🚨Tactile & Loomit: How Top Games Make Money From Ads

Yeah, I mean, it's a lot of AV testing. I think the starting point is usually the segmentation. Trying to to look at the different segments, audiences. So, the the how long you have been in the game without converting, your geo, platform, those kind of things. That's one one aspect, and then the other aspect is what do we want to show? Like which ad placements we want to use?
[music] [music] [music] [music]

[music] [music] [music]
Hello everyone, and welcome to a very special episode of the two and a half gamers podcast, where you can stay two and a half steps ahead of the mobile gaming industry. Today, like I said, we have a very special episode cuz we have two wonderful guests that are joining us to talk about Lily's Garden and their ad mod strategy. So, before we start, I'm Felix Broberg. I'm Yakub Remiar. And we are your hosts. Welcome so much to the show, Eric. And nice that you guys want to give a brief introduction about your for yourselves. Uh I'm Eric, and I work at Tactile Games, and I am yeah, mostly focused on everything related to administration in Lily's Garden and in the whole

portfolio, so we don't only have Lily's Garden, we also have a couple of other games that are both in-app and administration focused. And we are using yeah, Lumit. Thank you. I'm Ignasi Ortiz, I'm co-founder and co-CEO of Extreme. We are the makers of Lumit, and for those unaware, Lumit is a ad mod strategy layer that sits on top of, and I would argue turbocharges your mediation stack. That means delivering anything from dynamic floors, advanced segmentation techniques, swappable formats, and in general, anything that you need to get to user level, maybe even request level optimization for ads. Cool. Welcome so much to the show. I'm quite excited about this because Lily's Garden is one of those match three games that has ads and properly has a real ad mod strategy, which is pretty much you guys and King, so I really am excited for this. Well done. I think we'll start off by just doing a bit of a review and basically have a look at Lily's Garden, what it looks like for our viewers, and then basically maybe Eric, you can do some commenting and interrupt Yakub if he

gets anything wrong. Yeah, I think we we covered it at some point, or at least maybe briefly, but I don't think so in the separate episode on the podcast. So, Lily's Garden, I think it's especially in our circles mostly known for their very very signature creatives that a lot of companies took pretty much into their portfolio account, but as for the games itself, it's click to match genre, so not match three, but click to match. Uh you play through the levels and slowly you know, move on with your I don't even want to call mansion because again, it's more of a garden as as the name stands, and you slowly of course unfold the story, which is quite important in this case, as I said, directly tied to the creatives of the game. There's lots of twists and like love triangles and stuff like that. And then of course, there's pretty robust live op system that's tied into it and stuff around the usual kind of life cycle of a match Not sorry, match three

game or the puzzle game that I mean, but that's there. Anything that I missed, Eric, as for the very brief intro? No, I mean, we usually refer it as a more of a blast game, and yeah, with puzzle pretty nice uh pretty nice story and some decoration elements to you know, pretty similar to the the scape Garden Scape, Home Scape titles that you have on the market. And yeah, it mostly focus on on in-app, and indeed ads come on top of it. I mean, we'll cover this in more details, but yeah, it's pretty it's pretty robust in terms of live ops, and uh you have lots of story to to uncover. Yeah, as I said, like one of the things that really comes to mind when talking about Lily's Garden is its signature creatives portfolio, because I think I literally saw some videos where like literally people are looking through the whole story as it goes from the start to

the end, because it like there's literally a story behind the creatives, so that's like one of the major innovations I guess that the team did there. But yeah, anyway, like good job on the game, definitely. Yeah, yeah, I do remember the creative. I mean, maybe from four years ago that were yeah, a bit edgy made us a bit famous. I think there were some New York Times articles about this, or like some major publications talking about this. Yeah, go ahead, Felix. I [snorts] think from my perspective, right, there's a lot of games in this particular genre that don't like ads, but basically you guys have quite a lot of ad placements. Like how does Tactile think about ads, and how did you guys decide to roll out rewarded ads in Lily's Garden? Yeah, I can confirm like setting up setting up ads was was and still still is a challenging. We come from an in-app heavy world, and we definitely want to preserve that. Well, and it's it's for obvious reasons, very high CPIs, relatively long recoup window, and uh

you know, you have a lot of competitors that prefer not to have ads. I think you mentioned the only ones we have out there, like Candy Crush, but there are actually other ones like Twin Blast, Matchington Mansion, or Project Makeover. So, you still have a few major players that do have ads. So, it can work, but as with yeah, most things with mobile gaming, it's it's a lot about AB testing, obviously. And yeah, thinking about what what defines a non-payer in your game, because at the end of the day you have a lot of users that do not convert, do not buy purchase in-app in our in our game. And there are a couple of parameters, aspects that you can rely on, like like level progression, time since install, time since the last transaction to to determine what makes a what a non-payer. But yeah, I I don't have a secret sauce to to share, but in our case, the the the main focus and the main thing that we try to do is how can ad revenue

bring incremental revenue on top of in-app given the pretty wide non-payer user base that that we have. And how do you test that? Do you basically just run AB test, or do you look for different things to basically decide when you start showing ads? Cuz yeah, it's a bit different from company to company. Yeah, I mean, it's a lot of AB testing. I think the starting point is usually the segmentation, so trying to to look at the different segments, audiences, based on pretty pretty um um basic aspects that most publishers know, so the the how long you have been in the game without converting, uh your geo, platform, those kind of things. That's one one aspect, and then the other aspect is what do we want to show? Like which ad placements we want to use? What do we want to offer? And then AB test it. Uh one of the main concern is

obviously in-app cannibalization, and and a big challenge is how long do you want to test those to make sure that things are statistically significant. It's always it's always a challenge. Like I would love to run a six-month AB test, but we don't have that much time. Which ad placements generally work the best in this category? You don't talk about Lily's Garden specifically, but in my experience at least, it's usually the the pre-level boost and the post-level boost tends to be like majority of the impressions. Is it the same here, or Yeah, I mean, those places are where well, users they will end up in those in those screens, in those events quite often, so we try to focus on places where we can get high ad engagement and high frequency to really maximize the potential. So, we usually try to use three different pillars on on RV placements. Uh they have to be easy to discover, easy to to understand. And then they have to to really bring value, so they they are appealing to users, so they actually want to engage

with it. And also, they have to be really really rewatchable, so there is a frequency potential, but long story short, they have to be bringing engagement and frequency, and the ones you mentioned are are obvious choices because you get those quite often. Depends what you give away for an rewarded video placement, but yeah, usually you can bring value to users in those places. But at the same time, they bring a lot of in-app, and they are good conversion triggers, so you have to find the right balance. There are obvious places where we make a lot of in-app and so, yeah, we have to be extra careful in those places, so AB testing and and monitoring the the game economy well enough to make sure that it's it's bringing incremental revenue. And I know you guys have like been on pretty much tested a lot of the mediation platforms out there, and I guess we can tag in Ignasi you but like what's your experience with the different mediation platforms and where are you currently mediating? We have tested used all the

big ones, so Google AdMob, Level Play, and in Max. And that's Well, in the past we would do a switch from one to the other, and obviously it wasn't the best way to do it because you would just compare things before and after. And the incentive package would definitely help take the decision. But with Lumit, and that's how they How do you say? Put their foot in the door by with them the AB testing capability of mediation. And with their tool we were able to test two uh two of the big mediations and and what's the upside on on ad ARPDAU, eCPM, but but also check other other metrics like ad engagement, frequency, retention, and other game engagement metrics that are quite important because developing an internal tool to AB test mediation seemed a bit too complicated. That's not really something we would use on a

regular basis, and that's why we we went with Lumit to to AB test the mediations. It's actually not the only case where people come in just because of that specific feature, which we know is attractive because it's just so hard to do, and it causes such a big headache for for studios. And then of course our job is to show them that we can do much more. Like even going back to the very quickly going back to the cannibalization that we were just discussing, I thought it was interesting because we you know, we we actually just had a client reduce interstitial ads for 10 for his 10% most valuable user base. And of course he got killed on on ad ARPDAU, but he actually almost made all of it back through the increase in conversion rates to to paying users and how much those users were spending. The cannibalization of course was there. He couldn't make it all the way back, so now he's, you know, testing more segmented approaches and seeing if he can actually make that work. But you know, there's a lot of talk about how these things people think happen, and

then it's interesting when you actually do a test and you actually see things, you know, we got killed on the advertising side, but IAP just went extremely high. Like incredible. So So yeah. And how did you think when you were building this product, right now, Narsi, because you guys have your roots kind of in gaming as well? Like which product problem do you say you mainly you solve when basically you allow a publisher to do an AB test inside of Lumit instead of doing it on their own? That's a good question cuz the honest truth is that when we started, we didn't set out to to fix mediation AB testing. We actually set out to Well, we at the time we thought we were going to fix mediation, right? We we launched with a mediation. We got into a lot of trouble by launching a mediation. A lot of fun. But the truth is that we quickly realized that the battle wasn't necessarily there, and there's a lot of reasons why launching mediation is extremely difficult. But we did quickly decide that the reason that we wanted to launch mediation was that there was a lot to be built, right? There's a lot of I mean, you discuss this every single week. Like all the things that you can do, all the

strategies that you can implement, many of which require coding. And we said, "Let's just build a platform to allow all of that." And it's what we have built internally. Given that we were not going to be using our mediation as much as we had initially expected, we quickly had to find a way to use more than one mediation within the same app. But we still do this. Like you can of course there's there's terms and conditions that you need to comply with according to the partner, so not that easy to to implement, but one of the many uses of the technology we built is mediation AB testing, which is ridiculously simple to to do on Lumit. Literally a toggle. You go back and forth. If the adapters are in place, there's absolutely no more code, and the data is just right there. So it's it's very easy, but then you get into, you know, segmented mediation, some talk of prebid now. You can actually do that as well. So all of these things are are easy to to do running the same technology, which is just how you manage multiple mediations within the same app. Do I get that right that it's like I think like 60% or 70% of the usage of

the platform then because especially in this scenario where like IAP heavy game putting ad placements, which like I guess cannibalization is the main kind of concern. Like everything pretty much needs to go through AB test all the time and forever. Yeah, everything we do needs to be AB tested, and I'll take it a step further. I don't want to say declared war on AB testing in general, but I think it's it's improvable. So on one hand, everything we do, not just mediations, need to be AB tested. But then on top of that, we we were sitting down last year at a meeting, and we realized that when you look at AB test in general, like you're just running a mixed group of users in one bag and a mixed group of users in another, and you pick out the average, and that's put into production, right? And And then that works because in general you will increase whatever KPI you are targeting. But the truth is that within both groups, you have subgroups where maybe the feature that you're testing is actually detrimental to performance. So we initially started doing lookback

windows, 14 days, figuring out the subgroups where you should actually not use the feature. So let's say AB test comes out, you need to turn a feature on cuz you tested it, it works. You do the lookback, and you say, "Okay, but we turn it on except for these small groups, right?" And we were doing that. And that has its own problem, which is it's static. Because you are looking at what happened back when, and then Eric came up with a new creative, and all of a sudden, you know, he changed the UA campaigns, and then it changes again. So actually literally last week, and and we're just starting to talk about this now, we started releasing you can think of it as an always-on AB test. So you decide the feature that you're testing, and then in real time machine learning models will decide whether on a user level basis that specific user should be on or off. So you test a feature first, but then you decide the effective implementation of the feature is done at the user level. And you actually turn it off more than you would think. So anything you turn on is likely being detrimental

to some subgroup of your users. So something like smart auto segmentation or something like that. Yeah, we we we've called it so many things. It was like hyper segmentation. No one knows what we're doing. We call it So I mean, we'll brand it at some point, but yes. Nice. And Eric, you said that you basically you started with this AB testing feature, right? And which other products did you end up using cuz you said they got their foot in the door, right? Yeah, going back a bit, we used the the mediation AB testing tool because we didn't want to develop it internally cuz we're not an ad-focused company yet. So for now the decision process of when should we build or buy {slash} pay for a solution, in that case we lean more towards paying for it. If it brings consistent uplift that will justify the cost, we're so good. But I would love to hear other companies and bigger companies how do to know how they approach it. So yeah, in the case of mediation AB tested, it didn't really seem worth it given our stage and and size. Maybe my CTO will have a different

opinion, but but then beyond the mediation AB testing tool, they have other other things that that bring value. First thing they you obviously have to to add their SDK, and and then you will get a lot of data visibility, and it will be a good way to challenge a bit on how you look at your metrics from ad engagement to ad quality. Then I think we really needed to catch up a bit on ad monetization, catch up on all the best practices out there such as what you mentioned like multiple ad units to optimize our ad auction. So that's one one thing we've been using. And then Lumit also has a few things in the background to optimize the ad caching {slash} ad loading. So so the smart buffering so we don't miss on any impression opportunities. And I think overall it's also great to have a point of contact with Lumit, an account manager to to exchange

objectively on what we should focus on as a publisher to to improve the ad monetization. So yeah, we really benefit from from them developing features for the industry as a whole. And hopefully get more clients so you can get economies of scale, and and we can enjoy lower fees. He's negotiating fees live. This is This is I'll take it. I'll take it. He's Yeah.
And Narsi, what what does that actually mean, smart caching? What does that actually mean? It's That's also funny because it started off basically meaning that the buffer looks at the device quality to decide how much resources we can actually pull without affecting game stability. Essentially, once you turn it on, you don't want to miss out on opportunity, so you're optimizing for price, right? So you're loading several ads. You're maybe firing multiple requests. But then you can tell the tool to not do this if the device has certain characteristics which you think will lead to instability. You can do it for it. You know, you can set rules where it will behave differently however you want. The most commonly used

one has to do with stability. But that said, you know, that was back when we called it smart buffer. Now we're talking about the smart smart buffer. It's getting a bit Increasingly, we feel that ads are not a commodity, and there's a lot of talk about ad quality and what you can do there. So like there's a lot of variables that I believe clients would consider when deciding what ad to show, right? And so we're trying to see what we can build in that regard such that if you have more than one ad loaded, then it might not just be a matter of just show the highest price one. Because if you like we we don't render ads, we don't know what's going to be shown, so that's very hard to control for. But if we could, then I'm sure that clients would love to say, "Show this ad instead of that one," right? So short of that cuz we can't build that, at least not yet, we are looking at different ways in which the the buffer itself can be more than just, you know, load multiple ads and you're going to show the highest price. So, if I'm understanding it correctly, right? Just for the listeners, right? So, basically, a big issue when you're always working

with ads is that some of your users, particularly on Android, have not-so-great devices. And basically, if they crash or have an ANR because of ads, that ca- counts towards your overall, I guess, organic score and how many organics you get on Android. And basically, a big issue is always that if you load ads on devices, then essentially, that causes crashes. So, if you can exclude them or even basically do networks that have more stable SDKs, that essentially means that you'll get it back because of organic downloads, correct? That's exactly right. And I'll add one more thing. When you increase the complexity of the setups, you're of course potentially adding latency, right? Which also becomes an issue if you get to the opportunity and you don't really have an ad loaded. So, the buffer, to give you just other use cases, it also allows you to trigger a very simple auction as soon as the session is started such that you actually have an ad to show if you need it. And in the background, you're running a more complicated setup to get a potentially higher price and then you

can show that if you haven't shown the the lower price one. But that ensures that you get the first show, for sure. You're not You're not missing out on opportunities. Same thing goes with Like with multiple ad units, if you want to remain compliant with Google, you can't be calling them many times for the same request. So, we can actually split the request in two, call them once, call the every- everyone else through multiple ad units, and you can run a strategy which is compliant but also getting you the best price. So, the smart buffer was was not necessarily considered the cornerstone of the product, but increasingly, it kind of touches on a lot of things that we do. Interesting. I'm just going to go back a little bit, Eric, and and ask you, like, how do you guys think about ad quality? Cuz it's like, you know, it's a race to the bottom and ads are only getting worse and worse in our industry. Like, how how do you guys think about ad quality, especially if you moved between all the mediations? I'm sure you've seen worst of the worst. Yeah, it's especially important for us coming from a in-app focused world where we really focus on the on the user experience. So, it's it has been a challenge and it's still a very big challenge,

especially on some mediations. We try to be as proactive as possible by by generally ask the ad networks to use their more conservative templates. So, we have decent ad quality and lower ad duration, not the most annoying end cards possible, and and try to make it uh try to make it on par for every network that we work with. But it's it's a constant thing to look at. So, we're also quite reactive with it with obviously a lot of user tickets. There are a few tools out there that are that are great. I mean, I have a source of quality to look for the ads that are problematic. Also, on LuMit, you have a few metrics such as ad duration to to monitor that. Actually, the parameters and the that you ask the ad networks to comply with, do they actually comply with them? So, yeah, it's a little bit of proactive

measures by asking networks to use conservative templates. But unfortunately, a lot of reactive things on answering user support tickets and blocking uh inappropriate ads or ads that are just user cannot close or things like that. That's it's a challenge because every ad network works with it in their way. So, sometimes you can you can use their platform. Sometimes you have to reach out to your to your account manager, and you can block specific creative, or you can only block the whole app. Um yeah, I know there are a few solutions out there, but we haven't found the silver bullet. I guess the the the main issue here is basically like ad networks benefit from having really long end card templates because that's how they earn money, right? On conversions. So, the issue you find with all networks, right? And this is the annoying thing that drives me wild, right? Is basically, you can ask

to be put on conservative templates that are, you know, 30 plus five seconds on rewarded, and then you're happy. But then, usually what happens is that networks always update their templates to change them to basically keep them fresh for users to keep clicking. Usually, when they do AB test on their creatives, what they end up doing is basically just picking random games or apps that they want to test it in, right? And then, if you happen to be in conservative and they're testing a more aggressive end card, they don't care. They just roll it out, right? And then, if you don't say anything, that means you basically are now no longer on conservative. You're on like another thing. So, I guess in our situation here, like, yeah, tracking or even giving this data on ad length is very good for basically publisher. revenue per second data point is one which people tend to love because there's also a matter of, "Okay, fine. I'm getting the highest price." But when I actually look at how much of the session time is being spent watching ads, then maybe the ad revenue per

second is not necessarily the highest, right? And what are you missing out on? What would the user have done had he not been forced into a 90-second interstitial? So, yeah. Yeah, and to add on top of it, if there are any ad networks watching, please give us the ability to AB test these two things. One, the template. I would love to test those templates to actually justify using them because they bring higher revenue. And and the second one is um competitors blocking. Um yeah. It's just it's a matter of philosophical reason now where we just do it without being 100 100% sure if it's better for us. But yeah, it would be great to AB test those. I don't know. Like, it always in my experience when I'm testing the most aggressive templates compared to the least aggressive, like, the ECPM loss is maybe 12%. So, that's not revenue, right? But actually, what happens in the end is basically your retention goes down so much that it's never worth it to be on

the most aggressive one. And it's really painful if you're a small developer cuz then they're never going to change it for you. They always used to ignore you, right? So, yeah, it's a real issue.
see cases where the the conservative ones are so like shorter enough that you're actually getting an extra impression at the back of the session, which essentially means that, like you're saying, in the naturally, you're actually not going down. Or you potentially you're not going down. It really depends on how how it's based. Yeah, it's it's definitely a challenge and one which it's it's tough to convince networks to to change because that's, like you said, Felix, that's how I make money, right? So, which also speaks to the need for a I would guess a neutral party to manage or audit how ad quality is working. Yeah, I guess like one thing we've actually been doing is basically for some clients where it's like high IAPs, we just call an interstitial for the first couple of times in the rewarded slot. Take the revenue hit, but basically, we see that it creates like a better feedback loop cuz it's shorter ads. So, people tend to watch more ads overall. But yeah, it's definitely an issue.
Yeah, I know. Look, or even segmenting

mediations. I don't want to touch on any any specific provider as being better or worse. But you could definitely decide to send your traffic initially to one mediation and then switch to the other. Especially like in a case like Tactile where they're not overly reliant on advertising, which means that they don't overly rely on on certain types of, you know, optimization UA campaigns. You can essentially not send 100% of your traffic into the same mediation all the time, and that will potentially have an impact. But it definitely the swapping formats into the video is is powerful. We've seen that as well. Yes. Yeah, definitely swapping format in a second that makes sense and it's great for the UX. And if you set it up uh in a smart way, you can actually improve a bit your impression per DAO and overall be better off. Yeah, I would love also to do different mediations based on on on user level or the the the UX that you want to have. It's just that you need to add so many SDKs and the production team when you add like 20 MB,

25 MB to your game, they start becoming angry. Um but yeah, we would like also to to be a little more sophisticated on the bid floor part and and do it at the user level. And I think that's something we'll we'll soon test with LuMit. I believe there is something cooking. Yeah, yeah. We it's it's it's interesting. Now, there's a lot going on, right? AdMob with key value segmentation. And then, whether you have or don't have access to to managing client prices, demand price floors on the client side. So, there's a lot of things going on, but definitely, I mean, Felix, you've said this again in past episodes plenty of times. The right the smart use of multiple ad units or dynamic floors or however you're implementing just is is a powerful tool to improve performance. Yeah, absolutely. Like, I don't know if you're able to go in and maybe just explain the concept a bit, Eric. You don't have to explain how you're setting it up cuz I know that's your secrets. But basically, if you can just explain like what it actually is for the audience. Uh explain what, sorry, Felix?

Bid floors and multiple ad units. Yeah, we we're trying to optimize the ad auction, right? So, with bidding, uh we we ended up with very little control, and hoping for the best. So, it used to work like you send a request, and then the bidders determine what's the highest ECPM for for that ad request. It's it's convenient, it's automated, but then you don't have a lot of control on the ad auction. What bid floor does, and coupled with multiple ad units, it it it tries to send multiple ad requests with different bid floors. So, it's kind of waterfall with bidders. Uh so, you send first request for instance at $100 eCPM and then if you get an answer, great. If you don't, then you go down the chain. And that's that's what we're trying what we're doing with with Lumit and the

multiple ad unit solution is to yeah. Request a couple more ads to to to find the the best value for for the those users. And can be geo and platform specific. Um, and it can help improve my eCPM quite a bit. I'll add to that one one another not so minor thing which is like what when this got started and I agree with Eric we were very much in favor of waterfall. We struggled with the move to bidding just because the optimization windows help studios I think in general. Now if you if you run the multiple ad units just like plain vanilla set up set them up and you and you just call them from high to low, you will eventually run into trouble. The reason being of course your high priced ad units will be filled less and at some point networks will consider that you're spamming them. So mediation particularly AppLovin has been vocal about asking clients not to you know have very low fill rates on their ad units which means that you can't always call the high

priced. So part of the trick here is having a model that allows you to decide when you call the high priced ones when you don't such that you get to let's say a 30% fill rate on at least a 30% fill rate on all of the ad units that you have. There's a limit to the ad units that you can have so that's 20 which means 19 cuz you have a backfill. Um, and that's essentially a trick. When when you're running the price floors on the mediation side you need to know what price floors to set and then you need a model to decide when you're calling each of them to get the highest potential price without getting into trouble. Not to mention the two things I mentioned earlier, right? So you need to be compliant with Google cuz you don't want to leave Google out of the play but you can't call them more than once for the same impression. And then you don't you want to make sure that what you're doing isn't so complicated that the latency adds up when you start missing out on a potential impressions. So that's where the buffer comes back in and you get the fast first impression. So so I think that that's the the first step. I think the evolution slowly is becoming the client side decide the the floor the floor, right? So I don't need

to have multiple ad units with a lot of price floors. I just need to have the ability to send a price with the request the price floor with the request and that effectively becomes having infinite price floors because you can just set the right price floors. And again the trick then becomes how do I decide how many times I can call them with what prices such that I comply with fill rates and I get the right prices and so on. So there really is a lot of tinkering I guess is the word. Sorry? Tinkering? Yeah, there is a lot of tinkering which is I mean if possible best left to us. I hope Eric can enjoy the fruits of our labor without needing to kind of spend a lot of time figuring all of this out. But it there's a lot going on. Yeah, it was good to start working with Lumit on this because yeah, we wanted to catch up with with those best practices. Uh, there were a few pitfalls that we didn't want to fall into the 30% rule from AppLovin. Those kind of things that you know it's an industry where you you don't ask you

don't know. And yeah, knowing that we have this limit of ad units and and things like that when the collaboration was pretty good to to catch up a bit faster and and get the the uplift that that we got with the with the tool. This is the part which I really hate on the industry where Felix always talks about this and like how do you know this? Oh, you need to ask your rep. Yeah, that's how you know it. Yeah, unfortunately we're probably in one of the most closed industries in all of like in one of the all of the world, right? But still final question right here because basically anyone who just listened to this bid floor talk and basically don't understand what's going on, rewind and re-listen to this because this has been the best AdMon improver or like in hack so to speak of the last two years, right? Setting bid floors and multiple ad units. So if you set this yourself, right? Ignacio where basically you're setting one ad unit and you're calling a second one, what is the issues that usually publishers run into of basically just having that instead of a multiple ad

unit format with more than two? It's really hard to find out what bid floors to use and when and basically yeah, it's really hard to actually get the uplift but the uplift is there, right? E- exactly. I mean like I said there's several issues that will come into play. First is you need to find how to best use the ad units that you were given. And what we find is those that typically start doing this in house will for understandable reasons probably not use all 20 ad units. And if you give me 20, I'll use 20. If you give me 100, I will use 100 because there is incremental value to use with that granularity, right? At at some point look when we were working with waterfall and you'll still understand why the analogy what we were doing is we were predicting the price at which we thought an impression should sell. We would call that the heat map and then we would we would set the calls around that heat map such that we would be very granular trying to get the very last penny out of the networks. And

then of course you have a backfill. You can't take it to that level of granularity with only 20 potential units. Sort of like work into what how best to use the prices that you have available and then kind of manage around when do I actually have to call the high priced ones? How high do I need to go on the highest price ad unit and then how often will it be called? And how do I make it so that that is actually converting into the highest potential ad revenue offer for us. So it does take a lot of tinkering over time. You know, you setting it up the first time takes a lot of work but then I think the real work is in figuring out how to to evolve that over time so that you are always kind of adjusting even for changing seasonality, changing your whether your UA has changed and that changes the type of users that you get because all of this will have an impact on fill rates, on price and so on. So there's a lot that goes into getting that right. And then like I said there's a lot of invisible complexity into making sure that you don't you know run into trouble. You said during one of the last episodes how scale tends to deliver

almost a 10% lift because you get to know things which maybe others don't. And I think that's part of our job. Like we know things because this is everything this is the only thing we do. We work with a lot of people and we make sure that our clients can always test the latest trends as fast as they can without getting into trouble or at least you know, where we can. There's in addition to being a closed ecosystem I think there's a lack of explicit regulations which at times make it difficult to operate. So we get a sense of what people are doing and saying but that's it's a very nebulous ecosystem. So scale gives you leverage in this ecosystem.
Exactly. Which is very I think on that note we should end here. This has been another fantastic episode with a lot of insights. Anything you want to add there? I I want to add one final comment going back to something that Eric mentioned which I think is interesting and I apologize for the for the interruption but he did mention that he wondered he was working with us because build or buy, right? And and it's a very common conversation to have.

My I I would say two things on that front. First is I don't think tech is a moat anymore. I think AI has just completely obliterated anyone who thinks that technology will be a competitive advantage and I think many large studios used to have ad tech as a competitive advantage and we've had pitches with ridiculously large studios that dislike what we do and they told us to this to our face because we are giving out the technology that used to give them a you know different way to work and then exactly. So tech isn't a moat for us either. The only reason we exist is we will need to grow at scale, bring fees down and and work with volume. That's it. Otherwise we're dead. So we are always innovating and hopefully delivering as close as we can to cost. So with that in mind my my answer to Eric's CTO and every CTO out there is I know that CTOs love to build but the truth is that now it's a waste of time and money. You need to focus elsewhere because that's definitely not going to be where you get your advantage. And and maybe a final part of that is AI has

also made it so that it is no longer an option whether you want to optimize at a user or even a request level. I think that we can argue about that today. In two years time that will be obvious. I don't even think we'll be arguing about whether automation is you know a crucial part of any AdMon strategy. So people really need to think about whether they have the infrastructure needed to do that and that's where we come in. Hopefully that's what we are trying to do and what we've been building for the past four years. Perfect. Great. So again good luck with keeping that scale as it gives leverage. Yes, yes we we're growing. We're definitely growing thanks to Eric and and several others. And it would be it would be nice to to to if if other AdMon people reach out after this episode just to because I'm I'm kind of eager to understand how they they approach this build or buy and how they think about the tools they want to use in AdMon in general in general. And I'm

also happy to chat about Lumit obviously. Perfect. This has been another fantastic episode of the two and a half gamers podcast. As you mentioned if anyone has any questions about Lumit, please reach out to Eric or Ignacio. And we will catch Ciao. Have a nice day. You win. Bye. Thank you. That was the best timing. Uh, wait a second because it's still I think Felix didn't stop the recording. Yes, he quit as a host. He'll be back in a second. So then Didn't didn't stop the