingo with Social plus language learning or social plus learning in general, if you like. Again, very social. There is a feed you can follow, friends you can join, clubs you can learn together and that makes the product better. I mean there is how many language apps are there, how many fitness apps are there? But those two stand out. One of the reasons is the social features and the ability to actually interact with your friends as you do the activity. And you know there are some benefits, right, of what social brings, right? Or let's say integrating social features in your product. One is this natural growth loops. If I am on the app and I can invite my friends and I can collaborate with my friends, that's great, right? Naturally, right? People will start inviting people they want to collaborate with on this app. You know, Figma is example, Canva is example, and the two others we just talked about also examples, user to user engagement makes your life easier, right? If it was just, let's say Strava was just a fitness app, you probably would need to create a bunch of content interviews, bring in some personal coaches who would talk about the benefits of a certain diet or exercise regimes, etc. You get it for free. Basically, users exchange information with each other, comment on each other, generate content. So, so, so it's just cheaper to keep the users engaged. Retention. If your friends are there, you're there. And then of course, competitive mode. Because once you build this community of, you know, millions of users who are engaged, who have their data there, who have their friends there, even if the competitor comes up with a better product, it's not that easy to switch. Right. The switching cost for users increases and obviously as a business, this is what you want. I mean, you know, we kind of touched on this a little bit. Why, why social matters is because we are in the attention economy. People have short attention spans. They don't have patience or time to actually explore, search, find things which are relevant for them. If you don't give them what they're looking for within, you know, some data here talks about 15 seconds, 5 seconds, whatever, right? You lose them, right. And it costs a lot because there's so much competition to bring those users to your website or to your app. You really don't want to lose them. So you better off showing them what they want very quickly. Right. And then of course, one of the developments of the past years is that there's a lot of machine learning, data science, infrastructure modeling, testing, deployment, MLOps, et cetera, startups and companies out there. And this is kind of actually not a full screenshot of the landscape made by one of the VC funds. There is a lot. Right. So if you're not an experienced data scientist, machine learning engineer, it's quite hard to find your way. But there are solutions here which enable this real time machine learning at a fraction of the cost that it used to cost just five years ago. Right. And then we touched a little bit on the personalization and maybe I'll skip this one and just go to the numbers of Strava and Duolingo. 100 million users, 1.5 billion valuation for Strava, 50 million users for Duolingo, already a public company with 3.3 billion market cap and quarter billion in revenues last year. So this kind of personalized, relevant social experience makes products better and of course makes the businesses better as well. And you know, I'll skip how to build it. We kind of touched about this, but it's all about understanding where in your product the social experience can actually make the product better itself. Right. And how to bake those. One important aspect is data maturity and that's interesting, Right. A lot of companies say we have the data, but actually it's not very usable. It's maybe scattered across multiple systems. And actually the first step would be to make the data centralized and put it all in one place and also have the right identifiers to be able to work it. It's quite interesting. One of my first projects at McKinsey was with a bank that grew through acquisitions and as many banks do, right. And Basically they had 19 different IT systems and each of them had different identifier and a different way of classifying, identifying different assets. Now, with all the regulation after the financial crisis, they had to report their, what's called risk weighted asset position. Basically say, how much capital do we have on our balance sheet to offset any big financial crisis? Or drop in the value of our assets? And they had 300 people who were, well, I haven't said the name, but who were pretty much doing nothing for 24 days a month. And then in the first six days of the month, when they had the kind of six days to report to the regulator their position, they were working basically 24 hours a day going through millions and millions of Excel rows, which were all of those outputs from all of those 1990 systems to try and match which assets they actually hold, which assets can offset other assets, et cetera. And you know, the interesting that they would miss about 20% of the value. Our business case for this McKinsey project was actually, well, this 20% that you skip just because you don't have enough time to go over, it could be worth 200 million a year for you. So why don't we kind of put it all together and have one system? So centralizing the data is one important step so you can understand what it is. And a lot of companies are in this kind of third phase of dashboards, okay? Especially kind of the newer companies and smaller, younger companies. They have dashboard. They look okay. So my user is clicking on cat videos. Ah, great. So I'm gonna push more cat videos to the feed, right? Or our users are more interested in running, so let's put more running content. But first of all, it's quite manual. And second, it's very asynchronous, right? Maybe users like cat videos now because there was something on YouTube and that's why they're reacting to it. But until you respond, it might be too late. Right? Again, maybe cat videos is not a great example. But you know, this whole dashboard thing is good, right? You definitely want to understand your KPI, you definitely want to track them. But in today's world, it's not enough to actually deliver value to the users, right? And that's where we come in and we help deliver the ROI from the data when it's actually used in the product and in real time, right? So user clicked on cat video now. Great, I'll show them more cat videos. Right? Again, maybe not the best example, but, you know, if you think about jobs, right? User clicked on this job, show them more jobs like that, right? Not even more like that. More like that. And also in line with what they clicked and engaged with before and applied for before. Because then you increase your application rate, you, user is happier, your customer, which is the employer who is posting the job, is happier and you know, it's just a better user experience and you want to do it in real time. If you send them an email, you know, a day later, oh, here are three more jobs, maybe they're already gone, maybe they've been on another website, maybe they, you know, they're not interested anymore. And then the algorithms and here there is a decision step, right? Not everyone need a real time, right? And again we talked about E commerce. Batch processing is fine for that, right? It's not necessarily depends on the ux, of course, but on a regular E commerce website you don't necessarily need a real time system. And then whether you need machine learning or rule based is enough. And here the rule of thumb is if your segmentation based on rules is good enough, so you create segments which are not 100 segments, right? Number of segments is reasonable. So if you look at the news website, you might have people who are interested in business, people who are interested in entrepreneurship, people who are interested in politics. You go to a BBC website, you have all of those tags on top, you can click on them and then you'll get all the articles with that tag, right? Maybe that's good enough, right? But if you go to social media or if you go to Jobs, it becomes much, much broader and it's much, much harder to manage those segments and also make a segmentation that makes sense. And that's where machine learning can come in. It basically creates those segments on the fly, right? You look at the piece of content, you have your profile, boom, I'll put you in the segment. It might be a segment of one, it's fine. It might be a segment of a few users and then you will get whatever fits that segment, right? So whenever the segmentation is not easy to predetermine and also the number of segments starts to scale significantly, that's when you want to start thinking about machine learning. Finally. Well, what is this super linked thing and how it works with the platform? So you have the customer product and you have this whole data that we talked about and this might have the user data. So the user profiles the content on the platform, whether it's Jobs, whether it's boss, whether it's other people, if you're talking about an event app. And then events is actually the behavior of the users. So all of this information needs to be Synced to superlinked. So we become this kind of a user database for you that holds the information about the users, about the items on the platform and about how those two interact historically. Right. Model config is something we discussed about. So this is where your specific labels, your specific logic come in, where you can define features which are relevant for your product. And then the model is being trained. Right. So we start with pre trained model. But of course, as users start interacting with the content, we train the model specifically to the product and then we have the kind of the super linked engine. So the pre trained models are already here. Right. They also feed into this and in the end what you get is an API through which you can show those recommendations. Right. And the type of APIs we have is recommend content, is recommend users. And also who are the similar users on the platform that you might want to make the same offer to. Right. And that one is for instance suitable for E commerce cases. Okay, so this user like this product or bought this product. Great. Who is the similar user for that? So just basically a summary. Right, so it's kind of real time machine learning personalization infrastructure for social plus use cases. We have pre trained models, so you avoid the cold start. So a user comes to your app first time, they already get something somewhat relevant. Right. And of course as they interact and engage with dapp, it will become more and more relevant. It's about one day of developer time to set it up. You just connect to rest APIs and with the current setup you don't need data science, machine learning expertise. But if you have it, we can work with you as well and you can deploy your models based on our infrastructure and you save the whole effort of actually setting up the infrastructure, managing all the pipelines, making sure it works in real time, which we had a team of machine learning and data science engineers working on that for about a year. So that's quite an effort. And then some of the results we can talk about. So for instance, in the social media app, if people scroll through the feed. Well, if people, if we double the scroll depth of the users in the feed in the first week, that contributes to 50% higher retention rates over 30 days, period. Right. So we talked about this kind of the huge churn of 95% of users actually leaving the app after seven days. And here, you know, you give them more relevant content, the scroll further, they find value. Great. Your chances of retaining them are increased by 50%.

That's a major uplift. That's, that's incredible. Retention uplift if, if that can be realized for, for your customers. Speaking of your customers, this was the question I was going to ask which was with respect to your pricing model, how, how do you price? Super linked.

So we price proactive user basically. So if you have a number of users, one of the reasons to do it because we do a lot of pre processing to make sure that it works in real time. So you sync your user information, you sync your content, you sync the interactions and then we pre process and prepare the model to whenever the user comes in so they get immediate response and immediately relevant content. So basically we charge two and a half dollars for a thousand users. But of course with our design partners have kind of custom models, there's a.

Lot of work going into all of this. I was obviously looking through the company before our call and there's quite a, there's a number of people working on Superlink right now. How is this currently financed and are you looking to raise money?

That's a good question. So we raised our precedent round last year from three London based VCs. One of them is RLC, now called Concept Ventures, Episode 1 and MMC Ventures. MMC and Episode 1 are more kind of infrastructure, AI focused funds. So very, very good fit there. And the idea is that we will now prove the use case, get the KPIs right, have those design partners operational, working happy and experiencing the KPI uplift and then we plan to go towards the next round, towards the end of this year probably.

Are you looking to grow your team out as well between now and your next fundraising event?

Absolutely. Actually we were able to build quite a significant pipeline of clients as I mentioned before. Also quite different use cases, quite different verticals. And at the moment what really stops us from serving all of them is just capacity. Right. So we'll definitely look to grow the team. One of the things where we're kind of waiting for is a bit of use case focus. So we can also hire people who are specifically trained and experienced in that particular use case because I think that would be important. But yeah, we'll definitely look to grow the team, especially on the tech side of things and get all of those whole plan that we discussed as well of letting the client do the configuration so we don't need to do it for them. So giving them enough control and an interface to actually do it, giving them an interface for evaluation of the models, ABC testing. If they want, they can deploy multiple models on our infrastructure and then of course kind of strengthening the models themselves. So we'll be Definitely looking to hire, I'd say, well, probably before the round as well, but definitely after the round.

Ben, we're coming up towards the end of our time together and I always like to kind of give this section over to founders. If there's anybody or there's anything that the founder themselves wants to ask of the audience or anybody who's watching this, maybe prospective design partners or, you know, collaboration partners or whomever it is that you're looking for. I kind of just want to hand this time over to you and, and, and see if there's anything that you're looking to ask of the audience.

Absolutely. So we are definitely looking for more design partners, especially in this kind of social fit space, which is where we kind of started, let's say. So that's, that's very interesting for us. So if you are thinking about how to prioritize content and show relevant content to your users in real time, happy to hop on a call. At least explain how we do it right. And if it makes sense, work with you to deliver specifically on a specific API.

So, Ben, thank you so much for coming on the show. It's been a real pleasure having you on learning more about Super Linked. I think what you're doing is very compelling, especially for companies that are looking for that kind of stickiness and the high retention figures that some of the most popular social media applications out there have. And I can't really wait to see what you do next, especially as you build out more design partners and, and look to take on more investment towards the end of this year. So thank you so much for coming on the show.

Thank you so much. Nick, thank you for thoughtful questions and it's been a pleasure.