# How Much Is Your Data Worth?

- Channel: [Vana](https://streameth.org/vana)
- Date: 2025-03-06
- Duration: 27:32
- Watch: https://streameth.org/watch/67c977dddfe4387924220caf

## Transcript

I think this is a panel that everyone wants to listen to because it's answering the question of what is your data worth. So this is going to be, this is my favorite panel of the entire East Denver because the most common question I get asked as one of the co-founders of VARNA is, okay, so how much can I sell my data for? How much is it worth? And every single person on this panel has sold data. Every single person on this panel knows what your data is worth or what our data set is worth and we're going to be exploring that a little bit today. A couple of DLPs are represented here and then a couple of folks who have joined the VARNA team because of their expertise in data sales and data are on the panel as well. So we're going to go talk, we're going to talk through that today and hopefully answer the question for you all about how much is your data worth. I think we'll start off by talking about each of our histories and what we used to do or what we currently do such that we are capable to speak about this topic of how much is data worth. So for those who don't know, I used to work at Appen with this guy over here. Colin, maybe you should introduce Appen and then we can talk about what each of us did. Yeah, for sure. Appen is a big data broker. So they work with, they have a huge crowd of like 8 billion people who all either label or add data to their platform and then they sell it for lots and lots of money. I was head of crowd there so I was in charge of making sure we had the right people to fill the projects and that they were happy enough and were providing good quality data. And I'm chief of staff at Avana. Awesome. And so I was head of impact sourcing at Appen and my job was to try to find the most difficult data sets that were the most expensive because the tech companies really wanted them to fill gaps. Things like niche language data, niche image data. I became very known for being able to just pull data out of thin air, to be honest, and for a lot of money because I'm a big platform. So that's how we know data. Maybe I'll go to Darren and then I'll go to Roger. Darren, how do you know data? I've been in a startup ecosystem for like 15 years. I've done a bunch of different stuff over those years, including a social network out of Boston. And in that, for example, we would sell the information to Nike and Under Armour would be our customers in those instances. There are disconnects in the markets in a lot of data. So there's a lot of companies that sell to a middleman and has no real comprehension where that product gets shipped to. And you see this problem space over and over and that disconnect of understanding because they don't sell the primary product creates opportunities for data sales. And aggregation. I also run some data labs and have tried to source information. Funny story is watermelons look a lot like safety helmets. And when somebody brings a watermelon onto a building site, it blew up my model. I had to go get pictures of workmen carrying watermelons. Right. And that's a strange data set to need. But that was a highly valuable one. And it was it was more difficult to get than you would think, because when you ask someone to carry a watermelon so I can take a picture, they're like, where's this going? You know what I mean? But the reality is that it was still a highly valuable data set, as well as the larger data sets that which are sort of locked up that we couldn't get access to. I know, like when we were doing the insurance company, there was data that we knew existed, but we just weren't allowed access to it. So I think these walled gardens are also something we need to consider when we think about value. And then I'll go to Ryan. Talk about your past life and your current life in relation to data. Sure. So and great to be with everybody today. So started my career at LinkedIn and I come to the data business through the advertising business where the LinkedIn profile data, your job title, seniority level, the company's size, education, skills and certifications are used for a couple of different business lines. One of them is through targeted B2B advertising. So if you're an IT decision maker, someone wants to sell you cybersecurity tech, they would target you with advertising on and off of LinkedIn, also for recruiting purposes and also for the sales navigator product. So your profile data has a lot of commercial value for staff and advertising. Then I moved over to data partnerships at Meta or Facebook, Instagram. And this is where it really kind of came to life for me. So one way to think about Meta's business is they are very afraid of Apple and Google. Apple can change the Apple iOS policies, which can result in not understanding how a user is interacting with other Apple apps. So it's very difficult to personalize advertising to that user. Mobile gaming is like a huge advertising category on Facebook and Google. So people want to know what games you're interacting with or did you shop in a shopping cart and like not transact. And then Google has been threatening to deprecate web cookies. So think of this as like cutting off the oil pipeline to Facebook to allow for personalization of advertising at scale. So you could think of the Oculus or the virtual reality as like, yes, it's the metaverse. It's all about virtual reality. It's a future new computing platform. It's also something that's like not owned by the Chrome browser, Google, and not owned by the iPhone, Apple. So it's a way for Facebook to have the data supply chain kind of controlled by them. And then, you know, from that, I'm now, you know, went to Coinbase, went to Polygon, was very interested because people hated Facebook so much and could you use tokens as a way to control governance over the apps and services we use and also have economics in them. So now I'm building a data liquidity pool, DLP, on VANA, with VANA for vehicle telemetry data and very interested in the exercises of understanding what is the annual subscription value for that data set? How can you enrich that to command a premium? What's the difference between maybe a smart city or Department of Transportation's needs and use cases and what they would pay for it versus a battery company that wants to build software to make batteries more performant for like the Tesla battery versus a company that wants to do battery insurance? So these are all different use cases. But I think this is going to be a really big commercial opportunity for our business application of tokens to create these data to service businesses data devs. Amazing. Thank you, Ryan. A lot of expertise in data sales, as you can tell. Roger, do you want to talk about your data? Hey, everyone. I'm Roger, the co-founder of Diffusion. I come from the data side from a forensic background. Very early on, I helped purchase Mint.com for Intuit, which was a nine-month-old company, sold for $180 million, and that was basically because they're getting a lot of financial data on consumers, right? And so zero revenue, $180 million thing. That was like 2008, 2009. I spent the majority of my career in FinTech in China. I was both a data buyer for credit data or individual data, and we were actually aggregating. We're a FinTech lending platform, so again, we were aggregating a lot of data and selling it to different types of credit bureaus or unofficial credit bureaus in China, essentially. And then I got involved in, post that, fast forward a little bit, got involved in crypto asset recovery investigations. And then through this, it was really kind of like, what is the value of information? If you're really thinking of the value of information, you know, one link, we handled over $100 million worth of recovery cases, and we've seen a lot of people being socially engineered into kind of clicking a wrong link or depositing coins or fiat into a fake exchange. And then again, what's the value of data? What's the value of information? And so that became the genesis for starting Diffusion, which is all about creating a community-driven source of truth and aggregating validated, high-quality language data, otherwise known as knowledge, basically. And so when I met Art, and talking about aggregating private chat messages and things like that, that was very synergistic. And again, it's like seeing what is the value of understanding truth, value of the truth, and kind of uncovering what could be a fraud or what cannot be a fraud. So we came in from that angle. And then, so a bit new to kind of the pure data broker side of the business, essentially. Yeah, so we've got two data devs, and we've got three folks who work at Banner now, who used to work in data in some way. Folks, like, I think we understand this concept of the difference between private and public data, and why private data is really valuable. But I think that's also another common question that I get from folks, of like, why is private data actually valuable? What's your take on why private data is valuable, and what's your take on public data? I'm going to actually start with you, Ryan. Sure. So as it was, you know, let's take a step back. So how do you get, you know, how do you get to vehicle telemetry data from the Tesla API when I think about private data? If you go to a service, let's say Whoop. I don't know if anyone here has a Whoop device. So I am a weird person. I go to the terms of service, the terms and conditions. I'm not a lawyer. But I go to the terms of service for different hardware devices that collect commercially valuable data. And then I look, in their terms and conditions, do they explicitly prohibit you from doing an export? Although we can talk about whether they're even allowed to do that. So that they can control the data, so they can commercialize the data. And within the context of Whoop, I had a Whoop, and I actually stopped wearing it as a result of this. Their terms and conditions indicated that you could not, you might be able to bring it to another data controller, but you can't commercialize it on your own behalf, which suggests that there is a commercial benefit to them for doing so. Another interesting example is 23andMe. You're not really supposed to export your 23andMe, sold, I believe, a couple hundred million dollars of data to a pharmaceutical company. But you're not able to export your own 23andMe to, or at least within their terms and conditions, to commercialize it on your own behalf. So it suggests that these companies do see their data set as a competitive advantage. And so being able to free your data and then commercialize it on your behalf is sometimes explicitly prohibited within their terms and conditions. And that's why I think we're seeing a lot of success within the deep-end category within crypto, where you're just owning the hardware end-to-end. You don't have to worry about some sort of third party having issues with them. But I think that's to say, like, yeah, the private data is super important. And if you're in conflict with the company's terms and conditions for even enabling a user to export their own data, like, that can be really scary for a startup. So there's a lot of commercial value there, for sure. Yeah, I think let's have that debate at some point in time over the legality of those terms. But I will say... I'll give legal advice. I also cannot give legal advice anymore. But I think that what is clear there is that the companies realize the value of the private data. Otherwise, they wouldn't put these moats up. They wouldn't put these walls around their data. Darren, how does that relate to public data? Well, if you think about public data, it might be in the public domain, but it may not be accessible or usable. And if you think about the value exchange of this public information, the previous approach of having, like, a long-tail Google has sort of been eaten away by a lot of these foundational models who've incorporated a bunch of that knowledge in. And that sort of value proposition of public data being able to be monetized through advertising and so forth, that opportunity is just literally going away, as that information is no longer being sent to that website that may be the best website in the world on guitars with an Amazon affiliate link. Those opportunities are just being eaten up, and that information is being sourced directly into, like, an AI model and being pulled up to you. And in doing so, it actually breaks the economics of the public system. So what can we do about changing those economic systems? Well, one of the biggest things is about changing how this value flows with public information, because that information still is created by someone or something, but that value is accrued only into the publishing platform. You know, you think of any social media, any sort of activity that you're generating data for, which is just through usage of their platform or product, then that should potentially come back to you as well. And I think enabling that is going to be more and more important as a society as we move forward with more and more automation. So how do we change the value flows and you start making money from the data that you're already going to generate? So I think that's the thing that we really need to think about when we get to public data. So, Roger, you mentioned that DTGN collects telegram or language data, right? And you mentioned that your goal there is to understand like a source of truth about how we think about the world. Why can't you get that from the public? Why can't you just get that from the public internet? What's the difference? Yeah, 100%. So basically like, if you want to get that from the public, what I tell people is basically if you care about privacy and everything and you really care about the value of your data and things like that, like don't use Google. Don't use Facebook. Don't use telegram, et cetera, because then you're hypocritical because you're going to lose a lot of... These guys are selling your data for hundreds of millions of dollars and then you're not basically actually participating in that type of economy. And furthermore, if you want to actually get value, you have to, as part of the data model, aggregate all this, aggregate it into a large repository that is actually more value as a group, essentially. So on that side, like, yeah. Just on that, like when I was previously doing data sales, the aggregation of data is why it becomes valuable, right? Because at that point where we've grouped that information together into a larger data set, we can actually get signals from that grouped data. So, you know, we need platforms like this in order to help the monetization. We need the data guys to sort of centralize us putting our information in so then we can help that group monetization. I just wanted to highlight the importance of the aggregation. Yeah, I'm sorry. Just to add is like you as an individual don't really have a say in the price of your data and how it's being sold. So part of this is the promise of a data DAO is the governance model where, you know, as a group, as a token holder of this DAO, you can negotiate as a group together with the administrator of the DAO to say, okay, let's sell this telegram data or WhatsApp data or telemetry data to a large corporation at an X amount of price. So I think that's something quite game changing too. So Colin, I'm going to ask you a very targeted question about our roles at Appian. So like, you know, we bought and sold data for large tech platforms like Think, Face, and a book. Why were they interested in buying data? Like don't they have data? Yeah, but my understanding is that they couldn't train without the user saying they could on their private messages. So we actually had, I believe, we actually had Facebook buying Facebook data from Appian. Because they really wanted that private data. And to the question of like, why, if you're just trying to build, if you just need language, why are you going for private messages as opposed to something that's public? Well, that data is totally different, right? Like the way that I text my wife is not public anywhere, you know, and it's really valuable data that can't be scraped. That isn't, even if you have the app, you're not training on that data unless the user gives the provision, right? Yeah. And I think that that was actually, that was one of my projects that I worked on. So the large tech platforms, because they wanted, so the big difference is you think about the public internet as all of the words that exist if you do a Google search, right? And as Colin said, those words are very different from how I would communicate with you in this forum, how you would text to each other and text to your loved ones. And the source of human conversation isn't necessarily the billboard that exists on the public internet. And so oftentimes they would do these projects with large tech platforms that would try to buy their own user data and get the consent from the user through that purchase because they were unable to just train because of the regulatory restrictions. They themselves saw enough value in private data that they would buy their own data just as a way to get consent. And that was a wild discovery to me moving into this space as well when I was at Appen. That's a really nice segue into the next question of what is the wildest or strangest data set that you've sold or that you've seen being sold on the panel? And actually, Darren, you kicked off on this one with the builders and their watermelons. Is there any other data set that you've seen being sold or you've sold yourself? There was actually a data set which we purchased but then couldn't use, which is an interesting story. So it was for Maxwell MRI. We were doing computer models on prostate cancer. And we actually got a bunch of MRI data but then we legally couldn't use it because we couldn't prove who they all came from. And it actually takes a very long time to collect this sort of data because you have to partner with a hospital. So we had the information. We literally had the data set but we weren't allowed to use it. And we couldn't reach back and get easy permissions and we'd have to go from the start, which would then take four to five years from the point of when people are coming in to get their scans to sort of opt in. So that one hurts, if you know what I mean, because we had the information that we couldn't use. So I think if you move outside the crypto-crypto aspect, there's a lot of value in the audibility of the individuals agreeing to stuff, the DAO structures, the fact that there's a license that you've agreed to, that you've got rights and that you've got a wallet that you have got access of. And so that whole supply chain of sort of verifiability is massive. The other interesting one, I got a bunch of game designers to create me silly building site issues because you need negative pictures as well for sort of computer vision models. So I had like basically GTA V in a building site of nonsense just for AI models. And we had like a bunch of game engineers creating those synthetic images because those images did not exist in reality. Because when people are doing silly things on the building sites, it only ends up on TikTok, unfortunately. So it's hard to source that information. Now you can get it off TikTok because they either contribute it to the TikTok beta now. No, it wasn't available then. Yeah, now you can. Ryan, what's the strangest data cell that you've seen? Yeah, I mean, I think this goes back to what got me interested in Ivana earlier. But I was partnering with Reddit when I was working at Polygon. And when I saw the $250 or $300 million licensing deal as a Reddit user to Google Bard and then OpenAI. And then I think it was within Reddit's S1, in addition to advertising, they had data cells as a new business line, which is something I imagine Alexis Ohanian and the founders would have never imagined. That was like a potential business opportunity for Reddit. And what that signaled to me, and then I'm in conversations with a company called CulpData, which actually can underwrite a loan against the data set. So they look at a lot of companies that have interesting data sets. And then they will actually be a data broker and sell that data set on behalf of the company and take a cut of the revenues. And they can actually underwrite a loan against that data set. And what was interesting to me is if Reddit, if this is now like a business line within Reddit, in addition to advertising, how many companies with hardware sensors, how many companies similar to what you're talking about with like financial data would like add even revenues but have something valuable, could have a new business line of data as a service. And I think that's going to be this emerging category within like WebT or just like traditional businesses. And I think we have this interesting opportunity for data devs to participate. And that's, to me, that was kind of like an aha moment, like, oh, well, I think there's going to be a lot more of this. Another thing, and this is not as cool as your watermelon example. I was like trying to, like, oh, do I have an example as cool as this watermelon construction? I don't have one as cool as that. I just think it's wild, LiveRamp's business model. LiveRamp is a company that does an identity resolution to power personalized advertising scale across lots of different platforms. And also a similar business model to customer data platforms. But essentially, like, they're resolving in a way that's adhering to GDPR and CCPA, your identity on the Internet across a lot of different platforms, your home address, and all the different flavors of advertising and targeting that could be used that they connect and like pour into this digital identity that you don't even realize is being built about you. And it's just wild, the level of sophistication that goes into that and how, like, just wrapping my head around, like, our partnership with them was just, like, super, super interesting. So that was probably the craziest thing, but not as cool as the watermelon. Roger, do you have a crazy data set? Well, maybe one very basic is, I guess, diffusion, the way it works is a data-on-demand driven platform and folks can create subnets to attract the data they want. And one guy, one entertainment company wants to create a data set on weekend plans globally. So what people do on their weekends in different parts of the world. Colin, do you remember any crazy data sets we did at Appen? I remember a lot of, like, we need 10,000 people to react to this thing we're going to put on screen and film yourself and then upload yourself reacting. So, you know, very strange data sets. Although I will say the one I've heard that might be coming as a data DAO is a feed pick data DAO. Love that. That is a pretty crazy one. I don't think it exists yet, but I have, you know. Look, diffusion models come through feed picks. Yeah, right, exactly. So be on the lookout for the feed pick stuff. The craziest one was Appen was helping a company train, like, a model that you could take a picture and then it would estimate your BMI. And the training data set for that was, like, a bunch of people wearing a nude suit walking into this device and, like, spinning around and then coming back out. And, like, we had to get it in a lot of different categories. So you needed to get it in different ethnicity categories as well as age categories. And so, like, we had a real problem. I did the problem sourcing, right? So I had to get, like, try to find 60 to 75-year-old men who would wear a nude suit and walk through this thing and then come back out. I was like, where am I going to find these people? So I had to go to, like, our local, like, veterans club and be like, hey, guys, does anybody want to come do this? It was the most wild experiment. But these are the crazy, wacky data sets that exist out there that could be satisfied. I don't know if that one could be satisfied by data now, but who knows? Maybe, if it was worth enough. Speaking of this, I'll just go with one last question for everyone in the group, which is the question that I've asked everyone at East Denver, right? And it's, what is your dream data set and what would you use it for? And you can't use the DLP data sets that you're actually working on. So what is your dream data set and what do you want to use it for? Darren, you start. Well, as someone who does like to be a bit of a degenerate on the charts, I would probably like a data die of all of the VC deals and unlock periods for obvious reasons. Very cool. Yes, Colin? I think my answer is kind of boring and it's actually one that's being built on Bono right now, which is Amazon. And I think the reason I'm saying that, I know that's not the craziest answer, but I vampire attacking data walls. And I'm like, this is the best data wall to take data from. Like, I want to see Amazon get really mad. And I think this is a good way to do that. I don't. Well, yeah, I'm not in charge of the problem. I actually do want to see Amazon get really mad. Someone related to our DLP. So we have a group of crypto friends on biohacking. So we like the biohacking bros kind of group chat. So when you get a data set on all the different types of biohacking tips and verifiability on which one of them, because some of them are like really weird suggestions and we don't know if they're good for you. You should definitely talk to... Where is he? Mikhail? From the biohacker data. He's probably going to help you with that data set. Oh, yeah. There's a data dog. Yeah. You should definitely get that one. Yeah. It's a really good... And there's a sleep one as well. Ryan? Yeah, I'm going to... I have to plus one with the biohacking. So when I was narrowing in on the specific data product that we're building at DLP Labs, I was speaking to a data scientist at a German pharmaceutical company called Behringer Ingelheim. And they purchased billions of dollars of clinical trial data for the purposes of drug development. So the question is, could you pair 23andMe, like an export, like basically be Vampire Attack 23andMe, or ideas, it's fantastic, or you create some sort of variation of whoop, or you just Vampire Attack whoop. And you have that data and you pair that with, let's say, a clinical trial around questions around your interaction with the health system and around diabetes, which is very popular. There's a lot of drug development around diabetes. So many people have it and struggle with that. And that would be extremely commercially valuable. And the revenues from that could be used to support a Dow. So I think the health category is so interesting to me because pharmaceutical companies pay a lot of money. And it's very difficult to get clinical trial data plus basically to inform drug development, anything informing drug development. It's definitely outside of my area of expertise, but seems to be a pretty significant area where data is purchased. So that's some inspiration for some data nows out there. These guys know data. They know what data is valuable. They've built things with it. So if you're interested in setting up a data Dow, I recommend talking to one of these guys on the panel. Thank you so much. I really appreciate your expertise here. There aren't very many of us in the crypto space that have done data sales and understand how people's data actually gets bought and sold. So I'd like to say that you guys are probably the best. You're the best in my eyes. Thank you.