# S3 E5 T60 Andreas - Pentaleap

S3 E5 T60 Andreas - Pentaleap 

[00:00:00]  

[00:00:00] Tom: You are listening to the Middlemen podcast with  

[00:00:03] Tom: me, Tom Limongello  

[00:00:05] Tom: and my co-host 

[00:00:06] Tom: Hey, Scott. 

[00:00:09] Scott Messer: Hey Tom, nice to hear you again and see you. 

[00:00:13] Tom: Um, today is, an episode that I'm very excited about, as usual, but, this one specifically, I really, I kind of feel for Andreas. So we're, we're talking to Andreas Reiffen from Pentaleap. He has been an entrepreneur in this space. He created a company called Crealytics and now Pentaleap and. He admits early in the episode he failed at this idea and he's doing it again, and now he's successful. 

[00:00:41] Tom: And so not only am I rooting for him, but he's fighting a really good fight. And that fight was one that the retailers weren't ready to think about at the beginning of retail media, which was, I have this merchandising strategy of curating products, and then I have. A retail media strategy now, uh, that pretty much is diametrically [00:01:00] opposed. 

[00:01:00] Tom: Like it really has a lot of conflicts to the two and he's fixing this problem and you'll hear sort of the technologies involved and all, and all of that. But for me that's what's exciting about this episode because we run into it all the time in our consulting business. 

[00:01:14] Scott Messer: Yeah, it's certainly a, a difficult piece. and you know, as I always put my publisher hat on here, um, the thing that sort of exploded my brain this week, maybe that'll just be a whole section for us. 

[00:01:24] Scott Messer: Um  

[00:01:25] Tom: Scott's exploding brains. 

[00:01:27] Scott Messer: uh, was. Again, how important relevance is inside retail media? A regular digital publisher really doesn't care what ad shows up in the sports section or the finance section. 

[00:01:40] Scott Messer: As long as it's not fraud or malware. Um, you wanna put running shoes, healthcare like whatever it is, it's fine, go ahead and buy it. But in a retail setting. It's a much more curated experience. And not only is, are you trying to deliver what the user is expecting, you're also really trying to drive outcomes. 

[00:01:59] Scott Messer: So [00:02:00] relevance is important and most likely you're getting paid on a CPC or CPA basis, even internally where it's a conversion and a purchase. So a CPM display doesn't help you that much. Um, and that was what was really interesting about Pentaleap was not only are they thinking about how to. Balance, yield with your directly sold retail media. Ads, but also with third party demand coming in from DSPs like Amazon and Andreas will sort of show us what that landscape looks like. But it also compares the yield of your organic search listings and how well that those sell on the site. So now you can really run a yield experiment across your entire search grid. 

[00:02:47] Scott Messer: And not have the, the two silos of merch and ads, just sort of taking their space and not knowing what the other one's doing. And it's really driving some interesting outcomes for [00:03:00] retailers when you can do this. And that's, that's where publishers live. We are great at, uh, mediating yield across demand sources, but it's kind of a newer thing for retail media. 

[00:03:12] Tom: Yeah. 

[00:03:13] Tom: and I mean this is definitely like a leveling up or a maturing era, uh, where customers that are using Pentaleap are no longer ripping and replacing, uh. You know, one ad tech vendor for, for the next, for me, first of all, just the Rippl and replace, I'm starting to feel like I might have suggested this as a consultant, and I'm starting to feel that like I'm hearing the. 

[00:03:35] Tom: The tape being ripped off. Like you're, we're, we're basically asking the client to be waxed if they're, if they're, you know, you just switch from one ad vendor to another and I, I feel like Pentaleap's approach might be, you know, a little more sustainable. And so I think that that's a, a maturing area where retail media is, is leveling up, but I think it's for the whole ad business. 

[00:03:55] Tom: And, and that's it's always fun to talk with you about sort of where retail's going, where [00:04:00] publishing is going, because. The best companies, the most interesting ones, are doing things that merge those capabilities across, you know, now search is now part of it. And so I think that, that for me is what's, what's exciting, but there's also some, some scary stuff. 

[00:04:13] Tom: And, you know, if you, if you stick around to the end, Andreas is gonna talk about where, uh, LLMs and AI capabilities and how that new surface, uh, might be threatening to retail media. It may not, not, might be threatening, but it is really threatening to retail media. 

[00:04:31] Scott Messer: Yeah, it's a doozy. So if you don't wanna get your front end stolen, you might want to hear Andreas's lesson at the end. Um, thanks for coming on and enjoy the episode everybody.  

[00:04:43]  

[00:04:44] Tom: We are here with Andreas from Pentaleap. Andreas, you're an interesting character. . You work in both Germany and New York. But tell us a little bit about. Pentaleap. How you got there, I knew of you from Crealytics, so tell us your story. 

[00:04:59] Andreas: Yeah, so we, [00:05:00] we started, in retail media, under, my former company Crealytics, a longer time ago actually. It's, we had a plan in 2017 and, the plan was exactly what we are doing today. But we were not successful at all and the whole model failed. So we, we, the idea to quickly summarize that you ask what we are, we wanted to build a mediation platform that basically serves ads alongside organic results and allows retailers to plug in. 

[00:05:39] Andreas: Different front ends, different demand sources to break free, be independent, and basically replicate what we see in the Google ecosystem or Amazon ecosystem where, um, those companies own everything that happens on site and then have multiple companies competing for, for the front end. And, um, [00:06:00] we had several retailers who wanted to do it. 

[00:06:02] Andreas: Um, there was Google who wanted to get into the game and others. And, um, we last minute, um, there's one person who didn't like the model and it flipped in the last second. So what we did, we sunset the product, waited, had a me too solution, and at some point we realized that the market, um, became just much more mature. 

[00:06:25] Andreas: Um, got into the second, um, phase and we said, okay, maybe we try that again. And now we want like, I think more than 30 retailers. Um, we are growing very fast, big companies, so we, we have a, let's say, technology layer that fits every stack. We are not a replacement technology for Criteo or Koddi or, um, Kevel. 

[00:06:49] Andreas: Anyone, anybody else. 

[00:06:50] Tom: So just so the audience can understand, 'cause this is pretty complicated, I think I'm just starting to wrap my head around it. The reason that we first heard of you was, [00:07:00] when another earlier iteration of retail media PromoteIQ had a bunch of retailers, they were acquired by Microsoft. They were starting to, sunset their, their product and just sort of, , they were throwing in the towel and I think Orange Apron Media, part of Home Depot was looking for a seamless way to keep their retail media going. 

[00:07:20] Tom: And that ended up being you, , at Pentaleap Leap. Kevel and Vantage, and my initial reaction was, how can all of these companies work together? And I, I come from this sort of, you know, I came from Quotient, which was very turnkey, very monolithic. We were just doing everything. And so for me, hearing somebody playing nicely with others sounded great, but I was like, how can this work? 

[00:07:40] Tom: So how are you okay with not being a monolith, first of all? 

[00:07:45] Andreas: Um, so, um, you, picked a very good example. And, home Depot was truly an innovator in, in the space here. Um, their situation was they were on Microsoft PromoteIQ, there [00:08:00] were several companies who were not a hundred percent satisfied with, um, what the, the product could do, and hence Microsoft decided to sunset the product. 

[00:08:09] Andreas: So Home Depot needed, um, a solution that worked better for them. And the original idea was basically rip the whole thing out and bring something new in. And I do remember those conversations I said. It's actually not the smartest idea because it's, I mean, it's a sizable big retail media operation. 

[00:08:33] Andreas: They're running and if you rip and replace the whole thing, you have, even if you work with a a monolith and you switch to a new monolith, you have to rip out the ad serving tech. So actually. Um, the tech that is responsible for surfacing sponsor product and putting them into tiles and, and making sure that the stuff is relevant, but also you have to rip and replace everything that [00:09:00] is brand facing. 

[00:09:02] Andreas: And this is where the nightmare starts. When Walmart moved, um, to an in-house solution, it was wild. They, they were, the hotline was busy. People wanted to understand, how can I solve this? How can I solve that? And you have to imagine thousands of brands at once in a, um, in a midnight operation, moving them into a new front end. 

[00:09:24] Andreas: It causes headaches and it will never, ever go smoothly. So my recommendation was. Why don't you take a phased approach? And the phased approach is, um, the actual ad serving has nothing to do with how campaigns are being managed in some front end. It, the front end, um, pushes campaigns to the actual ad server and then the how. 

[00:09:51] Andreas: The website is being rendered and put together. This hapPentaleaps on the other side. So what Home Depot did, they implemented the Pentaleap [00:10:00] ad serving that comes with an RTB technology. So essentially an ad request goes out in that, um, in that instance to PromoteIQ. PromoteIQ, um, responds with a range of potentially relevant products and bids 

[00:10:17] Andreas: We fetch that in real time, but the actual decisioning where which ad goes and which ad gets filtered out is then on the Pentaleap tech side. And once that step was completed, they could. Gradually migrate brands into, um, a superior front end. And they chose Vantage, which is a very comprehensive product where you can run, um, well everything on onsite, it plugs into Pentaleap. 

[00:10:48] Andreas: But you can also run your offsite campaigns on Google and Meta and whatever Yahoo Programmatic and, um, install, um, CTV and so on. And that's, um, [00:11:00] how Home Depot did it. And, um, we just announced a client went today and they did exactly the same thing. So rip and replace is a bit clunky and there's a much smarter way now to just take a step by step approach and sleep well. 

[00:11:15] Scott Messer: It is, uh, coming from the publisher side. It's really interesting to hear this, and this is something I talk about on, on the Middlemen a lot is how. Retail media and, uh, ad serving and RTB, which is a much smaller subsection, is, is one different than publishers, uh, and how much more challenging it is because of the relevancy problem. 

[00:11:36] Scott Messer: Um, and so I was curious if you could talk about like, how relevancy is hard for retailers and like why it's different and how Pentaleap sort of fits into the relevancy model. 

[00:11:50] Andreas: You, got the key point. It's if you have a traditional publisher, there's content on the website, and then there are ads on the edges. A retailer is [00:12:00] fundamentally different because the content is actually products and the ads are also products. And, um, the, like the. Core thinking here is that what retail media has done in the past from a user perspective makes just no sense at all. 

[00:12:22] Andreas: Because what we did is we said, well, we, we push a few ads into pre-selected fixed tiles and this is paid, and then we have sponsored. A user seriously will never care whether there's a sponsored label or something else. They want to get to the right product. And the downside of running this in silos, so the way it works is there's a high-end algorithm that puts your website together and ranks products based on an expected margin per impression. 

[00:12:57] Andreas: So the higher the margin per [00:13:00] impression is you, you compute. The higher a product would rank, and then it goes down the grid in a descending order. And then you have paid, which uses very simplistic algorithms to have a rough idea, is something relevant or not? And you paste that into your website. Now, the problem if a product that surfaces isn't super relevant. 

[00:13:24] Andreas: You don't win the click so you don't make money on the click. But even worse, you also don't sell the product. And now imagine you run a store, whatever, fifth Avenue in New York, and you put your slow sellers into your, um, into your shop window. It will, you make some add money that way. But you kill your whole retail business. 

[00:13:49] Andreas: So it's a retail media. I don't wanna know how much this actually delivers an incremental impact bottom line to a business because [00:14:00] it, on the one hand, brings money in. On the other hand, it drags the conversion rate down. And so the benefits from retail media get eaten up on the, on the retail side. And this is. 

[00:14:12] Andreas: In the very core, the problem that needs to be solved. So we took a different approach here and we said, let's break with that division of paid and organic. And in the end, a retailer wants to make money and a product now can have a retail margin or retail margin if you sell it Plus. Ad margin you make. 

[00:14:36] Andreas: And while the compound margin defines how products rank and if, um, a product is, um, organically listed rather in the, on the, let's say, in the middle of the page, but it has, uh, um, tremendously high bid. It then, based on the same logic, deserves to rank higher, and this is if you operate in in a silo, in swim [00:15:00] lanes, there's just no chance you can maximize the margin your website delivers per page call. 

[00:15:08] Scott Messer: Um, this is me just processing that, uh, I, I followed it. It's, it's super fascinating and I think for like smaller retailers that have a very narrow endemic single focus, it might be easier for them to manage. This process. But I think as you start getting bigger and you have multi-category and you're selling campaigns for different things and you have different performances and you know, a wide range of CPCs that go on across your products and categories because of the price of the products or how popular they are, it gets incrementally more complex and you need a solution. I'm curious though about like bringing in third party demand as well now because it's one thing when the retailer's sort of managing all of this and they have their team on it and there's data scientists and merch teams and ad teams sort of working together. But [00:16:00] then when you wanna bring in third party demand, which is kind of new to retail media, uh, it gets really complex because you're doing it at, in real time, RTB. 

[00:16:14] Scott Messer: Uh, how does, so one, what is Pentaleap's stance in sort of the RTB world and what are you doing there? And then let's talk about how you blend that into, uh, the front end sort of yield framework that you just talked about. 

[00:16:27] Andreas: Yeah. Before I start, maybe let's clarify a common misconception. When people hear about RTB, it's immediately, oh god, like old programmatic, this play world and um, fraud and. Um, and bad ads showing up and, um, we don't want traffic to link out. That's all like simply not true. RTB in the sponsor products onsite retail media case is simply, um, a vehicle, a technology, a tool [00:17:00] to um, connect any. 

[00:17:03] Andreas: Demand, any demand source that can respond to ad requests, um, to, to a retailer's website. So sponsor products show up and they are all endemic. It's just who actually funded that. To give you an example, you have probably read the Macy's news, um, Macy's is, has now connected Amazon. So, um, there's an ad request going to multiple parties to Criteo and to Amazon, and they both, um, respond and, um, then the technology, um, renders the entirety of the grids organic and paid, but it treats the, um, the different sponsor products that are suggested from different sources in exactly the same way. 

[00:17:50] Andreas: It, it built a unified ranking. And it's invisible for the user. So it's not like, um, that any traffic would leak or [00:18:00] would go out. So it's a. 

[00:18:01] Scott Messer: And that ad request they're making is for, in, in like the Macy's example is a product that Macy's sells. So it can't be something that Macy's doesn't 

[00:18:09] Andreas: Exactly. only there's a product feed matching happening so it's only viable for products that are actually being sold by Macy's. So you can play this and this is where, um, we are going with this. You can. Send ad requests to potentially 10, 15 different parties and they all would check, do we have somewhat relevant demand? 

[00:18:35] Andreas: And they would push back a set of, let's say, 20 to 50 potentially relevant products that, um, that we take. And then the decisioning happens. Um, and it's some we don't even show because they're not even listed anywhere in the organic ranking. So the AI that is already operating has disqualified certain products by [00:19:00] simply not surfacing them. 

[00:19:01] Andreas: And, um, and yeah, in essence, that's how, how it works. 

[00:19:06] Scott Messer: Those products are re are returned with bids as well, right? So you can do the, the revenue margin calculation. 

[00:19:13] Andreas: Exactly. So, um, basically compute a relevancy. We get, we get a bid, and this is how the entirety of the experience gets, gets rendered. And the, um, there are two. Mechanisms. There's one that is essential, which is we, you, you will hear like from all the vendors, oh yeah, we have the best ai. And this is like, oh yeah, this is so sophisticated and we've invested in our AI capabilities. 

[00:19:44] Andreas: The truth is. There are a few companies who are really ahead of the, um, of the ad tech companies, and that's, that is, uh, whatever, Algolia or that's a Google onsite search. That's companies that, um, yeah, [00:20:00] their breaded butter business is to deliver relevant grid of products that users will buy. 

[00:20:06] Andreas: So our tech is built on that intelligence. And doesn't happen in a silo. So if you like, the search product, inherit information to us, we capture that and we make use of it. If products are not contained in the organic ranking, we can still show them and then we need to apply our own algorithms. 

[00:20:28] Andreas: But that's. Not the norm and it should not be. And this is the, the big shift we are seeing before we lost a lot of business because retail media said, well, but I have advertisers who want to show that product. So we have to show it. Yeah, you can. But then you cannibalize your retail business and you cash in a little bit on the advertising side, but you annoy people. 

[00:20:52] Andreas: So that's no longer the the way to go. 

[00:20:57] Tom: what was interesting there, and this is [00:21:00] something that Scott and I have talked about a lot, which is you work with, uh, either a retailer or a publisher and they're trying to consider what ad server and potentially order management system that they're gonna use. And generally the search provider could be Algolia, like you said, or somebody else. 

[00:21:16] Tom: We're always kind of like, well, why aren't they do, why aren't they in this game? it sounds like you're one of the first companies out there that is holistically looking at all of the sources and, and potentially the best source of information for ranking, um, as part of your core offering. And so that's, that's very interesting. 

[00:21:31] Tom: Now, does that solve the problem of the fight between the merchandising team and the paid retail media team? Or like how do they work with your technology? Is it, is it sort of automated or they still have their hands in there? Or like how does it work? 

[00:21:45] Andreas: Quickly to your first point, why aren't those companies also doing that? Um, the reality is, um, there is one who has started, um, doing exactly that because they have the intelligence, they are [00:22:00] very good at, um, at also we are providing a unified ranking for the very same reason. So we approve. 

[00:22:07] Andreas: It makes a ton of sense. And there. We'll be very good at exactly that. There is one key difference though. Um, we've built our stack to be a front end agnostic so we can work with any front end, but we are also search platform agnostic. So we see, um, among our clients, search providers are being replaced. Once a year, probably all the time. There is some change if you build your whole media business, um, on. Exactly your search platform, you will experience a big hassle to again, rip out the whole thing and replace it by something else. And are trying to leverage the knowledge independently from which search provider that is so that you can [00:23:00] flip well our product, but also the search providers without, um, having an issue to your second question. 

[00:23:07] Andreas: Sorry. 

[00:23:08] Scott Messer: Hold, hold there. Keep that. I just wanna highlight for our, our listeners, like what an important concept that is of not inter tangling your technologies and your stack. If you want to build in the best way for your future, it's a composable stack where you can take. Different pieces and swap them out as easily as possible. 

[00:23:28] Scott Messer: I mean, it's still hard to swap some of the individual tech, but it's much harder, sort of like in your Home Depot example, when you have to rip and replace the whole thing in order to change one piece. Uh, so I'm really liking this solution that's sort of fitting here. And Tom and I, if you, if you're watching, you're seeing a smirk because we hear this problem with a lot of retailers, um, and it's sort of, I think the. 

[00:23:52] Scott Messer: One of the most advanced challenges and that you, your retail media business, has to get to a certain point when [00:24:00] you realize how much more money you could be making or how much more easily this could be operating for you. 

[00:24:07] Andreas: Yeah, absolutely. There's so many advantages, of not having everything in with one single company in one stack, even. Even if companies are modular, I'll give you an example. If, if there is someone serving ads, but also providing a front end and making money with that front end, um, it's very unlikely that that company will come to you as a retailer and say, Hey, you know. 

[00:24:34] Andreas: I've seen now there is whatever, um, some other company, X, Y, Z, and they have now a much better front end. They have connected this to all different channels. Why don't you switch over to them and you will also even pay less. You will always try to get those retailers to stick with your front end because you make money that way. 

[00:24:56] Andreas: We made a conscious decision to. Be [00:25:00] seriously agnostic. There are companies who use our front end just to steer onsite campaigns, and we did that because sometimes we need it because retailers don't wanna hire another company and they may only want to do onsite, but what we tell them is. Well, as a fallback, you can use our front end, but if you would like to do a full funnel play, there are so many good companies now go to them. 

[00:25:29] Andreas: So we are like, it's a strategic decision we made to, we made to not do that. And yeah. Um, 

[00:25:37] Scott Messer: And I think I interrupted you on that second one, which I think is how you're talking about balancing the merch teams versus the ads teams. 

[00:25:44] Andreas: It, it wasn't as difficult as you might think because, if you roll out our tech, you will see. Benefit on both sides. Let me explain. So before there was a like fight, [00:26:00] internal fight and horse trades being made between retail media, give me more inventory um, and, um, and the merchant organization know, I I I'm incentivized by selling products and retail margin. 

[00:26:13] Andreas: And this was a like. Big friction we saw. So how we do it as we, um, usually come in and layer our tech, um, in conjunction with, um, incumbent ad servers on top of those ad servers. We run tests and we've done dozens of those tests and we, um, we take a look at different KPIs, we take a look at. Overall, what is the impact on the compound margin from retail and advertising? 

[00:26:47] Andreas: We take a look at what happens with retail margin, site conversion rate, and we also take a look about, um, on like ad revenue, how does it scale? And what we've been seeing consistently [00:27:00] is that even if we show more ads. The site conversion rate, and that's back to your point, rather on the merchant side, increases even with more ads. 

[00:27:11] Andreas: And there's a simple reason for that because previously you might have operated with, let's say, six different tiles, but with highly irrelevant products that had an impact on the overall conversion rate. So now as this is. Basically delivered by the same high-end AI that works in the background that we, um, make use of the overall site conversion rate. 

[00:27:36] Andreas: Um, has a tendency to slightly go up. So the merchant organization is fine with that. On the ad side, what we see in the tests as we run those AB tests, we have a range. I think the lowest improvement was 66%. The highest was 154% in ad revenue increase. And the simple driver [00:28:00] is number one. 

[00:28:01] Andreas: Clickthrough rate. Um, and, um, we price this, um, with a CPC and not a CPM. So high clickthrough rate means that sheer traffic volume increases and, um, and so there is a positive, positive impact on the merchant side and a positive impact on the advertising side. So those. like fights we saw previously, they stop quite quickly when we rolled this out. 

[00:28:31] Andreas: There's another impact. Usually before there was always complaints. Oh yeah, why do I see this product here? The user search for red dress and this is a blue dress, and how can that god. 

[00:28:44] Tom: Yes, the variant discussion, something we're having right now with a client 

[00:28:47] Andreas: Yeah 

[00:28:47] Andreas: there are edge cases where this still happens, but our answer is, well, your onsite search platform delivered that product. So the [00:29:00] philosophy is no longer, um, you have to fix your search and you have to fix your retail media relevance algorithm. Focus on getting your search right. And you will pull over those improvements, um, to the ad side. 

[00:29:13] Andreas: So it's, it's a streamlined focus on better relevance and you make use of it Asana at a different place. 

[00:29:22] Tom: So I've got one more for you. Search providers are getting swapped out because there's a lot of challenges and changes in search, and it's becoming more conversational. Is this model gonna work in that new world? 

[00:29:32] Andreas: I think there are two, two thoughts on this. The first is if we get into a game where retailers lose their front end to an OpenAI or other AI companies, Gemini. Which by the way, is a huge question mark because many have tried to steal the front end from retailers. 

[00:29:55] Andreas: It's, um, social platforms and um, and Google tried with a [00:30:00] marketplace model and nobody liked it. Retailers pushed back really hard because it degrades them to become logistics companies, but if this happens, retail media is dead. And not only that, like retail is dead because retailers are no longer retailers. 

[00:30:17] Andreas: The, it's just a logistics play in the end. If, if everything happens from the agents, will this happen anytime soon? It will not happen anytime soon. So we, we will stay in the game. Users won't change their behavior immediately. So the second scenario is, and we see this, I just read, home Depot launched a new AI agent that is a Home Depot specific one. 

[00:30:39] Andreas: Then we have another client. They also launched their AI agent and with the very same methodology and we start doing this, um, with one of our clients of. Basically plugging into the intelligence from that AI tool. In the end, this AI tool has a prioritized list of [00:31:00] recommendations again, and it's not like just two, three products and you can, um. 

[00:31:06] Andreas: Merge bids with that logic. And if product five and six are similarly relevant and you're almost indifferent, you can change the order and give the one product with a bid, some more leverage. And that's why I don't, I'm not concerned about that. We, the, the only thing we need to see is how FAST AI gets adopted and will those. 

[00:31:33] Andreas: AI agents, external ones, start taking over the front end, and that's more than a retail media disruption. That's a, that's a very big thing, and it's not gonna happen that fast. 

[00:31:47] Tom: All right. Well, we really want to thank you for, uh, taking some time with us. Um, that was a fascinating discussion. And, uh, yeah, we look, look forward to hearing more updates. You guys are doing some, you guys. Some big news recently, so [00:32:00] congrats. 

[00:32:00] Andreas: Thank you so much for, um, for the chat and for all the questions. Really enjoyed it. 

[00:32:05] Scott Messer: Welcome, Andreas. Thank you very much. Have a great day, and watch out for the Middlemen. 

[00:32:10] Andreas: I will. 

[00:32:11] ​