Section Insights
Introduction to Data Brokering
What is the background of the speaker in the data brokering space?
The speaker, co-founder of Polyhares, discusses how they started their business during the data licensing process and achieved a million dollars in revenue in their first month. They emphasize the rapid growth of the data industry and hint at the insights they will share regarding the industry's needs and future direction.
- The data brokering industry is experiencing rapid growth.
- Polyhares achieved significant revenue quickly, indicating a lucrative market.
- Understanding the industry's demands is crucial for success.
- There are insights about the industry's future that are not widely known.
Commission Structure for Data Leads
How can individuals benefit from referring data leads?
Individuals can refer leads to Ryan and receive a 6% commission on closed deals. The process is straightforward, and Ryan encourages brokers to add value to the data presentation to maximize the deal's potential.
- Referring leads can be a profitable venture with a 6% commission.
- Brokers should aim to enhance the value of the data they present.
- Understanding the data presentation process is key to achieving better valuations.
- The average deal size is around $300,000, leading to significant commissions.
Data Requirements for AI Labs
What type of data are AI labs looking to acquire?
AI labs typically seek comprehensive data sets rather than specific types of data. They may require full corpuses of data for training models, indicating a preference for rich, contextual information.
- AI labs need full corpuses of data for effective model training.
- Specific data types like Slack or Teams are less relevant than the overall context.
- Understanding the data needs of AI labs can guide data providers.
- The demand for diverse data sources is increasing in the AI industry.
Deidentification Process of Data
What is the process for deidentifying data before selling it?
The deidentification process involves using third-party services that apply algorithms to remove personal identifying information while retaining the relational context of the data. This is crucial for both safety and maintaining the data's value.
- Deidentification is essential for protecting personal information and data value.
- Third-party services are commonly used for effective deidentification.
- Improper deidentification can lead to significant data value loss.
- Liability issues arise if deidentification is not handled correctly.
Exclusivity Terms in Data Sales
What are the typical exclusivity terms when selling data?
Exclusivity terms for data sales usually last around 24 months, allowing buyers to monetize the data without competition. Longer exclusivity can lead to higher payments, but sellers should avoid perpetual agreements to retain future value.
- Exclusivity terms typically last 24 months to protect buyers' interests.
- Longer exclusivity can increase the selling price of data.
- Sellers should be cautious about entering perpetual agreements.
- Data sales can impact the overall valuation of a business during due diligence.
Transcript
0:00 Ryan Lock. So, there's a lot of hype online right now about people making money brokering data or selling data to the AI labs. As I understand, you're someone who's been in the space for quite a while and you've made a lot of money doing it. So, first of all, why don't you tell us who you are and like what is your track record in this space in particular? >> Yeah, so Polyhares got started in around June or July when my co-founder and I were going through the data licensing process on our own businesses. and then really just took off. You know, this space is growing really quickly and so in our first month, you know, we tracked a million dollars in revenue.
0:38 and it's just been really growing from there. Insane. So, a million in your first month. And we'll get into the actual business model that you guys are following. I think there's a lot of questions that I have personally. I think there's a lot that the audience wants to know about whole how this whole world works. But, you know, give us maybe one or two takeaways that if somebody listens to the whole thing, what are they actually going to learn?
1:01 >> What they're going to learn is a lot of what the data industry is looking for from, you know, representatives that are looking to make money. You know, there's a lot more than just making intros to be really good in this industry and to make the most money. and then I'd also just say, you know, where the industry is going and some key things that I might know that aren't really public knowledge that could be of use.
1:26 Some I'm super curious about this because we see companies like yours and like, you know, the micro ones of the world just seemingly come out of nowhere and do I mean, you you guys did a million or, you know, seven figures in revenue in your first month. I mean, that's a huge feat even for somebody that has a track record like you. So why don't we start by can you just give us the like the elevator pitch of this business model like how does it actually work?
1:52 >> Yeah. So these frontier AI labs they need data to improve the models. They've used all of the data on the internet. They've bought as many rare books as they can buy. And so they need private company data to continue their improvements. And so what we do is we help with securing the rights to those data sets through acquiring them from business owners. We do the deidentification work to to remove any personal identifying information and really just get the package of data training ready so that the Frontier Labs can come buy it from us and get it immediately on to work with it for their researchers and save their researchers valuable time. And so we're really like a procurement partner for those labs.
2:36 >> Got it. Now, like how much money are we talking here per deal? cuz I've I've seen some crazy numbers being thrown around. I've even heard of people who have gotten pitched these crazy numbers and they're like, "Look, this seems too good to be true. Like, what? You're just going to hand me, you know, 300K or 500K or a million dollars just for my company emails or Slack? I mean, that almost seems too good to be true." So, I guess the first question is like what is the average deal size that you guys are seeing on on your end?
3:04 >> Average would be around 300,000. we we've done, you know, seven figures in just in a single transaction, so that's definitely out there. and then, yeah, I mean, of course, the Spirit Airlines deal, went for $10 million to Google, you know, and that was for a company that's not even in business anymore. So, there's a lot of, you know, upper market potential for this as well. Now, can you kind of run me through all the like because again I think on Twitter right now again there's a lot of hype. People are saying like, "Oh, hey, just go find a company that's been in business for 20 years that has 30 plus employees and then you can just license their data and turn around and broker it to an AI lab." Like obviously it's not that simple. So, can you give us the actual like what are the almost like the line item of things that you guys do to transact and to actually close a deal because I think it's way more in-depth than people are making it out to be online.
4:02 >> Yeah, let's see. I have a good slide for this. Okay, so just making the intro is only step one. the second piece of this is the real technical due diligence side of acquiring data. just a big stack of files is not what's interesting to the AI labs. If you know you just have if you had a wound down company but you don't have all of the context from their full business operations. So if you're missing their Slack for example, that's a very important part of the process. What the AAB labs want is to be able to see how every employee operates inside of a company and the different types of tasks that they do and being able to watch a workflow from maybe a client email all the way through to how an internal team talks about the email, fixes some resolution, what solutions they propose, which ones do they not go with versus which ones that they do.
4:58 sometimes that's even more interesting than than the final answer, right? and then how does that get all the way back to the client? So that full trajectory is what's interesting and that that can happen across multiple different systems. And so when we're doing these technical due diligence calls, a lot of what we're trying to understand is how the operations of a business work. you know, for my own company, I wish I had known this, but we mainly operated on just cell phone. Just pick up the phone and call the homeowner. Just pick up the phone and call the contractor or call the sales rep. And a lot of that data wasn't in like recorded anywhere. and so the actual workflow the AI could view just looked like ticket created, ticket closed.
5:41 >> Oh, I see. Yeah. >> And then my business partner, for example, he had, you know, all of his subcontractors even on Slack and you know, so I couldn't even get subcontractors to send me invoices on time and let alone that. And so he had full richness of data as everyone remote only was communicating on Slack. And so that was really interesting for the labs. >> Got it. So it seems like they want to see kind of like you said, they want to see the workflow end to end. And the more context they can get around that workflow, the better. So, it's like, you know, Slack messages alone may only be worth so much or maybe nothing at all, but if you can get the Slack messages and the CRM entries and the tickets that have been created along the way and the emails that have been sent internally along the way, like that's kind of the full package to to kind of oversimplify like is that fair to say?
6:34 >> Yeah. Yeah, that's correct. >> So, one thing I want to say, Ryan, so obviously like there's so much that goes on under the hood. It's not as simple as people are making it on Twitter where it's like, "Hey, find a company that wants to sell their data, connect them to an AI lab, and cash a six-f figure check." Like, there's so much that you guys are doing around like anonymizing the data, which I want to get more into here shortly, packaging it correctly, really even understanding what you guys have access to. So, one thing I want to say for the audience, because there seems to be a lot of people that want to get into the space as it's getting more popular, if people want to like refer leads to you, Ryan, like if somebody's like, "Hey, I've got a company that's interested in selling their data. I don't know what I'm doing, but I'd rather just hand the lead to Ryan, and let him take it to the finish line." I think they can get what is it like 6% commission if they use the link that we're going to put in the description and in the show notes. if you want to just send Ryan your deals, let him close them, and then you get 6%.
7:32 Like, is that how that works? Is it that simple for them? >> Yes. And we we encourage sellside brokers to not only take our finders fee, but to really be value additive to the process. that's what my recent article was about that I wrote on X was just about how to be value additive and help these business owners organize their data similar to a sellside for a a business, right? preparing their financials, presenting them in the correct light as opposed to potentially leaving money on the table.
8:01 >> and so a lot of these business owners, just like selling their business, don't know how to present their data. And so it's makes it more difficult for us to even correct like get to the best valuation we can for them. so if you could be value additive and we could pay you a finder fee and they can help pay some type of like management or you know advisory fee, then that's that's that's even better. >> Awesome. So yeah, we'll put that link in the description for sure. If people just want to like send Ryan their deals, let him close them and then you just get a kickback. That's I mean that to me you can make a killing doing just that cuz I think before we recorded you had said that your min or is it your average deal is 300k. So somebody would get basically 18K on average for sending you guys a lead that ends up closing. Is that about the average deal size?
8:47 >> Yes. Yeah. >> Okay. And another thing that you mentioned that I think makes a ton of sense is almost looking at this industry like like you're a business broker, right? Like when you made that comparison to the M&A world, there's I think there's going to be a whole industry of companies that pop up that represent companies not in the sale of their business, but in the sale of their data. And a really good broker representing a company in the sale of their data is going to do exactly what a good business broker does. Advise them on here. Here's how we got to package your data. Here's how we've got to market your data. Here's how we've got to present you guys to different buyers.
9:25 not just go with the first buyer that knocks on your door and offers you a six-figure check. So, that makes things a lot clearer to me. >> Mhm. >> Now, okay. So, my next question, and this is something I've seen, you know, being commonly echoed across the timeline is the runway for this industry. Like, I'm seeing people say like, "Oh, this has got three to six months left before it's just, you know, the the labs have bought all the data and this is no longer an opportunity."
9:54 What are your thoughts on that? >> So, there's there's a few things. I think that the timeline's going to be longer. and the main reason why is due to exclusivity on the data. So, usually a company's anchor buyer is going to require that the data that we buy goes to them exclusively for a period of time. which means that if you count the number of frontier labs, there's you now need several of every single industry just to feed like a base layer of demand. so I do think it's going to last longer than just three months.
10:30 and especially the more rare and niche the data sets are, it'll probably just move from where more common companies are less interesting, but then still if you have a 1,00 employee distribution company, well, that's going to still be interesting. There's not as many of those. so that I think there's just going to move up market in terms of what's interesting to the labs. Additionally, there's other types of data and this data as a revenue stream for businesses is going to become I think a more exploding idea. Ali from Micro One was talking about it recently.
11:03 that you know next thing could be egocentric data which is basically wearing headcams and collecting robotics training video. And so companies that have bluecollar workers can have their c like employees wear these headcams and get paid hourly to do that. >> That is incredibly interesting. So, okay. So this you're kind of convincing me that this space has more legs than I initially anticipated because again I think people are I mean again you would know better than me but from what I understand like the prices are already starting to come down as far as >> the offer that you might have made to a business in June might be a lot higher than the same offer that you would make today just because there's obviously more competition. But you're saying that overall there's still going to be a demand. the demand is just going to shift to either more niche specialized data or to this maybe harder to get data which is like actual I mean call it like blueco collar data for lack of a better term where it's like hey let's go strap strap a camera to our our line workers or to our technicians and get this robotic this data that we can train robots on like is that kind of the the gist of what you're saying there?
12:17 >> Yeah, that's and that's only one more way. there's another you know theory out there that enterprises are going to start to need data to do their own model training. like Fortune 500s are all you know kind of taking this idea. there's a few like the Thomas Ruters the law firm and financial firm. I think they're the financial firm but they they are training their own models and that's going to potentially become more common. And if you're an example is like a $20 million business, well, you might need hundreds of millions of dollars worth of transaction data to train your model.
12:55 And the only way to do that is to aggregate several of these data sources to train your own models. So if that if that happens, then the world of buyers explodes from maybe 10 to hundreds of thousands if enterprises start training their own models, which is another theory for the market too. That is such such an interesting insight because you saying that makes perfect sense. I mean, I think we're already starting to see like the Fortune 10s or the Fortune 100s or even the Fortune 500s of the world starting to train their own proprietary models. And that's I mean that's a moat for them. That's a huge asset for them.
13:32 >> So, it makes sense that they would be willing to pay for these other private data sets to increase the speed at which they can train their own in-house models. therefore increase the size of their moat. And to your point, like that opens up the the pool of buyers. Right now, maybe there's just like 10 main buyers that in terms of the labs accounting for like 95% of the data spend, but two years from now or five years from now, there might be thousands of buyers. And it might be it might just be like a a landscape just like a the business buying landscape right now.
14:05 Like, yeah, there's big corporations that buy up tons of small businesses, but >> there's tens of thousands, if not hundreds of thousands of people actively looking to purchase small businesses in 2026. >> And you're saying that the data market might go that direction. Is that correct? >> Yeah. Yeah. I think everyone's going to look to train their own models because of the cost savings of open source and they'll need the data to do it. So, >> dude, this is so insightful and I have a billion more questions. One thing I'll say for the audience, so again, I know we're like we're covering a lot of topics here. Ryan's sharing his screen. He's got some slides that he's sharing here about kind of how they operate as a company. We're going to take essentially everything we discussed in this episode here and we're going to just condense it into like an easyto read almost like playbook style summary that we'll just put in the description and in the show not. just like a free download if anybody wants kind of the 8020 of this conversation so that they can either go and look into this industry themselves or again look to go and partner with somebody like Orion and just be his lead gen more or less. So we'll have that resource available for you guys. Now okay so Ryan another question that that I see people have all the time is what kind of data are the labs looking for?
15:26 >> Yeah. >> in terms of the sources, I guess to to even be more specific, like, and not to not to hit you with too many questions at once, but like is Slack more valuable than Teams? Are we looking for email primarily? Like, what actual data are we are they buying? >> That's Yeah, as far as that, it's going to be the full corpuses. There might be some labs that work on a specialty project where they might need 10,000 PDF files for something. but for the most part, we're going to need the full context data. And so this is something I wrote about in the article as well where yes, a company might have 20 years of history, but if within the last three they've changed all of their softwares over and only have partial context for anything past the last three years, we can only really score the last three years for for our valuation for the full context. So that's super important.
16:21 >> Got it. Okay. And I see one of the kind of the footers on your slide here. Depth of history beats raw volume. >> So that leads me to the question is what is like your perfect what's the perfect fit company in your mind? What what criteria does that company meet? >> Yeah. I would almost say like remote first is an advantage in this world. you know, I think that it's not to say that this is operationally correct because we got a lot done by being in person at my company, just by even using whiteboards to help make things like just happen, right?
16:56 >> but it's really for the data world is it's not even about what happened. It's how it gets recorded and and documented the whole chain of of a of a workflow. And so probably remote first attributes, white collar would definitely be more interesting because a lot of the work happens in emails in Slack messages on you know word files and things like that. so those would probably be the two big attribute points is is remote first white collar.
17:27 >> Got it. Remote first white collar. Are there is there like a minimum number of employees? Is there a minimum amount of time that they need to be in business? like could you get a little more granular with the criteria? >> Yeah. so the market keeps moving up as far as what is more likely to transact. and so 50 employees is kind of the if you go above that number, you're probably going to be able to get a deal. anything below that becomes, you know, a little bit more difficult to transact. There's a lot of fixed costs involved. And so then the smaller deals are very similar to like how M&A works. There's the same amount of legal and due diligence involved. So why not just do bigger deals?
18:10 >> U so 50 plus employees, three years in business ideally with all the same softwares. we prefer US-based. There are opportunities for non US-based but English is currently required. you know all the deidentification softwares and things like that are currently being built out for English language and so even the possibility of translating it incorrectly and missing something that was personal identifying is is just too risky.
18:43 and then industry-wise it rotates and fluctuates around. The labs change their demands regularly and so you know couple weeks back there was a demand for heavy construction businesses and they wanted you know the these huge commercial contractors and then recently the demand has been video game companies and couldn't tell you why, right? They don't they don't necessarily give us all that information, but they just say go chase it. And >> that's super interesting. >> Yeah. Yeah.
19:13 >> Okay. So, it sounds like the industry there's almost like a flavor of the month when it comes to the industry, but what about what about the like the super regulated industries? So, like banking, finance, healthcare. Obviously, there's HIPPA compliance issues, but I from what I understand, some companies still look to buy that data like >> Oh, yeah. >> So, so I guess that's the first question is how how do you guys treat those industries? Like let's say if you've got a healthcare company that's willing to play ball, how would you approach that?
19:46 >> Well, it's yeah, healthcare is a little bit easier. there's currently a lot of practices in place because of like like re literal research purposes for for healthcare purposes. that have been in the space to deidentify patient trajectories and patient files so that they could be used to train the next batch of doctors. So that one's you know a little bit less concerning just because those those already exist as workflows and and processes in these companies. the it definitely gets more hairy when you start talking about high finance and and those other professional services where confidentiality is typically at play.
20:28 and non-disclosures where yes, just because something isn't personally identifying, it's potentially still materially valuable and it it difficult to even scope out for PII. And so those high finance legal type of deals we work with a third party to do on prem deidentification. It's definitely more expensive to do that, but just for the risk of what could happen, the only reason you should do this is if you feel like it's a safe transaction. And that's something that we add to be able to do that.
21:06 >> Okay. So, when you say onprim deidentification, that's that was one of my main questions, too. And something that I've heard people asking is like, well, how I heard a guy last week say, you know, okay, if you're going to license my data, how do I know I'm not going to end up on the headlines of CNBC as being the guy that, you know, let my my patient data or my customer data >> not be deidentified properly? So, like >> I'm sure there's so much that goes into it, but like how do you guys act like what does the process look like for for deidentifying data?
21:38 >> Yeah. So first we use third parties and there's two routes. You can either transfer the data to them to run on their servers and deidentify which is slightly cheaper but still expensive or the on-prem is the most expensive option and they're using you know series of algorithms to look for any personal identifying information any nickname associations any addresses locations emails phone numbers and they're replacing them. So anywhere that it would say Corey for example would become employee 632 and that way the relational context of the data is like retained and that's that's really important that relational context piece because some people ask if they can do the deidentification themselves and it's a hard no because you can destroy the the full value of the data set if you would properly deidentify it. so it's a very complex piece of the pie right now and so it it's important to do right obviously to protect the safety but then also to protect the value of the data set and then who's liable like if let's say the third party company makes a mistake and we forget to you know anonymize Sally's name and then her health information leaks out and you know it ends up in somebody's clawed like search like prompt results. I I don't know if that's can actually happen, but you get the idea. Yeah.
23:03 >> In that case, like who's liable? Is it you guys? Is it >> the data anonymizing company? Is it the company who sold the data? And I'm sure that this is covered in the contract, but like from a general perspective, who's liable? >> It would based on the contract, everyone in the chain would be liable. >> Yeah. and you know, there's some protections for the business owner. we have been working with a insurance provider to ensure these transactions.
23:32 and so we've got a >> yeah we we underwrote a specific policy for this and so there's a pilot we have going on to test this out with them but just based on the contract, you know, it's most likely everyone. >> So I feel like that's a insurance product that is going to explode. Like there's going to be I'm sure there probably already are, but just and the reason this comes up for me is like my dad's been an insurance agent for 30 some odd years. My cousin and uncle have been in the insurance industry for 14 years. I worked in an agency for, you know, six or seven months at one point.
24:08 So that in my mind is like, dang, I kind of want to just go up and start up a specialized carrier that just just sells these essentially like data brokering insurance policies, if you know, for lack of a better term. I mean, that that in my mind is going to blow up in the insurance world. It's just But it's like how do you underwrite it? I guess there's not enough transaction data at this point to accurately underwrite.
24:32 >> Yep. And so that's what the we're currently doing like a very small pilot. So it's not like a full product yet that anybody can get. so we're testing it but then if it does come to be you know a product. I think it's also really interesting for insurance companies because then now you're you know it's basically free lead flow as well for them. >> Right. >> To win the full policies. >> Right. Right. Right. Dang. This is so interesting. Okay. So what are the what are the most common objections you get?
25:01 like you've got a CEO on the hook. He's got a hundred employees. You're you you're dangling a million dollar payout in front of him. He obviously wants the quote unquote free money for the asset that he already has for virtually zero effort on his part, but he also wants to make sure that he's not putting the company or the shareholders or the board in a spot that is he's going to later regret. So like what are all the questions him or the CFO or the CTO are nailing you with when you're in the later negotiation rounds?
25:31 a lot of it's going to be just some of those contract terms. what are their protections in the contract timelines? We haven't really had a no. typically that's kind of self- selecting throughout the funnel. Like, >> right, >> if they weren't interested in potentially transacting then, >> so we haven't had anyone be a no. Most of it comes down to just the negotiation of terms and price there at the end.
26:02 yeah, the the big one I would say is just the the deidentification and it's kind of a real big separation of what we do differently is this third party. it's how we win a lot of deals over some of our competitors. you know, when you talk about them receiving your raw files, they have even less incentive to delete those because there's other valuable data sources in there. that's potentially valuable. Not to say that anyone's doing this, but I'm saying it's it's tough to know what they're doing once you transfer them. So, >> yeah, that would be those questions around deidentification definitely come up and that's what we've optimized for.
26:39 >> And so, you typically answer that saying essentially walking them through your process and being like, look, we use vetted third parties. There is stipulations in the contract that protect you to an extent should anything go wrong in that process. Like, and and that that's usually enough for them. They're like, "Okay, yeah, you know, sounds good." >> Yeah. Yeah. The and the third party, the benefit of that is they're only going to have the raw files for call it a week for the time it takes to deidentify it and then they get a you know, they get deleted >> and they get it. And so that protects the business owner to know that before I receive the files, >> it's all been deidentified. And so it protects me and the the company owner.
27:21 So it's I'd say that's that's the really big one. Yeah. >> Got it. Okay. Now, are there any questions that I haven't asked that you you that you get asked all the time? Like anything I just blatantly missed? Cuz I could probably sit here for another two hours and just pick your brain on this, but I want to make sure I always try to be a standin for the audience and say like, you know, what are what are people going to ask?
27:47 whether they are a company looking to sell their data, whether they're somebody who sees this as a business opportunity to be a middleman, anything we're missing here? >> The exclusivity periods are are probably one of the more talked about things. And so some people are are asking for perpetual licenses to these data sets, exclusivity, >> which I think that's probably a bit rich for for the buyer side to be able to to get that. and that potentially carries weight in an M&A transaction. so I would that would be one thing I'd recommend any any sellside broker helps establish for their clients is like, hey, we're not going to sign perpetually.
28:30 >> right. >> And then moving then forward, what is the term? Is it 12 months, 24 months? typically it's going to be a 24-month ask from the buyer. and and that protects their ability to then turn the data set around and make the money that they're supposed to cuz they're they just need to protect their ability to turn this into a product and then sell it up the chain to the Frontier Labs. and if somebody else gets it before the Frontier Lab before they sell it to the Frontier Lab, then the Frontier is going to be like, "Hey, I already have this." So, I would say exclusivity terms typical is probably 24 months.
29:14 yeah, that would be the other big piece. >> And obviously the the longer you're willing to be exclusive, the more the buyer is going to be able to pay, but then there's a trade-off at a certain point, right? Like to your point, I would never sign a perpetual deal. Like you that that in my mind is probably the worst move you could ever make as a seller because you don't know what's going to happen down the line. you don't know if that might be more more valuable at a certain time or another question I have is like what if I sold my data or licensed my data and then I want to sell my business like is that going to affect the value of my business should I want to go and physically sell my entire business obviously that's something I'm going to have to disclose during due diligence I mean how how does that work >> yeah it does it does come up in due diligence and and so that's where we're we're more on the like deal making side.
30:05 You know, my partner and I's background working through M&A transactions, we're definitely more of dealmakers and understand what the that's a bigger deal for business owners than making a few hundred thousand. So like if that's context that we're given from the sellside broker that this is something they're looking to do in the in the near future, we can time our you know expiration terms to the exit timeline. and you know help even guide the business owner that's buying the business through what these terms mean.
30:39 I have another article, I think I wrote it like two or three months ago at this point about this kind of problem, about what this is going to look like for M&A. but we'll definitely start to see this come up more as a due diligence item. because like even outside of this one licensing opportunity, there's probably going to be several others where data as a revenue stream come into play. >> Got it. And that makes sense. I like I like your framing around like we can use that as a deal point. Like if we know this particular owner, they want to sell their data today and maybe they don't want to sell the business for another two years, then we can make sure the exclusivity period is for about 2 years so that when it's time to go and sell the business, yeah, they'll disclose during due diligence. Well, yeah, we licensed our data two years ago, but that exclusivity period is up. So, hey, Mr. New buyer, you're buying the business, but you're getting this asset that you can then turn around and sell again with it. So like technically they're getting a nice kicker if that if you if you structure it right. Is that kind of what I'm hearing?
31:41 >> Yeah. Yeah. And then I I don't know that I and just understanding on the financial side you're probably not going to be able to add it to the EBIDA value. >> Right. Right. Right. >> Oh look we doubled our we went from 300k in profit to 600k this year because of >> Yeah. Because it's a one because it's like a it's a one time. Right. Like these are not recurring transactions as far as I understand.
32:00 >> Yeah. Yeah. Right. That's the structure. >> Got it. Okay. No, that makes perfect sense, dude. This has been super super eye opening and I'm I'm glad that I was able to hear it from someone like you who is in this space because I have a feeling with this I mean kind of being like the wild wild west right now or at least it feels like I'm sure you you probably know that better than anyone. >> There's going to be a lot of people that pop up talking about this thing that don't really know what they're doing. So good to hear it from you who's someone doing this at a high level and and actually having success with it. So just a couple reminders for the audience. So again, if you want to bring Ryan leads, like one I think one opportunity right now that people have is if you're good at lead genen, like let's say you're really good at cold email or you're really good at content or you're really good at paid ads, be a a lead genen shop that just funnels leads to people like Ryan and take a 6% commission on every one. And if the average deal size is 300 grand, that's an 18k commission you're getting. And I know it's not going to cost you 18k to get a qualified lead in the door. So, that's kind of where my mind goes, always thinking about how to monetize this opportunity. So, again, we'll put links in the description, links in the show notes to like one click take you straight to Ryan's site and you can start sending him leads.
33:16 We're also going to put that playbook that I mentioned in the description and in the show notes if you want just like a nice summary of the 8020 of everything we covered. That'll all be there. Ryan, any closing thoughts from you? I think that this is going to be a great space and there's going to be a lot of room for great people to be a part of it and I want those people to be with PolyShares and I'm going to be here to help support them as they try to make money in this industry.
33:44 >> I love it, man. Well, again, you seem like you know your stuff. I can tell you definitely do. anywhere else you want to send people whether to follow you on X, follow you anywhere else, >> follow on Twitter is great. >> Perfect. And we'll put Ryan's information in the show notes and in the description. So Ryan, thank you so much for your time and for everybody that tuned in. If you enjoyed this, like the video, leave a five star. I mean, this was one of the most interesting ones that I've ever done and this is like 300 plus episodes at this point. So appreciate your time, Ryan, and I'm sure we'll be talking soon.
34:14 >> Awesome. >> Ryan Lockach, so first of all, who are you and why should people listen to you? I know you're deep in the data brokering trenches, if you even want to use that language. You're, you know, you're buying and selling data on behalf of the big AI labs and making a ton of money doing it. So, like, >> you know, why should people listen to you? What is somebody going to take away? >> Yeah. so first, I I don't mind the data brokering language. We operate as like the layer that helps aggregate the data, turn it into training packages that then sell it on to the frontier lab. So we are doing a value lift in the process. But I mean I think that in every industry that's of this size of deal, there are brokers just helping guide people through the process. And so I think that that's just appropriate.
35:02 You know, as much as the label doesn't you know, some people like to avoid it, but it's just appropriate. So I think that that's okay. and you know, we've been around prior to the Big Spirit Airlines transaction. we were going through this on our own assets, me and my co-founder, back in July and June. and that's when we kind of understood the space and he's from the venture world and so we've had some easy connections into the frontier. Excuse me. Can we reshoot? Sorry.
35:33 >> Yeah, no worries. Let me I'll even stop. Let me stop.
Summary
- Polyhares generated $1 million in revenue in its first month by brokering data for AI labs.
- The company acts as a procurement partner, securing data rights from businesses and preparing it for AI training.
- Average deal sizes are around $300,000, with potential for much higher values depending on the data's uniqueness.
- The data industry is evolving, with increasing demand for specialized and niche datasets as AI labs exhaust publicly available data.
- Effective data brokering requires understanding the full workflow of a business, including communication and operational processes.
- Remote-first, white-collar companies with at least 50 employees are ideal candidates for data transactions.
- Data deidentification is crucial to protect privacy and value, often requiring third-party services to ensure compliance.
- The market for data brokering is expected to grow as more enterprises seek to train their own AI models, expanding the pool of potential buyers.
Questions Answered
What is the background of the speaker in the data brokering space?
The speaker, co-founder of Polyhares, discusses how they started their business during the data licensing process and achieved a million dollars in revenue in their first month. They emphasize the rapid growth of the data industry and hint at the insights they will share regarding the industry's needs and future direction.
How can individuals benefit from referring data leads?
Individuals can refer leads to Ryan and receive a 6% commission on closed deals. The process is straightforward, and Ryan encourages brokers to add value to the data presentation to maximize the deal's potential.
What type of data are AI labs looking to acquire?
AI labs typically seek comprehensive data sets rather than specific types of data. They may require full corpuses of data for training models, indicating a preference for rich, contextual information.
What is the process for deidentifying data before selling it?
The deidentification process involves using third-party services that apply algorithms to remove personal identifying information while retaining the relational context of the data. This is crucial for both safety and maintaining the data's value.
What are the typical exclusivity terms when selling data?
Exclusivity terms for data sales usually last around 24 months, allowing buyers to monetize the data without competition. Longer exclusivity can lead to higher payments, but sellers should avoid perpetual agreements to retain future value.