# S1 E12 - Neuralift.ai

Episode 12 - Neuralift.ai 

[00:00:00] ​ 

[00:00:05] Elgato Wave Neo-24: welcome to the middlemen. I'm Tom Limongello, and I'm here with Todd Sawicki. We, as middlemen, live at the intersection of media and e commerce, and we would like for you to join us in our discussions where we turn that chaotic intersection into your comfort zone  

[00:00:20]  

[00:00:20]  

[00:00:20] Tom: this is a welcome back for us from Labor Day. We recorded this prior to Labor Day, but I think from listening to it, I got the feeling that this was a really interesting, pivotal moment in understanding all the recordings that we've done, all the podcast episodes we've done previously, the retail media series which was three episodes the episode we did on Oracle, all of those were focused on the shift in the advertising market from third party data to first party data. 

[00:00:48] Tom: But we did it from the background and the perspective of a retailer or publisher where it was all about getting their transactions set up so that those [00:01:00] monetization and for sales attribution. But what we didn't talk about was how do brands live in this new world. And I think that's why this discussion with Jonathan Mendez from Neuralift is going to be very interesting and you have a long history with him. 

[00:01:16] Todd: the Jonathan's background in online advertising. And he really started in his early in his career, looking at how you leverage third party data and optimize for that is really interesting. And I think also an important point is to, it speaks to the radical shift that's going on in the online advertising and online marketing ecosystem from third party data to first party data. 

[00:01:39] Todd: We cannot understate how big a change this is. Where almost all on advertising and marketing was built around third party data and Oracle's business to refer to that and listen to the story of what that was about, which is Oracle shutting down its Oracle Advertising Technology business, which was doing [00:02:00] hundreds of millions of dollars revenue, but it was built entirely around third party data and helping businesses market using third party data, and that's going away. 

[00:02:08] Todd: And I think Oracle sees the writing on the wall that with that business going away. Meaning third party data targeting shifting entirely to first party. They just gave up and said, you know what? We got better things to do with our time. And. That radical shift actually goes back. It didn't just start this year or last year. 

[00:02:25] Todd: It goes back before the pandemic. And Jonathan's story, that radical shift goes back before the pandemic. And Jonathan's story speaks to that. And he, Got into the business of helping brands build out CDPs, right? That was really the beginnings of this first party data story was, how do you help, how do brands do that? 

[00:02:49] Todd: They put it CDPs in place. Jonathan built a business around helping brands do that. And then once he got in there, he realized that's great. That the brands have now started aggregating all this first party customer [00:03:00] data, but it turns out. That doesn't help you very much. And then you have to figure out how do I use that data? 

[00:03:06] Todd: How do I build that data into segments or cohorts? How do I export that data and transact on that from a media or marketing standpoint? And it turns out that the CDPs are great for storing data and pushing that data out to other platforms, but they're not very good or make it very easy to segment. 

[00:03:24] Todd: And build those audiences that you want to market to. And so Jonathan's got a new startup, which is really meant to solve that problem using AI to help make it really easy for these brands to build those segments in really interesting ways, leveraging AI, it's a really cool product. Jonathan's had a chance to walk us through that product. 

[00:03:43] Todd: And so it really, I think, speaks to a class of products and tools you're going to see emerge to help brands. Interact and do more with first party data and react to the world that's changing very quickly, where they're going to have to, if they want to do online marketing, they're going to [00:04:00] have to do it around and leveraging first party data. 

[00:04:02] Todd: And so that's really what Jonathan's story is about today is you brands, this is your future and you're going to work with tools, like the ones that Jonathan's built. 

[00:04:11] Tom: Yeah. And I think that a lot of, people like me who work more on the retailer side more recently, We were pushing and selling personalization, but the truth is that's not really the retailer's job. It's more likely the brand is going to have to take their data and say, this is what I want to do with these types of cohorts and these segments. 

[00:04:28] Tom: So I think, your point about, this is the time when those that other side of the marketplace lights up. And so I'm excited to allow the whole middlemen community to listen to what Jonathan has to say. 

[00:04:43] Tom: we want to welcome Jonathan Mendez, who's currently at Neuralift, and we wanted to really get to what his journey has been in the customer segmentation space, because I think you have a lot of really interesting. Companies that you've been in and those have all [00:05:00] informed where you are now. 

[00:05:01] Tom: So can you tell us your journey? 

[00:05:03] Jonathan Mendez: Yeah, absolutely. And thanks for having me. This is great to to talk shop with the middlemen. So for me, it's always been about first party data and segmentation to do that. I started I had, I go even back before I got into tech, I had built an e commerce company called vitamin lab. 

[00:05:22] Jonathan Mendez: com and we acquire customers from search. The, 

[00:05:26] Todd: I hope you kept that domain. That domain would probably worth a lot right now. 

[00:05:30] squadcaster-di5i_1_08-23-2024_121712: I still have the 

[00:05:31] todd-sawicki_1_08-23-2024_091712: There you go. 

[00:05:32] squadcaster-di5i_1_08-23-2024_121712: Yeah. Yes. But I unfortunately let go of vitamin multivitamins. com, which I saw some years ago was sold for 40, 000 which was one of those, 

[00:05:43] todd-sawicki_1_08-23-2024_091712: Could have been a nice car. 

[00:05:45] Jonathan Mendez: Yep. Yes, probably worth more than the 240 domains I still have But with search, people are self segmenting immediately when you get into a search engine and you put a keyword into the search engine as a query you've [00:06:00] immediately segmented yourself and the whole search index works that way, right? 

[00:06:03] Jonathan Mendez: You're it's matching. Based on a rule, that rule is the query. And when you come from search to a website like we had to buy something, back in the old days, Google used to be past that keyword along to you. And so you could know that somebody was coming to your site looking for multivitamins versus looking for vitamin B 12, for example. 

[00:06:23] Jonathan Mendez: And so that was a really powerful kind of segmentation technique to use to improve conversion rates and show people what they're coming for, what's relevant to them. When I joined offermatica I joined, they had been using that same type of technique around segmentation and most of all, looking at all of the data. 

[00:06:41] Jonathan Mendez: What offermatica did was we would look at everybody coming to the site and then we would slice them up into People who came in the morning, see how they performed. People who came in the evening, see how they performed. People who came from paid search versus organic search, right? 

[00:06:55] Jonathan Mendez: You'd have all these different segments of people that you could look at, [00:07:00] you could see performance, and then based on the performance, and there's always differentiation of that performance based on, how you want to slice up The audience and the visitors you could start to get ideas for what they, what off market did incredibly well as a B testing, multivariate testing and targeted content or what they call personalization now. 

[00:07:19] Jonathan Mendez: And when I came in very early, I had the search background, they were doing that In session in sight. So they were doing recommendations. And what I helped do is bring that into landing pages. And then, of course, homepages and really everything from search. And this was about 2006 2007 when search was really taking off commercially. 

[00:07:39] Jonathan Mendez: And So I came in to help get brand successful doing that. And everything was, as I said, we would look at, how people got to the site. We would look at the keywords that they were using. We would look at the performance metric, everything. We're always optimizing the conversion rate. 

[00:07:53] Jonathan Mendez: And and that type technology did incredibly well by really creating the market for testing. But, [00:08:00] testing is really about segmentation, too. And and so that was, that was my beginning of first party data because all of that data, again, this was site behavioral data, as they call it now or behavioral data. 

[00:08:10] Jonathan Mendez: There are segments within that entire data set that , show themselves, right? When you look at the results and you're able to segment those results by by different behaviors. So that was what we did it all from Attica. 

[00:08:21] Jonathan Mendez: Obviously proved incredibly successful required by amateur and then adobe and the product is still out there as adobe target. And doing incredibly well. So I think going back to the original Point is, with segmentation, you want to use all of the data you have at your disposal to understand who your customers are, what's important to them, what's relevant to them. 

[00:08:43] Jonathan Mendez: And with what's done in CRM systems or even now within CDPs or the data warehouse, you are using data minimization techniques by doing a SQL query. So you're saying, okay, I have all this data, but this column and this column or this [00:09:00] feature are the three most important things to create my segment. 

[00:09:03] Jonathan Mendez: And that may be true, but those are typically KPIs or metrics or something else that doesn't tell you what's relevant to those people. And that's why I think conversion rates have been stuck at 3 percent generally for e commerce for, the last decade plus is because everyone now segments in through that manner. 

[00:09:21] Jonathan Mendez: And I think it's wrong. 

[00:09:22] Todd: One thing, to continue your journey. To bring up, for those not aware, I went offer Madigan started a company called yield bot, which really was a a play in the shopper marketing world. And, it's not necessarily perfect analogy for retail media, but it definitely starts to begin. 

[00:09:40] Todd: I think what we now think of as retail media. And so it'd be good to understand, the story of. of YieldBot as well, 

[00:09:48] Jonathan Mendez: so the Genesis of Yobot was really came out of All4matica. Before we got acquired, we had started trying to optimize display ads in a similar way on site with first party [00:10:00] data from the retailer or from the publisher. And we did a lot of publisher work at All4matica too. 

[00:10:04] Jonathan Mendez: I don't want to discount that. We did work with around content engagement. We used to call content merchandising. So personalization just to get, higher engagement rates, more page views. And the idea for Yobot was what if you could, instead of take, drive people to the landing page, based on a keyword, you could take the landing page to the people based on a keyword and serve it out or serve an element of it, the headline or the call to action or, something. 

[00:10:32] Jonathan Mendez: And this was, it was and the cost of actually serving it and the cost of media with the rise of exchanges meant that you could do this relatively inexpensively. It used to be, you want to buy a display ad, it was 20 CPM, and you to figure, but now it was, five cents some places. 

[00:10:48] Jonathan Mendez: And so the It seemed reasonable that you could now, use the same kind of keyword based targeting in session, use that first party data as the rule for the [00:11:00] ad match, and then serve that ad. And that worked incredibly well for click through rates. And so to your point, we're around shopper Todd. 

[00:11:08] Jonathan Mendez: What we ended up doing is we ended up using that technology to drive a lot of traffic for brands. Into their shopper marketing pages on places like walmart. com or target. com or others. Because what they want is traffic and they want ultimately in people to purchase the product. And so 

[00:11:27] Tom: frame are we thinking about at this point?  

[00:11:29] Jonathan Mendez: this is 2012 to 2016 and yeah, it's funny. We you might remember triad the huge shopper company. So they were one of our first customers buying our By our media and it occurred. 

[00:11:40] Todd: working with the publishers, right? retailers, and effectively driving, right? This would be like what retailers are now starting to do with audience, like YieldBot being too early. All these RMNs now doing offsite audience targeting, YieldBot would have been like one of the perfect go tos for them. 

[00:11:58] Todd: And those not aware, YieldBot [00:12:00] ultimately shut down. It was one of those classic examples of Being early is still wrong. They were early by about five to eight years. Cause again it, with audience extension now is a key component of retail media. It would have been a perfect partner to 

[00:12:13] Tom: yeah. So I'm laughing because I'm laughing because I was a quotient and that was right after triad had their day in the sun. Quotient came in as basically an audience extension provider to retailers. It didn't have a ton of traffic. So 

[00:12:27] Jonathan Mendez: we talked Corp dev team at Quotient a bit. They were smart. They saw the opportunity for us. It didn't it didn't end up working. Actually, I think the guy left that we had was our main point of contact there. But absolutely. It it was definitely ahead of its time. 

[00:12:41] Jonathan Mendez: But the value was there for the publishers and for the brands. We were. Our largest customer was Meredith now dot dash, which of course is doing something similar now with they've developed a product that basically creates keywords based on site session, behavior and contextual signals on the page and all the stuff that that we did back [00:13:00] then. 

[00:13:00] Todd: One of the guys over there at dash running it is Pat from Xander. And that ties that together too, in terms of his background in programmatic and probably has familiarity with what you guys are doing. 

[00:13:10] Jonathan Mendez: Yeah we didn't do programmatic. That was part of our, that was part of the challenge. And that was actually part of why I ended up leaving is because we didn't want to take third party data. We didn't want to, we only wanted to use our data. We only wanted to use our tags. 

[00:13:25] Jonathan Mendez: We wanted direct relationships with the publishers.  

[00:13:28] todd-sawicki_1_08-23-2024_091712: See, nowadays that's all just first look, right? It's and again, this is the early part of the business is programmatic and the idea of like header bidding would have allowed you to do programmatic the way you wanted to do it. That would. 

[00:13:38] squadcaster-di5i_1_08-23-2024_121712: We did do header bidding. In fact, we did header bidding before, 

[00:13:41] todd-sawicki_1_08-23-2024_091712: That's the point is right. You were before header bidding and had access to publishers in mass to do first look and then do first party matching on first look. And bid on that, right? That was the missing piece. You were having to invent what we now call header bidding to do what you wanted to do. 

[00:13:56] todd-sawicki_1_08-23-2024_091712: And again, being early is still wrong. [00:14:00] Today it would be a lot easier to do those things. You front run. 

[00:14:02] squadcaster-di5i_1_08-23-2024_121712: run Google. We used to be able to get our bids in before Google and and because our average CPMs were about 3 we could get any impression we wanted. And and again, going back to just keyword based matching and the importance of words to understand the context or intent of somebody. 

[00:14:18] squadcaster-di5i_1_08-23-2024_121712: That was a very powerful combination for us. And yeah, and a company like Meredith, which has lots of intent driven content recipes. I remember all recipes, Seattle company taught was one of our huge. 

[00:14:28] todd-sawicki_1_08-23-2024_091712: familiar. 

[00:14:30] squadcaster-di5i_1_08-23-2024_121712: One of our biggest partners. I think we, we paid them, millions of dollars of revenue and and and again, back to shopper, right? 

[00:14:37] squadcaster-di5i_1_08-23-2024_121712: So who are the brands that wanted, people to get driven to all recipes, there was the McCormick spices of the world and, Tyson chicken was a big one. And so yeah, that world is, it's fascinating to see what has happened with retail media networks. 

[00:14:50] squadcaster-di5i_1_08-23-2024_121712: Yeah, I left because The, there was a lot of the board really wanted to take the business into the programmatic space and find programmatic partners. We had tried to [00:15:00] work with the trade desk early on and just the tech wasn't going to work really well. And  

[00:15:04] todd-sawicki_1_08-23-2024_091712: 2016 programmatic landscape is very different than 2024. I think that's one of the things that, people miss having, since I sold Samantha to Outbrain in 2017. The things you can do today with programmatic are so much more advanced and again with the rise of third first party to replace third party and the rise of header bidding, right? 

[00:15:24] todd-sawicki_1_08-23-2024_091712: That rise of header bidding really changes the landscape around first party. Targeting and bidding and selling right in a way. That's very dramatic. And I think actually, because one of the things we've talked about has been the rise of retail media. It's not rise in the sense of it's new. It's been a part of the landscape. 

[00:15:46] todd-sawicki_1_08-23-2024_091712: In calling different 

[00:15:47] todd-sawicki_1_08-23-2024_091712: It was called co op. Then it was called shopper marketing and now it's called retail media. But what's, and I think the question that we've been looking at, which is why is retail media in its current form, like [00:16:00] digital retail media really started to explode. 

[00:16:02] todd-sawicki_1_08-23-2024_091712: And I think some of the pieces that start to appear. The landscape are things like header bidding that really starts to take that off. It's like you bought was doing a lot of audience extension type work in 2012 to 2016, 17. And that was leading edge at the time. And today would be standard bear. 

[00:16:20] todd-sawicki_1_08-23-2024_091712: And if you're a smaller retailer, you really need offsite targeting to achieve the scale to make campaign sales work. And so I think that's when we put the pieces together, the rise of header bidding 

[00:16:33] squadcaster-di5i_1_08-23-2024_121712: Yes, 

[00:16:34] todd-sawicki_1_08-23-2024_091712: Really changes the landscape and enables this kind of one of the pieces that enables this kind of new form of retail media, 

[00:16:43] squadcaster-di5i_1_08-23-2024_121712: and the other thing I would say is the thread going back full circle is first party data. That's obviously a huge part of retail media. So understanding, being able to understand that these are, who these customers are when they're on site. And that's and that they're real people,  

[00:16:57] todd-sawicki_1_08-23-2024_091712: right?  

[00:16:58] squadcaster-di5i_1_08-23-2024_121712: There's a lot of first [00:17:00] party again, the, it's the, it's, it's a big area where data fidelity is important and helps performance. 

[00:17:08] Tom: Because we are moving away from it being super keyword focused to a world of LLMs and AI, , we had to hold on to these keywords as the messenger of intent. And I think that we're moving past that to a better contextual understanding we're now ready to talk about what is Neuralift and tell us that part of the journey. 

[00:17:30] Jonathan Mendez: Yeah. So after Yulbot, I started because I, again, was not on board with the programmatic future of of Yulbot. I went and started to what, in fact, what I wanted Yulbot to become was a CDP for publishers. And when I left YieldBot, I started building CDPs for retailers, not for publishers. 

[00:17:48] Tom: What was the original promise of because I'm I think this is going to be part of the theme and whatever we call this episode is going to be about. Like a CDP, a customer data platform was supposed to do something and it [00:18:00] really hasn't, so I would like to see  

[00:18:01] Jonathan Mendez: Oh yeah, I say that all the time. You have a CDP, so what? What do you do with it? It's a glorified data warehouse which is then now why the data warehouse space, this rise of composable CDPs, which is basically, just various components of the CDP put together with the data warehouse because the first party data needs to be a source of truth and it needs to have a home. 

[00:18:22] Jonathan Mendez: It needs a, you what They call a kind of gravity. It's got to live somewhere as a source of truth. And you don't want it in all these different databases. In fact, that's why CDPs became interesting. The idea of originally of CDP is to take, data from all these different tools that had their own tags and bring that together, unify that around a single ID. 

[00:18:42] Jonathan Mendez: Entity resolution or identity resolution So you could see this vision of a customer 360, you can start to see the same customer across all these different tools where previously you had that same customer with its own ID in four or five different tools. 

[00:18:58] Jonathan Mendez: You had them in your A/B Testing [00:19:00] tool. You had them in your email tool. You had them in your, site targeting tool. You had them in, all different places. And thus you had the same Customer segmented differently in all of these different places, which prevents you from actually speaking to that customer the same way across all the touch points, which is how  

[00:19:21] tom-limongello_1_08-23-2024_131712: yeah, I think what you're saying is  

[00:19:23] squadcaster-di5i_1_08-23-2024_121712: what brands want to 

[00:19:24] Tom: we've been talking so much about it, the rise of retail media, the rise of retail media basically keeps the brands in the dark and they don't have the power to be able to say, Hey, I actually know a lot about this customer.  

[00:19:36] Jonathan Mendez: Yes. There a little, I'll give you a little rub there because a lot of the buyers of retail media are CPGs and they don't have the first party relationship with their customers. So they need. Somebody to come along like a Walmart to say, Hey we know who your customers are or an Amazon, right? 

[00:19:52] Jonathan Mendez: It's, and we talk about retail media networks. We know that those are far and away  

[00:19:56] todd-sawicki_1_08-23-2024_091712: absolutely no. Amazon and Walmart are the 

[00:19:58] squadcaster-di5i_1_08-23-2024_121712: the two [00:20:00] big ones. And they know who those customers are and they know what they buy and the CPG brands don't. And I think they're trying. 

[00:20:06] todd-sawicki_1_08-23-2024_091712: do the CPG brands not know because they haven't invested properly in things like CDPs or they just don't know. 

[00:20:15] squadcaster-di5i_1_08-23-2024_121712: They haven't invested properly in first party data collection. So you think about, 

[00:20:20] tom-limongello_1_08-23-2024_131712: for them, like the retail, like the transaction still will live at the retail. This is the last episode we just had about Nike overshooting themselves trying to be half direct to consumer. That's very hard. Unilever and PG can't do that. 

[00:20:35] tom-limongello_1_08-23-2024_131712: They don't have PNG stores, 

[00:20:37] Jonathan Mendez: no they can't, but, they can collect data. Brands have, go back to the Columbia house record and tapes where you sent in your, penny and you got 11 albums and you gave them your name and address, right? And they knew. So they knew who you are. 

[00:20:51] Jonathan Mendez: They knew where you lived and no kind of music you're interested in. Coca Cola has done really well with this in yeah. Building and creating programs all around first party data [00:21:00] collection, contests, giveaways. So there are lots of ways to do it. And to at least get enough of a dataset where you can begin to model against so if you had, a certain percentage of first party IDs. 

[00:21:12] Jonathan Mendez: And then have the rest of your data in terms of sales data, which you have as a CPG, what's selling where you could start to build really good models to, to do that. But yeah, but I think most of the CPG brands are very far behind. I know we've been talking to them a few of them very far behind in first party data collection. 

[00:21:30] Jonathan Mendez: And so that's a huge challenge for them still. But. Brands on the other types of brands and other verticals, it's abundant. 

[00:21:38] Todd: Interesting that CPG is that far behind versus other categories. What do you think, to your point, what are some of the categories you think have done the best in terms of collecting first party data? 

[00:21:49] Jonathan Mendez: Travel hospitality is a great one. If you think about all of the programs around, that reward frequency that's that's a great, that's a great vertical. Anything online [00:22:00] with regards to retail that's transactional or, or retail or transactional has done a good job and, and obviously apps, anything, There's a lot of there, there's a lot of companies that have done well, gaming is another really great vertical for first party data with an abundance of it. 

[00:22:16] Jonathan Mendez: So there are there are some, there are others that are not as, sports entertainment as a vertical that we're active in at Neuralift. And that's, it's just, it's a little behind, I would say these other verticals, they don't, you buy your ticket, but you buy your ticket. 

[00:22:30] Jonathan Mendez: Through a ticketing company the arena and the team may or may not know who you are. And in fact, the ticket companies have built their own data businesses these days.  

[00:22:39] Todd: I think everyone with a transaction attached to it has built a data 

[00:22:43] squadcaster-di5i_1_08-23-2024_121712: yes, yeah. As well, they should. And again, that gets back to the first party relationship, which, again, wasn't important, arenas were happy to have companies. 

[00:22:52] squadcaster-di5i_1_08-23-2024_121712: sell their tickets on their behalf. I'm not sure they read the fine print that says that data belongs to the ticketing company. And, and now I think some of that [00:23:00] those contracts may be changed going forward because this is the back first party data is the backbone for I, I just think it's first party data is the most important data for any business to have to win. 

[00:23:11] squadcaster-di5i_1_08-23-2024_121712: It's, in a competitive environment, the best first party data will win. 

[00:23:15] Tom: So when we were talking previously, you had some examples of what a brand can do to use the data that they have, or even use the data that a publisher or retailer might have, and the order of operations matters, it's you don't just take a big dump of data, you segment them. 

[00:23:33] Tom: First, so tell us a little more about Neuralyft's value proposition and how does that work? 

[00:23:38] squadcaster-di5i_1_08-23-2024_121712: Yeah, so an earliest value proposition is the answer to the two, you'll have a CDP. So what, the, I have a customer 360. So what, it, what are you doing with it? How are you growing your business with your data? That seemed to be the missing piece. And I, I took an interim role as a Chief Digital Officer, where I felt this every day. 

[00:23:57] squadcaster-di5i_1_08-23-2024_121712: We had A CDP. We [00:24:00] had we had Google Analytics four installed. We had Google Cloud platform. We had Snowflake. We had a CRM. So we had all of the data in the world but nobody was able to say. who our customers were, how they are, how they should be segmented. And if I, as I tried to cause my charter was to come in and prove conversion rate. 

[00:24:20] squadcaster-di5i_1_08-23-2024_121712: The first thing I do is ask for, customer segmentation and okay, now I have to put a ticket in JIRA. I have to write an Epic it goes to the shared data team. I get in a food fight for prioritization. Of course, every other data need in the entire organization. And, three months later I get something back and I, data is a motor sport. 

[00:24:37] squadcaster-di5i_1_08-23-2024_121712: There is no three months later, it's the playing field has changed that data that those, that segmentation is probably worthless or certainly not, accurate based on the changes that have happened over a whole quarter in the business. So I thought that this was a perfect problem for AI to solve since the data is there the issue. 

[00:24:54] squadcaster-di5i_1_08-23-2024_121712: Is time to value from the data and precision from the data. And [00:25:00] so those are two things that AI does incredibly well. AI processes, really fast. And there are no limits to the amount of data that can be processed in fact. And so we could take a table. A neural lift of, 20 million customers and 800 columns and precisely segment that table in a few hours that's months and months of work for a data team to be able as a brand to have that done. 

[00:25:25] squadcaster-di5i_1_08-23-2024_121712: Incredibly accurately more accurately than humans could do it in for months to basically minutes of time the time to value there to action on that data while it's still relevant and important is a huge benefit to the brand as well as the precision, using all of the data to create segments rather than saying, Hey, these having some person say these five minutes. 

[00:25:46] squadcaster-di5i_1_08-23-2024_121712: Features in the data are important to build our model for segmentation that introduces lots of bias. It's usually the highest paid person in the room, gets to determine what data is being used. 

[00:25:58] tom-limongello_1_08-23-2024_131712: Yeah. So yeah, [00:26:00] I think when we first talked about this, it was Oh, sweet. So there's no challenges there. You just said, no brainer. You just sell it right in.  

[00:26:05] todd-sawicki_1_08-23-2024_091712: I think, it's interesting here. Is having, and for our listeners Jonathan has given us a demo of Neuralift and when you see it and how it segments users. It's it definitely opens your eyes like, Oh, totally. When he talks about the old fashioned way of doing SQL joins and you had to guess what this, what those joints should be and what the segments would be is that's how you made segments. 

[00:26:29] todd-sawicki_1_08-23-2024_091712: You'd make a segment. Did it actually work up that segment? Didn't actually produce any marketing results. I guess that's the wrong segment. Let's try it again and again. And so you're, that would take years because it was like once a quarter is all you could do is segment someone. And 

[00:26:42] squadcaster-di5i_1_08-23-2024_121712: it's a matching problem. 

[00:26:43] todd-sawicki_1_08-23-2024_091712: yeah, and then neural lift shows up and its ability to use AI to look at a data set and create the segments and hours without a lot of human intervention changes the game of how data becomes accessible and manageable. 

[00:26:56] todd-sawicki_1_08-23-2024_091712: I think that's another thing is, it's worth talking about Jonathan, which is [00:27:00] that, you talked about the timeline when you're a CDO and how hard it was to get segments is unless you've tried to do that, it is. It's hard to understand how painful it is to try to execute on a data strategy. 

[00:27:14] squadcaster-di5i_1_08-23-2024_121712: Yeah what we've tried to do is take combined two things. The first is the neural network that we've built that actually does the segmentation. And so that is to precisely segment a table based on affinity in the entire data set. So we're looking across the data to find the patterns that are important to create segments. 

[00:27:32] squadcaster-di5i_1_08-23-2024_121712: We don't have any bias before we do the run. We don't determine how many we don't know how many segments there will be. We don't know how large the biggest segment will be or the smallest segment, although we do have a minimum segment size that's programmable. We default to a thousand for use cases around custom audiences in, in, in meta and snap and tick tock, etcetera. 

[00:27:51] squadcaster-di5i_1_08-23-2024_121712: But the first thing we do is we have this neural network to do the precise segmentation. I think what's really interesting about and that's the way the [00:28:00] platforms, that's the way they're doing. There's no one at Google or meta sitting there, once you put your campaign in and saying, Oh, let me SQL query, our table of, Meta users and create the best lookalike audience. 

[00:28:10] squadcaster-di5i_1_08-23-2024_121712: No, that's not how it works. 

[00:28:12] Todd: mean they're not handpicking from a customer list of 1. 2 billion users? 

[00:28:15] Jonathan Mendez: right, exactly. So we bring that same power to a brand to be able to segment. But the other thing we do that I think is really novel and, that most people really get excited about is once we have those segments done, we've created a number of of generative outputs to explain the segments, what we call reasoning and explain ability. 

[00:28:36] Jonathan Mendez: So we show you with the metrics that the neural network creates, and oftentimes these are benchmarks or metrics that the brand has not even done themselves with their own analytics. But we present them and we show that these are the key data that created the segment and use large language models to, explain that to, even an intern would understand who these segments are, why they've been created [00:29:00] who's in the segment and what's relevant to 

[00:29:01] Tom: I hear that. And then I'm like, I look back to, The Blu Kai days, is it better than just margarita drinking empty nesters? Like what, all those like segment names, which are hilarious to read. Is that, is it, is the answer like more than just a title?  

[00:29:16] squadcaster-di5i_1_08-23-2024_121712: Answer. The answer, Tom, is always more margaritas. I think yes, it's more than a title. I, the titles this is, this is something LLM is doing incredibly well, right? And we've created the prompts so that the titles are memorable, but they're Based on the underlying data of why the segment is brought together. 

[00:29:32] squadcaster-di5i_1_08-23-2024_121712: But what's really important is the distinctions between the segments. And that's where and the other thing I'll say that we do that's important is we allow the brands to overlay their own KPIs on top of the segments, and that's all in the UI. And that's part of the the data process and that we do. 

[00:29:47] squadcaster-di5i_1_08-23-2024_121712: So the KPIs can be now seen. Across the segments that the brand has. So they, which surfaces immediate opportunities for strategies and where they should be investing their time and resources, if you see a segment that's [00:30:00] underperforming for, let's say average order value, that's incredibly important to get that type of information in, minutes to hours from a giant table is invaluable to a brand. 

[00:30:11] squadcaster-di5i_1_08-23-2024_121712: And for us, it's all, not only is it optimization of the res of the end result to lift the KPI, but it's the optimization of the time and investments that go in that these brands have to just operate and get marketing, that, that is going to perform. 

[00:30:25] tom-limongello_1_08-23-2024_131712: So does this fix some of the dumb KPIs that you used to see? I remember you telling us about a cruise line. Let's hear some stories about war stories about how you got to where you are. 

[00:30:37] squadcaster-di5i_1_08-23-2024_121712: I know, in case anyone's listening, I want to be a little careful about every company has bad KPIs. But I think that this shows you what I think Neuralift does incredibly well is  

[00:30:47] squadcaster-di5i_1_08-23-2024_121712: just understanding what's relevant to whom you, Todd mentioned the matching problem before. Optimization is a matching problem. So to understand, what I have to show to this group and that's different than what you have to show to this group, that's what we Madoka [00:31:00] back in the day and that's what we're doing with Neuralift, almost 20 years later, it's still it's what all ads are trying to do, right? 

[00:31:06] squadcaster-di5i_1_08-23-2024_121712: We're trying to match something we think that will be interesting to somebody. If it's interesting, they'll notice it. If they notice it, they'll take action about it. If it's relevant. This is basic advertising 101. It's just, now we have data to do it. And now we have AI to accelerate how it's done and to do it more optimally. 

[00:31:25] todd-sawicki_1_08-23-2024_091712: I think to going back to the point about KPIs and appropriateness of which are what, I think from an ad and marketing perspective, the we as marketers have had to lean into the best available KPI, even if it's still garbage and, click through rates or engagements you view through conversions, God, what, Jesus, that's probably the worst KPI ever invented. 

[00:31:48] squadcaster-di5i_1_08-23-2024_121712: Any KPI, Todd, that was invented by an agency is a bad KPI. 

[00:31:55] todd-sawicki_1_08-23-2024_091712: There's a good take. Let's drive over the, yeah, we're just throwing the agencies under the the bus [00:32:00] here. The body is getting squished as the wheels turn. Kaboom, kaboom. 

[00:32:04] squadcaster-di5i_1_08-23-2024_121712: Yeah, it sounded like a few of those are agency derived KPIs or or ancillary ad services KPIs. I think the one I've, the one I've ragged on a little bit lately is attention. We're going to be measuring attention now. And I think Advertising is there to sell people on a product and there's a place for brand, there's a place for performance they all have their KPIs, but. 

[00:32:28] squadcaster-di5i_1_08-23-2024_121712: At the end of the day if your advertiser should be growing your business and if it's not working and your investments are, are not returning ROI. I think, I've always been more on the performance side because it's listen, that's one of the great parts of digital marketing. 

[00:32:42] squadcaster-di5i_1_08-23-2024_121712: I go back to the beginning of why. I loved search, search was the greatest and is still is the greatest demand capture technology ever created. There's demand gen and I don't think search is great at demand gen. I think social has done a pretty good job of that as far as digital goes. 

[00:32:59] squadcaster-di5i_1_08-23-2024_121712: But I still [00:33:00] think, television and CTV and outdoor. And there are other great ways to to create awareness and create generate demand. But the other side of it is capturing the demand. And that is something that digital has always done incredibly well because of the experience. 

[00:33:16] squadcaster-di5i_1_08-23-2024_121712: It's just so easy. It's been come easier and easier now. I can just click on a button and one click by, or Apple pay. It's just, it makes it so easy to purchase and to transact. And I think that's a hugely underrated part of digital's impact. 

[00:33:31] squadcaster-di5i_1_08-23-2024_121712: Obviously we see it on companies like Amazon and, we see it in the market cap of Google and meta. But I think overall, it's not appreciated enough that digital is demand capture and other channels are really good at demand gen. 

[00:33:45] tom-limongello_1_08-23-2024_131712: Yeah, I think the reason we were trying to push in that direction was that I think you told us a story about how dropping into the checkout flow might, might be, not the right way 

[00:33:54] squadcaster-di5i_1_08-23-2024_121712: Oh, 

[00:33:55] tom-limongello_1_08-23-2024_131712: to measure, whether or not you're being successful. 

[00:33:58] squadcaster-di5i_1_08-23-2024_121712: I [00:34:00] yes I've seen a lot of gaming. Let's just say I've seen a lot of people game metrics and I think that's and KPIs and two lessons there are one, your marketing org should all be pointed to a Northstar KPI and everybody should be focused on improving that. 

[00:34:15] squadcaster-di5i_1_08-23-2024_121712: That KPI and probably it should be conversion or order value or something that could be measured very clearly and and succinctly lifetime value is a great metric, much harder to measure. It takes a lot of time.  

[00:34:27] tom-limongello_1_08-23-2024_131712: But you're as Neuralift on that side of the brand who potentially could actually know the LTV, whether they figured it out or not. Retailer has a tougher time with that because they don't actually care as much, they want aggregate, increase in value per category, but the brand specifically could have that LTV understanding. 

[00:34:48] tom-limongello_1_08-23-2024_131712: And I don't know where that's at today because  

[00:34:50] squadcaster-di5i_1_08-23-2024_121712: It's funny because you're talking about, yeah, you were talking about Nike and, they bought a company called Zodiac back in the day that was [00:35:00] specifically focused on being able to determine and predict LTV of customers based on, their behaviors and their transactions. And I don't know what ended up happening to the company. 

[00:35:10] squadcaster-di5i_1_08-23-2024_121712: I always admired it quite a bit. But it's I don't know. I have 

[00:35:14] tom-limongello_1_08-23-2024_131712: happy. Yeah, I asked because I. 

[00:35:15] squadcaster-di5i_1_08-23-2024_121712: What I find on marketing teams when I go in and, work with them Is that generally speaking, the email team has, KPIs like open rate or, click through rate or things like that. 

[00:35:28] squadcaster-di5i_1_08-23-2024_121712: And, the the product team has, something related to add, products added to cart or, time on site or return visits or things like that. And yeah. But there's nothing that kind of puts it all together. And so this is, one of the issues I found is that leaders, CMOs, SVPs, they are struggling strategically. 

[00:35:48] squadcaster-di5i_1_08-23-2024_121712: And the reason they're struggling strategically is that the teams that they have are all focused very tactically and they're comped tactically as well. And so part of what we want to do at Neuralift was say, [00:36:00] okay, Let's take a step back for a second. Your biggest problems are not tactics. Tactics are there. 

[00:36:05] squadcaster-di5i_1_08-23-2024_121712: You know what you could do. There's, the channels you have, you know how to leverage the channels well at this point, what you're deficient in is understanding based on your data, right? Your data should drive your strategy. And there's no product or tool that I've seen that does that. And so what we've done is say, okay, give us all of your data and let us show you top down. 

[00:36:25] squadcaster-di5i_1_08-23-2024_121712: You know who your customers are and they're once you understand that the strategies become emergent themselves. It's there in the data. You have groups of people who are, performing differently are interested in different things. By a different times. Churn, different rates, like all of these things are right there in front of you where strategies are easy to develop. 

[00:36:49] squadcaster-di5i_1_08-23-2024_121712: And I think that was the big thing as I went in, and got the idea for NerdLift was that's, that seemed to be missing to me in the market. And that seemed to be a perfect use of AI. [00:37:00] And that was why we started started playing around with Neuralift and, did a eventually did some initial built an alpha, built a beta, started working with some brands and seeing that, yeah, like this is, everybody was very excited at these, C level. 

[00:37:13] squadcaster-di5i_1_08-23-2024_121712: To start to get this level of detailed data driven insights on their customers, because again, these are the people that have approved the millions of dollars that went into buying the data infrastructure and tools and have waited, six to 12 months of an implementation to finally be able to use it and are sitting there and saying. 

[00:37:33] squadcaster-di5i_1_08-23-2024_121712: So what, what, where's the ROI, how are we going to use these, all this investment? And so I think it's fortunate the timing with AI now we'll be able to unlock those investments 

[00:37:45] todd-sawicki_1_08-23-2024_091712: One of the interesting things about your story and the journey is the focus on first party data and in many ways, yieldBot was early, begins to show us that value. And I've actually, started talking about we [00:38:00] should stop calling them retail media networks, because what's happening is anyone who has logged in users or transaction data is now launching retail media. 

[00:38:07] todd-sawicki_1_08-23-2024_091712: And I said, we should call it either a first party network, because that's really what's driving it. And, now that throughout now the rest of ecosystem with heterobating, it makes that easier to leverage that data. But really to me, retail media is about the rise of logged in users, about the rise of first party targeting. 

[00:38:25] todd-sawicki_1_08-23-2024_091712: And so we're really seeing the first party revolution. And so it's in many ways, the world is catching up with you. 

[00:38:32] squadcaster-di5i_1_08-23-2024_121712: or we're coming full circle. 

[00:38:33] todd-sawicki_1_08-23-2024_091712: Yeah, or coming full circle, right? It's, I'm fascinated about this too, right? And to your point about conversions as a better Uber metric or Northstar metric, the benefit of retail media or retailers is they have that, like they have a real conversion, not like a fake conversion, not some implied conversion. 

[00:38:51] todd-sawicki_1_08-23-2024_091712: They actually have a checkout transactional. Event that they can look to and that I think is an important aspect of [00:39:00] this, that, that helps people understand that now, to your point of Oh, you've invested all this money in starting to capture data and you can't actually use it. You're, if you're doing SQL joins and take a quarter to do any rev of it, it's horrible versus the ability to have someone come along like you and neural lift to say, okay, how do I say my users so I can actually take advantage of all these new first party. 

[00:39:19] todd-sawicki_1_08-23-2024_091712: Opportunities even outside of, and the good news from a marketer's perspective is the rise of this first party ecosystem means you, you maybe start to unlock channels outside of meta and Google. 

[00:39:33] squadcaster-di5i_1_08-23-2024_121712: I think there's, yeah, I think there's two things. One, you could be more efficient in those big channels because if you have precise segments that will feed a custom audience that's the seed, right data set that they're doing their modeling against. So you want that group. of IDs to be as high in affinity and similar as possible to get the, the highest resolution of the modeling that the platforms are doing so that, that helps you [00:40:00] perform better there. 

[00:40:01] squadcaster-di5i_1_08-23-2024_121712: But to your point, yeah, I think One of the things I think and, when I talk to to, CMOs is the challenge that they have alluded to earlier, which is, they see their customers across all the touch points. It's not just channel specific. And this is one of the, again, going back to one of the chat, the problems in marketing now is everybody's chat, everything, strategies are channel specific, tactics are channel specific. 

[00:40:22] squadcaster-di5i_1_08-23-2024_121712: The, the customer doesn't think in channels. The customer is walking into your store. The customer is went, gone, goes to your website. The customer might have your app, the customer sees you on in search. Like it's, you need to see cross channel. And I think one of the amazing things about CDPs is that they've, aggregated this data together across channels and unified it. 

[00:40:42] squadcaster-di5i_1_08-23-2024_121712: But again, okay, so now. What are we going to do with it? If we start to segment, start to see segments of our audience or customers across channels, that is incredibly powerful for a CMO in terms of understanding where to make investments, how to talk to people the right way, which people are [00:41:00] important to me, which people are not. 

[00:41:01] squadcaster-di5i_1_08-23-2024_121712: Some customers, are just not, shouldn't have the levels of investment of other customers in terms of whatever it is, marketing resources, programs, coupons. So how are you going to treat. People differently. We have to know who they are. And and so I think that this is a, progression. 

[00:41:18] squadcaster-di5i_1_08-23-2024_121712: The way I looked at, again, being very involved, having built CDPs, being very involved in that world, especially as it's moved to composable and data warehouse. To me, the next progression is okay. Now let's take those, let's take that data set and start to see our customers in a much more much more intelligent way that will drive our business forward. 

[00:41:37] todd-sawicki_1_08-23-2024_091712: One thing we also like to talk about is maybe it's a good closing topic, which is Where do you, what's your vision of the future, right? For first party, for CDPs, for this what does that look like in three to five years? How does the world change? And obviously you're trying to help that with your Neuralift in your efforts. 

[00:41:54] todd-sawicki_1_08-23-2024_091712: So what's the, what do you, what's your prediction for the future of the world that you're living in? 

[00:41:59] squadcaster-di5i_1_08-23-2024_121712: [00:42:00] That's a great question. I think more automation more precision. Faster time to value. I'm very, obviously I'm very long AI. We're doing work in Snowflake with an agentive AI company for activation. I think that's incredibly fascinating. Going from a dataset that is in the warehouse to automated segment and audience creation. 

[00:42:22] squadcaster-di5i_1_08-23-2024_121712: To automated activation and even automated reporting back of measurements. Now you have cappies, right? So you have, you can close the loop on things. So insights, suggestions about, this budget should be increased or, if we increase this budget, this will, we'll have this effect to, conversion rate or, volume or, these kinds of things are all going to be drawn out by A. 

[00:42:43] squadcaster-di5i_1_08-23-2024_121712: I. So I think it's I think it's a really exciting time again. It's only it will only be as good. The thing that never changes. It's just it's only as good as your underlying data, right? Everything you do. It's the old garbage and garbage that never goes away. [00:43:00] But the capabilities is. Especially with higher fidelity data are going to continue to expand and continue to be automated and continue, which will again, make things faster and more precise. 

[00:43:12] squadcaster-di5i_1_08-23-2024_121712: And I think for marketers, that's, those are key things, precision and speed in marketing. Is that's how you win. If you could do something faster than your competition, if you could do something more accurately than your competition you're going to, you're going to, you're going to get market share. 

[00:43:29] squadcaster-di5i_1_08-23-2024_121712: And, and if you know something about, customers before your competition, you're going to get market share. So I think that's going to be really quite something over the next, five years and even, being able to I don't want to say reverse engineer, but gain more insight from the platforms, based on the data that they give you back. 

[00:43:46] squadcaster-di5i_1_08-23-2024_121712: Based on performance, based on the types of, ad groups or audiences that you're and products that you're selling, I think AI is going to help there tremendously too. So it's it's an exciting time. I think I think we're going to have a. [00:44:00] There'll be it'll be rocky because there'll be a lot of folks who will be very scared of, technology and progress as there always is. 

[00:44:08] squadcaster-di5i_1_08-23-2024_121712: But there's no question everything is moving there and, obviously the hyperscalers are pushing all of this as well. So for enterprise this is really important and they'll, because they have the most data they'll stand to benefit the most. 

[00:44:21] tom-limongello_1_08-23-2024_131712: Great. Thank you again for joining us on the middleman. Really enjoyed this and we're gonna have to think about a lot where you talked about it, about the future. I think I need to process it a bit. So I'm excited to do that. 

[00:44:35] squadcaster-di5i_1_08-23-2024_121712: I'll process it on GPUs. NVIDIA will be happy to download your thought. You get the neural link and get to download your thoughts 

[00:44:40] tom-limongello_1_08-23-2024_131712: Oh, now I understand the name of the company. Okay. Very good. We'll see you next time. 

[00:44:46] squadcaster-di5i_1_08-23-2024_121712: Thank you, fellas. 

[00:44:47] ​ 

[00:44:52]