- Duration: 25 min
- Platform: zoom
- Source: Read.ai 01JPQNWR1NWGNQQ23KW88DGQRH (https://app.read.ai/analytics/meetings/01JPQNWR1NWGNQQ23KW88DGQRH)
Transcript
[UNKNOWN_SPEAKER]: Hey guys. Danish: So nice to meet you. [Maciek]: Um, on our side, I invited Danish, uh, who’s our CTO as well to join, uh, in case we talk a little bit more in more detail about the technical stuff. Yeah, sounds good. I invited Blake. [Michael Houck]: He leads partnerships for us. Um, so, uh, figured it’d be good to have him. Here too. [UNKNOWN_SPEAKER]: Cool. [Maciek]: So maybe to just get right into it, perhaps it would be helpful to frame the conversation. I understand that when I briefly described to you what our tech has been doing, this clearly caught your interest in some capacity. So maybe to frame the conversation, could you talk about in what context you find it useful or interesting? And maybe we’ll start from there. [Michael Houck]: Yeah, I mean, look, we have this creator network, right? And we sort of help people grow on social media by connecting them with these creators. And the creators can charge for that engagement, that access, those partnerships. So anything that’s going to increase sort of the fidelity of data that we have about the audiences of those creators will help us advise them on how to price themselves more effectively and also maybe allow us to do some dynamic pricing things on our side. So yeah, when you mentioned that we would be able to sort of understand who’s in that audience, we could then surface that in relevant ways within our app to our users and be like, oh, you want someone with influence in this sort of niche. This person is a better value for you than they would be for another user because they have influence in the specific niche looking for. So that was the first way. I’m sure there’s other ways too, but that was what initially caught my attention. [Maciek]: So would that be the, like, let’s say intention of like doing some dynamic pricing and just trying to understand like what, if there’s like a specific problem that you guys would want to solve with this or like a specific goal that you would hope to achieve with that so that we can try to understand if like this is feasible. [Michael Houck]: Yeah. I mean, I think the goal would be increasing the amount of budget and amount of posts that are sent through our system, right? We want to increase throughput if better insights around who somebody is going to be able to get in front of either allows us to increase pricing or just increase the amount of posts that are being sent and matches that are being made with creators. Both of those are outcomes that we want to drive. So in like an exploratory way, yeah, it’s interesting. [Maciek]: Cool. So is there something in particular that you would want to know about the tech we developed or I’m just trying to think like how we can structure the conversation so that it’s the most useful to all of us you know so we tap onto the topics that are the most important here. [Michael Houck]: Yeah I mean from my perspective just want to learn more about it so it’s going to be useful to me either way so if there are specific things that you want to know that would make it more useful just ask me and I’m happy to answer. [Maciek]: So, Danish, can you maybe give a brief overview of how the tech works? Danish: So, fundamentally, what we try to do is we try to take various accounts scattered all over the internet and try to put them into one graph so that we can assess how well-versed people are in a particular topic or how influential they might be, how their careers might be developing, what kind of companies they have connections with, what kind of people they have potentially connected to through their professional network. So this gives us a big picture understanding of a person’s value in their network and across different networks. And then we can do all sorts of different operations on it. We can search people that might belong to similar groups, like communities, essentially. We can do searches for hyper-specific professionals. So the kind of queries that, for example, would be really, really hard to do on a system like LinkedIn or Google, like who are the most influential investors in a particular space, let’s say. All of those kinds of things we can do very, very well. The way it works internally is a mixture of network science. So there’s a lot of really complexity analysis and building those networks in a computationally feasible manner. And on top of that, there is a lot of machine learning to sometimes solve the sub parts of the problem. So for example, how do we connect different accounts that we find scattered all around the internet, but also some big picture elements like how do we then assess overall prestige of a person or their overall career trajectory. [Michael Houck]: Got it. So it’s, I guess, what do you then, what’s consumable by us if we’re if we’re gonna be a user, if we’re gonna be a client in some way, what kind of data points are consumable? Danish: Frankly, for us, it’s very easy to give you way more data than you will ever really need because we have all sorts of different problems that we have solved internally that we have not even really exposed yet, right? So if we collaborate, what it will come down to is that we will start with some set of data and ask you, hey, try something with this. If it works, great. If it doesn’t work, here are other things that we can also give you, right? So it will be probably discovering with you what really works for your use case, right? [Maciek]: So something that basically what I originally mentioned in the email was that if my understanding is that your creators are, and correct me if I understand this right or not, your creators would expect their followers from Twitter and LinkedIn and upload them? Or do you just validate based on the number of followers they have and some declaration from their side in terms of what they post about? To what degree are you validating that audience? Are you pulling in that information? [Michael Houck]: Yeah, it’s the latter. We don’t have them upload CSVs of their followers or anything like that. We have a scraping bot that just confirms their followers. Account. And then we also have them, as you said, declare sort of what niches and topics they talk about. And we use that for matching with the right users and the right posts. So it’s not at the level of sophistication where we know who’s in their audience or can derive sort of trends and influence from that. [Maciek]: Twitter is really easy to export it, meaning that the creator could just go to their Twitter account, export archive and then upload the archive, you know, through a simple upload. And what could happen there is that you could then call our API. We could give you very detailed, precise data on who actually follows the person. Actually, even more than that, because in the Twitter archive, you also have interactions. So we could tell you who’s actually interacting with their posts and with incredible granularity. Right. So not only like which communities these people belong to, but also like what roles they hold in which companies, with geographies and so on. Right. Really high level analytics. Danish: And not just for LinkedIn and Twitter. So if they have like a email newsletter that a lot of people are subscribed to, we can ingest that data. If they have a lot of communication with industry influencers through their emails, we can ingest that data as well and give you some estimate of how valuable their audience in general is. [UNKNOWN_SPEAKER]: Interesting. [Michael Houck]: Okay. So basically from, I’m thinking about it, the UX on our perspective that they would actually go through. So when they would sign up, we could prompt them to say, oh, here, you know, download your Twitter followers by doing this XYZ steps or, you know, your email list or your LinkedIn audience or something. They need to do steps to sort of connect those. Danish: Yeah. So, or you can just tell them that, hey, go sign up with graph.one and we’ll just get that information directly from you. From graph.one. So we have already solved those user experience problems of connecting many different platforms. So I don’t think if we are just trying to move fast, it might not even make sense to try to implement a native solution. You can just ask them to start it and then using our API, you can just get access to whatever you need. Okay, cool. That makes sense. [Michael Houck]: But I guess like just thinking through the kind of the second So we’d get that data somehow, one way or the other. And then from that, you guys would be able to deliver to us information about that audience that we could then use to say, oh, here’s the range we think you can price in. Or if we built some dynamic pricing system on top of it, we could show prices sort of effectively to our users or those specific creators. That’s interesting. Yeah, I guess anything. [Maciek]: I’m realizing now that you could, although this would require, this would be a little more complex, but because in those archives, there are also, there’s also information about like who actually clicked like, or retweeted this post, right? Sure. So you could also, so when I tried your platform as a user, as a customer, right? I was somewhat frustrated with the fact that I got some interactions and I couldn’t really like the attribution wasn’t really there, you know, so it was like sort of like guesswork of like what actually came from the platform, what was organic, right? Or what was like from otherwise from Twitter. You could under the system, you could actually tell the user, well, look, we got like guarantee, guarantee to you that this click came because you paid for it. But this person has followed this account before or hasn’t engaged with their posts before. Therefore, there’s a high likelihood that this actually came from our platform, right? So you could actually do potentially attribution. That would require, you know, it would be more tricky because either this person would have to re-upload the archive or we would have to purchase the Twitter API from Twitter to do this on a regular basis. But in principle, you could do it, which obviously unlocks a whole new level of a product. [Michael Houck]: Yeah, I think the that we don’t do that today isn’t, you know, a technical reason or anything like that. It’s just because the creators on our platform, our kind of commitment to them is anonymity on the platform itself because, you know, they’re doing paid engagement and that’s, you know, at the best in a gray area in terms of service for these platforms. So we don’t want to get anyone in trouble and we literally never have due to our approach. So I think maintaining that’s pretty important. We probably wouldn’t go down that road unless there was a way to do it without you know, surfacing who those creators were, but it seems unlikely that there wouldn’t be. [UNKNOWN_SPEAKER]: Hmm. [Maciek]: Yeah. Maybe there will be, but it’s getting like, I think too complex for what’s the ROI on this. Yeah. Yeah, exactly. [UNKNOWN_SPEAKER]: Okay. [Michael Houck]: So what would be the, what would be the next step if we wanted to, if we want to move forward and tell us something, could I see maybe what it looks like with my audience and then, we can have some intuition on our side about what we could build on top of that? [Maciek]: Yeah, you could sign up on graph one. And so this, this tool where we have this, all of the infrastructure that Danish mentioned, it’s not really geared towards specifically, you know, like audience analysis, or at least it doesn’t show you this in the interface, but it will. Danish: Yeah, but we have tools internally to handle that. We can, if you have any kind of exports of how you are interacting with, if you want any kind of reports on how you are interacting with people and whatnot, we can just generate those and send a CSV at the moment, right? This tool doesn’t have a reporting dashboard of some kind. [Maciek]: Yeah, so I think the easiest would be if you sign up on GraphOne and upload your Twitter archive there, and maybe also like LinkedIn archives, like we have your whole network, right? So we can play around with that. And then there is a way the website that we previously built, Hive One, the communities are not showing up there anymore because we haven’t been really maintaining this, right? But this was the website we originally started with, where we published around 5,000 Twitter communities. We stopped updating this when we stopped accessing the Twitter API, but we still have that historical data, right? So we have a few years of Twitter historical data for hundreds of millions of accounts, right? And those communities. It wouldn’t be super up to date, right? We still have this community information, right? So if you have an audience of a few thousand people, we can tell you like which of these people are influential in specific communities and how influential they are in those communities, right? So that’s where this gets interesting in the sense that let’s say someone wants to get endorsement from people that have audience in let’s say crypto, right? Or I know like machine learning communities, right? Some positive community or whatever else, you could actually give them targeting where you say, okay, this person, we know that this account is actually influential, has followers in those communities. [UNKNOWN_SPEAKER]: Cool. Okay. [Michael Houck]: I’ll sign up for both. I’ll check out the older one, Hive one as well, but I’ll check out the… [Maciek]: It doesn’t work anymore, but what you can do is that you could go to Internet Archive and just see a version from a few months ago, and then you can see communities. We just haven’t invested any resources to maintaining So it’s just like, um, the data doesn’t show. [UNKNOWN_SPEAKER]: Okay. [Michael Houck]: Makes sense. Um, cool. Well, look, I think this could add a pretty interesting layer onto what we’re doing. So, um, let me play around with it and I’ll follow up with you guys. Uh, we can chat over email and, um, then depending on next steps, make sense to hop back on another call and, uh, and, and formalize something. [Maciek]: That’s good. All right. Sounds great. [UNKNOWN_SPEAKER]: Take care. Thanks.