Why Data Is The New Superpower In Gaming!

Hello everyone and welcome to GameMakers. Today we are diving deep into what just might be the most critical battleground in mobile gaming growth today, which is data. So since Apple's app tracking transparency shook up the industry, UA teams everywhere are grappling with one question. How do we keep our growth advantage when traditional measurement gets complicated? Well, the answer is in your data. And today we're going straight to the source. So joining me is Roy Nom, founder and CEO of Airbridge, a company at the forefront of helping studios master the postat world as one of the most dynamic and emerging MMPs in the market today. So Roy is an expert in navigating practical strategies like scan, cappy, web to app attribution and he's also pioneering AIdriven monetization with Airflux, helping studios turn machine learning based data insights into real revenue lists on their game products. So if you are a UA lead, a product manager, marketer or an executive trying to get ahead of the curve, this episode is for you. But

before we start, I do want to warn the audience that this discussion is not meant to be a highlevel superficial sort of handwavy talk. We will go in depth and in a lot more technical detail than usual. However, I do think if you really do want to get that real advantage in the market today, it is critical to really understand these things in a very deep and nuanced way. So, with that, welcome Roy and I think we can get right into it. So, first question Roy, I thought we could just start by kind of setting the stage. So where are we today in terms of and we can start with iOS but iOS measurement and performance tracking today. Yeah for sure JK. So first of all like thanks for allowing me to you know join the podcast today. Uh definitely uh just a little bit of myself and the uh airbridge and airflux as a company. uh we've been here in this industry for

last 10 years you know seen a lot of these changes happening uh right now and today definitely I want to talk about you know some of these changes that was made four years ago which was i 14.5 and at so you know it's been over four years uh since Apple introduced at with iOS 14.5 and honestly you know even now many marketers I spoke uh with uh in many of these places especially last week I was in um meu Las Vegas and there's all all still saying it's really hard to measure is performance even the bigger ones uh tell me tell me that as well so we're still very much in a process of adapting and the challenges generally fall into two categories I would say first is a visibility loss uh which makes it harder to measure campaigns performance accurately and the second we call it signal loss so which kind of affects the quality of user data going back to ad

platforms making targeting less effective. So we need to differentiate these two kinds of problems visibility loss and signal loss to better address uh you know these two problems respectively. So although there's there are still plenty of challenges when it comes to ice measurement we've also seen some real progress like both from Apple and from industry stakeholders like the SAN CSBs who've stepped up with compensatory strategies and the tools. So let's break it down. I would say you know some of the progress is definitely we need to understand that scan there has been a lot of major updates to scan itself. So we've gone from version 2.0 to 4.0 right now and now you get up to you know three postbacks which used to be like one post back covering a longer conversion window and you also get more granularity with you know more data like a fourdigit source identifier the source app ID and even the source domain so these additions help advertisers get better

insight while preserving user privacy through crowd anonymity and second progress could be you know major platforms like Meta and Google have built their own private Y preserving yet uh you know matching algorithm within their attribution frameworks uh like MATA's AEM aggregated advanced measurement or Google's ICM. So these are sometimes also called hybrid or advanced SEN solutions and they're giving advertisers more touch points and helping recover some of the visibility loss under AT. Third, I can definitely mention there's Cappy uh which is abbreviation of conversions API which is becoming widely widely adopted. So many advertisers are now sending richer serverside event data like purchases, signups with click ids, hashed emails or hash phone numbers directly to these platforms using their API. So this kind of like improves match quality and gives these algorithms better data to optimize

performance. We're just with two left force and a fifth. So force strategies like web to app are seeing more adoption. Maybe you have seen a lot of people posting about web to app strategies on LinkedIn. So these campaigns rely on web traffic to guide users into app often through a landing page with a quiz funnel or the playables on the landing page. It's a flexible and creative workaround that steps some of the limitations of ISAP install campaigns fibs. Nonetheless, more advanced advertisers are t turning to like holistic measurement frameworks like leveraging lip studies or mm marketing mix modeling. So these methods help validate performance and provide broader understanding of what's really driving outcomes. It's totally based on aggregate level data. So no longer you need to depend on some of these user level or individual level identifiers like ED ids. So all of these things combined, I want to say that iOS measurement definitely presents a lot of

challenges, but still there's a lot of more tools, more workarounds and smarter strategies than ever before. So it's no longer about relying on a single method. It's about combining all of these multiple approaches to get the full picture. So yeah, so that that was uh quite a lot to kind of take in, but just to kind of give a little bit more context to uh the audience. So sounds like with on the iOS side based upon the changes post, it's it's a more complicated landscape, right? And you mentioned a lot of things. Um maybe we could in just terms of recap, you mentioned, you know, scan changes to scan. You mentioned things in terms of what some of the sands are doing like Meta's AM and Google's ICM. Uh you have Cappy, you've got webtap, you've got mm, you've got all these things. And so like the landscape itself is very complicated and has changed post at least on the iOS side. We we can go into more detail and we'll go into for our audience. We'll go into depth into

each of these different areas as well. But when you think about the landscape itself and based upon the changes that occurred, generally speaking, when there's a change in the industry, there's generally like winners and losers. And so what would you say are the impact of some of these changes have been in terms of who some of those winners and losers are from your perspective? Who would those be? Got it. So I wouldn't call it winners or losers. Uh which is kind of a little bit binary uh in a way but it's it's definitely true that a lot of advertisers are you know getting getting like these ch changes are presenting more challenges to the you know advertisers in terms of visibility in terms of signals. So, a lot of things to learn. If you're a marketer, UA manager in a gaming studio, if you're especially coming from a smaller studio background, it's going to be really difficult for you to understand all of these things. But believe me, um, you need to because

all of these things, uh, there's no no such thing as a single source of truth or a single method that kind of solves everything. So as a UA manager in a smaller gaming studio, you have to actually spend time learning about all these new concepts and the strategy and the tools and try to mix them together, blend them well and use them well. Yeah. So then practically then when you think about all these different kind of approaches and maybe some of the in terms of the data signals, can you talk about what would be some of the most impactful data signals that you're seeing studios leverage to get the best UA performance? Yeah, I mean this could be a little bit of a bias answer since especially since I'm coming from the you know Airbridge side, the MMP side. Sure. But still uh you know great question. I I still believe the most impactful data signals that studios are relying upon day-to-day. Uh you know right now still come from MMP data. Uh unless you're actually using just one channel uh if

maybe if you're using just meta the single channel you can be actually relying upon the meta dashboard. If you're using Google as your single channel you might be just using Google's dashboard and using the reporting systems or the firebase to get the reporting. But at least if you're spending like 20 30 20 30k USD per months on UA have a various channels like at least two to three you must be using MMP because that's the only kind of source that you can get the cross channel insight on your campaigns. So you need to rely upon MMP data but when you do that uh basically there are two types of you know major attribution data that MMP provides to you. First is the classic attribution data. So this is like where it gets a little technical on iOS attribution is primarily right now based on probabilistic matching uh because on iOS you understand that ad ed adisees are you know can get only get fetched uh once the user consent on the

AT. So basically right and probabilistic is essentially a euphemism for fingerprinting right basically uh you know that word is prohibited in the industry I I would not dare to use that word okay of course like probabilistic matching is kind of a similar concept to fingerprinting using all kinds of these generic attributes like IP addresses device width height OS model and some of these data that you're getting from user agent or from the device site matching them together uh both from the touchpoint side and the conversion side. Uh but actually some of these probabilistic data um you know are designed to be in a more privacy you know preserving ways so that it does not actually you know go back to the old like fingerprinting days but it's kind of a similar concept. You're right about that. Yeah. just thinking about the the current state of data limitations and so um you know you you mentioned that we still at least on the Android side have the classic data signals iOS it's a

little bit of a different story but can you give us a brief overview in terms of when you think about the differences as far as data maybe you could make that a little bit more concrete for the audience as far as when you when we think about data limitations on iOS versus Google and now increasingly you you mentioned before some of these web to app and some of these other kinds of uh data that we're able to get whether it's from you know and we we can talk about it but on the server side or whatever related to the web so when when you think about iOS versus Android versus web could you break that down for us if I am to just like do the breakdown um I when it comes to Android um you know and when it comes to visibility Android is much more relaxed compared to iOS so you've still got the Google Play referral system uh which kind of helps with clickbased deterministic tracking and the advertising ID is really easy to access uh since it's opt out based so

you don't need to get the explicit constant first to get the advertising ID you just get it by default unless user opts out from the tracking system and no need to explicitly ask for constant like iOS for sure and Android just remains as flexible as it was before and advertisers generally feel they're getting enough visibility for their Android campaign. So if you meet people at the conference or you know if I talk with our clients nobody complains about Android visibility. So it's there um and and then it comes web. So you know web is definitely more trackable than just pure iOS because you can carry parameters with clicks. So whenever you clicks you know these reckoning links on the website or web campaigns a lot of these you know a lot of these parameters just tag along with the click itself to the URL of the website. Uh so definitely it it is very accurately uh tracked in most of these circumstances. However uh you know even for web because web is

dependent on browsers and browsers are sitting on the OS itself. Uh so the privacy measures are tightening especially on uh browsers like Safari because they're on the iOS system. So Safari there's something called intelligent tracking prevention ITP which kind of limits how long cookies last. Sometimes these third party cookies can last only 24 hours if they're set with URL parameters. So plus there is something called privacy relay on the IO system which can hide IP addresses making you know tracking with probabilistic matching even trickier but since it's like still better than iOS I would put web as a second most flexible system out there and lastly iOS. So I was kind of like you know halfway uh you know explaining about this classic attribution. So of course for iOS you can still rely upon probabilistic matching and also we need to you know think about this scan um which is also by default uh the holy grail of is

tracking right now of course with a lot of limitations so probabilistic matching scone and you need to also uh definitely utilize still the AT um so it requires you know app to get explicit constant to access the IDFA uh the problem though is that constant rator still low only about 20 to 40% uh of the users opting in. So it's a combination of all these three probabilistic matching scone um and also you know at based like uh deterministic matching all combined can constitute is tracking. So in summary Android offers a lot more flexibility when it comes to tracking. iOS has strict privacy controls with AT and web kind of falls somewhere in between. Still great for clickbased tracking but facing like increasing privacy restrictions by Safari. And maybe I think it'd be very useful if cuz for some of our audience who may not be familiar with some of the

measurement strategies that you mentioned and there were a lot of them. I I thought maybe we could just go maybe one by one and talk in a little bit more depth uh as far as as far as each of those strategies is concerned. And if there are other ones that you didn't mention, maybe we could talk about those as well. So could could we just kind of go down the list and then maybe talk about each one and just kind of give a highle description for some of the audience who might maybe they're not marketers, maybe they're product managers or maybe they're executives or someone learning marketing. it'd be great for you to kind of give us a little bit of a you know highle view as far as what each of those measurement strategies is and how they work. So uh it's kind of like uh maybe it's kind of a recap of what I just like uh mentioned uh earlier in this discussion. Uh so you need to split between this visibility thing and the signal thing. So when it comes to visibility it's kind of mix of all. So we mentioned the SCAN uh which

is the Apple's alternative attribution framework. We can actually deep dive into this SCAN and how to really raise the AT consent rate later on and we actually do have a web to app which is kind of a rising uh strategy. Pretty straightforward. You send users to the website or the landing page or using the links uh using the webc web campaigns to send users to the app store uh to download the app from there. So this is web strategy. We also can talk about you know advanced meshing or measurement uh algorithm that are provided by SAS out there like uh metas like am or Google's ICM some of these things are uh certainly giving additional touch points to the advertisers for better visibility then maybe some of the more advanced like uh use cases of the lift studies uh the lift studies what we mean by lift study is kind of a running the AB test the hold out test using some of the platforms features like meta's lift study or Google's like conversion lift is something that uh that is used by

some of the more advanced advertisers. And lastly uh something that we've been hearing quite some time recently is like a marketing mix modeling. So this is more like a strategical approach uh because uh you're leveraging your own machine learning model or the open source models to run the MM which is not dependent on the user level ids but so that is the reason why this being more appreciated uh nowadays. So on the other hand we can talk about signals. So we talked about capy the conversions API. Uh this is kind of relatively new and really powerful. It's a tool that allows advertisers to send additional signals directly to the platforms like meta improving optimization because it comes with a better matching rate. Can talk about that. And lastly uh just a little bit of uh you know light into signal engineering this concept. So you know because all of these campaigns out there are getting so much automated uh you need to be careful in choosing uh which user behavior signals to send to the

platforms in the first place. So this kind of signal engineering is also worth um highlighting and right and so just to kind of uh break this down a little bit more. So you you talked about two concept and I just want to make that clear for the audience. You talked about visibility and you talked about signals and visibility means how do you what kind of measurement strategy do you use to get more accuracy right? So like uh for example um it it it on the iOS side anyway it is hard to get accuracy and then on the signal side means how do we send more data in some way and so these all of these measurement strategies that you have described are ways to do one of those two things. You're right. Um scan would be used in the concept in the context of of you know Apple support for giving you some kind of you know attribution data. Web to app would be just just to like help describe these things for our

audience would be when you go from a website and you send them to try to install a mobile app, right? uh the SANS the self-attributed networks like meta and Facebook also have some measurement things which is you know the things that you talked about which we could go into more detail about and then lift studies is more about if I un if I understand this correctly it's more about like incrementality where you kind of see where you are and you try to um measure the impact of a of of some kind of campaign so you kind of see uh what the difference was and MM But it seems like at least some of the people I talked to just have told me don't worry about mm. Yeah, it it works but you need a massive scale for it. But maybe we could talk about that later as well. For sure. Um and then some of these other things that that you have mentioned, you know, Cappy, for example, is is a concept that is newer to me. I it seems like it's related to uh web, but wondering if may maybe we could um

you know focus in on Cappy because that may be something I think some of the other measurement strategies may be a little bit more familiar to our audience but Cappy in particular. Um could you talk about what is sent through Cappy? What is it and why is it valuable? Maybe Cappy might not might be a kind of a new concept uh unless you're uh really actively spending a lot of you know money on on some of these like channels like Google Meta, Tik Tok or Ax or Snap because Kappy are the ones that are mostly you know supported by these bigger SAS and and and the initial reason why KP Kappy came into place uh is because you know when we think of the modern um you know advertising platforms out there the very basic basic principle that they always go by is that they actually target the users that are most likely to make installs that are most likely to make IPS or the most likely to make ad impressions and how do they find

find these users who are most likely to do ABC is based on the past user behavioral data. So simply put like we just send the users who have installed the app so far and they based on this data of users who have already installed the app they want to find those users that are most similar to the users who have already installed the app. So you know when it comes to this basic principle we can understand that you know the more that we send the more accurate data and more matching if if that happens on their side uh the more training data that they can achieve to you know better utilize for training their models to find the users who are most likely to conduct you know action ABC. So in in this context the Cappy is just about sending more signals richer signals using additional identifiers like hashed emails hash phone numbers and click ids and uh why is it important is because you know some of these mobile post outs there especially for iOS

they're heavily heavily relying upon advertising ID and also on the website uh we understand that third party cookies tend to decay and they have a shorter life cycle lifespan So in order for us to increase the matching rate between the ids of ours and ids of theirs which are uh you know the advertising platforms it's always beneficial for you to send some of these additional identifiers like hashed emails, phone numbers or click ids. So this comes with higher matching rate which in turn helps you uh you know have a better uh efficiency when it comes to the targeting algorithm. So, you know, in the context of gaming, uh, Cappy is super useful for those, uh, especially PC console gaming companies. So, interestingly enough, uh, you know, there's a lot of like free-to-play games now nowadays on Steam, and it's truly a game changer. So, up until now, if you were running a game on Steam, you couldn't send any in-game signals to platforms like Metagle, Tik Tok, X or

Snap. They don't let you uh install pixels or SDKs on Steam games, right? But with Cappy, you can finally send some of these in-game signals and you can optimize your campaigns based on these in-game signals. And this exactly what we do as airbrush platform with uh you know Nexon uh with their first descendant came uh there's a case study on our website that you know we use Cappy to send signals to Tik Tok and Tik Tok use it to target those customers or gamers that are most likely to be you know interacting with those FPS uh looter shooter games which kind of resulted in enhanced uh accuracy. Right. So it kind of sounds like what's happening is by sending the these networks uh these additional signals this additional data it helps them with their audience develop for example lookalike audiences similar to what Facebook has been doing for a very long time and even though there's no UD ID or or things of

that nature by having these other identifiers you can then help identify who those players are and then use the standard types of matching algorithms that have already been used before through machine learning and things like that to get the better performance. Is that is that correct Roy? That's correct. Yeah. Okay. And so when when you talk about what you did with Nexon, could could you dive into that in a little bit more detail? So um what did that look like in in terms of what it took for them to roll that kind of a program out and what kind of performance increase that they saw? Mhm. There's no such like official metrics, but from my experience with our customers, Cappy usually help with small but steady like improvements. Typically around 3 to 5% of lift and row, which is not small at all, but which is not big enough. However, um you know, that's not always guaranteed. Sometimes the additional matching keys you send don't help that much or the matching rate could be still low. So, it's something that you should do if you have the extra

development resources, but don't expect a dramatic change overnight. So, now uh we talked about this Nex and First Descendant case, um which is a free-to-play looter shooter PC console game on Steam, uh and other platforms. So, uh it's it was most of the cases when it comes to Steam games, uh people actually run two types of campaigns. they want to just like send users directly to Steam platform with UTM primers on it and use the Steam's like UTM tracking feature to see how many you know wish lists are getting added or sometimes they they understand how many downloads they're getting from uh using the UTM uh parameters. However, uh the shortcomings of using the Steam's like basic uh UTM trackers is is is actually two things. First, it's only clickbased. It doesn't actually take into account view and impressions. Second, it doesn't really help you understand the downfunnel in-game uh purchases, especially as a free-to-play game. So, when it comes to those kind of free-to-play games, you need to actually

be collecting some of these events, in-game downstream events like a purchases or boarding uh on boarding complete, you know, tutorial complete, level 10 achievement, these kind of events, send them to the server side and in turn either the MMP or your own own developers just setting up Cappy can in turn send them again to servers like Metag, Google, Tik Tok, Snapchat, uh X and eventually you can utilize uh these these events that are matched with hashed emails or phone numbers that are gaining from your app uh gamers uh in-game gamers to utilize them for target optimization for your campaigns. Another kind of concept that kind of comes up from time to time in in a measurement strategy that you mentioned which is there's kind of two flows, right, which is like web to app and web to web to app. Could you talk about those two flows and how how is that related to Cappy if at all? Yeah. Uh totally. So this is kind of a new concept. Uh because most of the

times this web to app or even web to app to web uh web to app sorry web to web to app flow flows are mostly adopted by like subscription apps out there or the content uh like online comics or the short drama apps out there. But more and more we see that because of the crossplatform nature of these games like uh maybe HTML 5 games, mini games or the playables on the landing pages, we see this web to app or web to web to app strategies being more adopted by the gaming studios and we cannot you know we we cannot miss out you know the recent changes and external links on the app store uh due to the rulings of the court that the web shops are web shops are gaining more traction nowadays u in a gaming ecosystem. So because of these you know uh rising trends uh there are more uh I would say spotlight to web to app and web to web to app campaigns. So let me could you describe the difference between those two? What what is web to app? What is web to web to app? Yeah. Yeah for sure. So it's kind of there's

only a subance difference but let me explain. um in a web to app campaign uh a very strict web to app campaign users are sent directly to the app store through a tracking link either as a final URL or an embedded in a in a tracking template. So the key difference here is that you may like question yourself, oh like isn't it just the same like with the app promotion campaign? But the key difference here is that instead of using iOS app install or app promotion campaigns, marketers choose web traffic campaign first as a campaign type and then they use the tracking links to send users to the app store directly. So there's a whole lot of difference between you using the app promotion campaigns or the web traffic campaigns but to optimize for app downloads. So that's kind of a whole difference. And second is a web to web to app flow which is kind of a more kind of default basic settings here. It's easy. You send

the users not to the app store but you send users directly to the web page like a landing page or the quiz or the playables or the any kind of HTML 5 games. and then let the users uh redirect it to the app store via a CTA button like a install the app or you can just you know show them a popup when they play the game for like one minute you just like pop you just pop up or you just put a model asking them to download the app to continue kind of playing so now here's the exciting part recently you know Sans like Meta and Google have started attaching click IDs in their postspecs for web traffic campaigns so this means you can now tie inab events back to those web campaigns and even optimize your campaigns based on those in events not just like with the web pixel. So take it a step further you know you can layer on cappy to send even more enriched signals back to the platforms as well. So both web to app or web to app to app are you know

increasingly more adopted by the gaming studios which are originally come from subscription app space. I believe you know with the advance of advent of more of these web shots I think this can be also valid strategies for gaming studios in the future. Right. And do you have specific examples of uh highly effective funnels for either you know web to app or web to web to app and then maybe you can explain why did some of these work so well just for our audience you might want to think about implementing some of these flows. Yeah. Uh so the obvious one example is when you send users to a web shop and use banners to guide them into the app once they make a purchase or even when they're considering making purchases you can still uh show them this web to app banners like install the app or launch the game uh on mobile. So it's kind of a simple funnel but it's effective especially for games with users who are already interested in buying. You can even retarget your existing mobile users to the web shop. let them download, let them actually re-engage into the you

know into the app into game by using the deep links as well. Another example is using HTML 5 games or playables as a landing page. So these games, you know, can hook people right away and then use a CTA button to move them onto the app. So I've seen those cases with some of the lightweight MMO RPG games out there. So you just like when you actually are landing on the website immediately the engine runs on the browser and then you're just like put into the village you can start your journey right now and you know once you just kind of like kill some of these uh monsters around the village and you reach like a level two to three uh they will ask you to sign up and once after signing up they'll ask you to download the app. So this is kind of another web tof flow that I'm spotting right some of these subculture lightweight games you can also do that not just the MMORPG games and lastly uh I would say kind of quizzes or surveys it's kind of a pre-registration flow you can actually ask users uh to for example let's just say you're having a huge IP

let's just say I'm a big fan of Star Wars like let's just say that you do a quiz like uh and then like people select some of these answers like uh like what's your favorite show is it or is it like Rogue one or something like that, Mandalorian. You can, you know, click some of these um, you know, answers like uh, solving the quiz and eventually with these answers, you can also ask them to download the app or register for the upcoming release of a Star Wars IP based mobile game. So, these kind of flows uh, all fall under the concept of web to app or web to app to app kind of um, you know, user flow. Okay. Uh and then because it's web to app, I assume app store policies don't apply to these flows. Is that correct? Yeah, you're less restricted and uh important thing is once user clicks uh some of these URLs a lot of these parameters are deterministically remain on the website which can be collected by web SDKs or the smart scripts. So definitely you will have a better visibility and time to time even uh for

the uh campaign efficiency I've seen some campaigns getting uh bigger a higher rorowass uh compared to the some of these you know app promotion campaigns this is kind of you know what you will only understand after running you know multiple types of campaigns at once but I've seen many cases where webto campaigns or web to web campaigns even perform better than the straightforward uh app promotion campaigns okay And kind of going back to that big list of stuff that you mentioned, one of the other things that you mentioned are some of the measurement strategies from the SANS. And so Meta in particular has AEM for example. So could you talk about AEM and then uh they they've got like eight conversion events or something of that nature. I I don't understand that. Well, it'd be great to to hear from you what it is and then how do advertisers use these conversion events? Yeah. Uh, so when it comes to AM, you

it's actually better for you to understand it as just one of the three options that you can choose when running iOS app promotion campaigns. Um, the other two options are obviously scan scan uh and web app that that we just talked about. So you can just use just one of these three or two of these three or you can just mix them together and see how it goes with you know all of these campaign times respectively. So the important thing is not really necessarily about understanding AM itself. It's about testing all three methods and seeing which one works best for your particular game title and your marketing strategy. Okay. So you know a lot of marketers they tend to favor AM because scan uh sometimes is kind of bit more complex and it takes a lot of patience for you to get the results like two to four days of postspec delay. Uh plus like scan is really tricky. So a lot of marketers they prefer using AM nowadays. Uh so sometimes I even encounter marketers they don't use scan at all. But uh that said I just like

believe that you know trying out all three methodologies together and see which one works best for you is really important because I have also encountered marketers who say that scan campaigns work better for them. Like for for example, I've meet met some like midcore idol RPG marketers who told me that scan campaigns work really well for them. So it's kind of a testing and seeing the data to make a final decision, right? And I think part of the limitation with scan is just the the sparseness or the limitation in terms of the data. But with between two and four, there's more data, right? Like Apple's starting to open it up a little bit. Do you think they'll open up more in the future so that scan becomes even more usable? Hopefully in the future. Yeah, it must be cuz like um I know that Apple has been adding a lot of things uh you know from the 2.0 to 4.0 there there's there has been a lot of updates although I wouldn't really say that even with all of these updates uh people still think

that scan is really tricky and then hard to use but there's been some major updates so far like you know scan used to have just one postback but now it has like three post back the second and the third. Yeah. So it's three times as good, right? Yeah. Which is great. But you need to understand the second and the third ones, they do not have the fine grain um conversion values. So it's kind of a coarse conversion values like the high, middle, and low. So they're like less usable than the first one, but right right the third. And second, um you actually have you know fourdigit uh source identifier. We used to have only two uh digits. So you can you could only track up to 99 campaigns uh using scan. But now with with these two more digits you can track more. They if you actually reached the tier three across anonymity you can get even get the source um source app ID which is really the you know app store ID of certain app that you that show the ads of your uh your

your game title. You also have the source domain uh which is the website that show the ads. So a lot of these uh richer data came into place which which can be useful but still uh because of a lot of these restrictions like crowd anonymity uh random postspec delays and the very hard concept of setting the conversions value of the scan uh many marketers just feel very difficult to use gun nowadays. Yeah. Right. And then there's also like this issue around uh multi-postback windows. Could you explain that? That's like 0 to two days, 3 to seven days, 8 to 30. I I don't quite understand that very well. Could could you explain what that part? Got it. So the straightforward simple answer to that is you don't need to mind too much on the second or the third post back. The reason why is, you know, as a marketer, you need to make quick changes to your uh media mix. You need to actually every day you need to be scrutinizing the numbers, crunching the numbers of each creative uh campaign and

the channels and just waiting another like 14 days or 35 days doesn't really make sense. So that is a reason why um we would usually recommend studios and a smaller studios and a UA marketers just to focus on the first postback. So here's why that makes sense. Uh first the initial postback is delivered within about four days. So earliest two days I mean um latest four days. So within two to four days you just like uh get the results which is kind of fast enough to actually influence creative testing campaign optimization. You know you can be as patient as seven days right? So you you get the feedback while it still matters. And second, first postback is the only postback that includes fine grain conversion value up to 64 combinations which gives you way more insight compared to the course value that you get in the later postback, the second and the third one. And lastly, uh those fine grain values are also backward compatible with SCUN 3.0. So

many people believe that once it got updated to 4.0, it's all 100% done. But no, actually 3.0 0 still exist. Uh and then you need to have this compatibility between 3.0 and 4.0. And you need to remember that under 3.0 there were no such thing as the second postback or the first postback. So keep things manageable. We strongly recommend optimizing around that first postback window. It's where you get the most data, the fastest feedback, and frankly the highest ROI on your effort. But if you actually want to go further from that point, you can leverage start leveraging the second and the third post back and you know just interpreting the results from these two and um optimize your campaigns. Any other tips or tricks on on scan or should we move on to one of some of the other measurement strategies that you talked about? Yeah. Uh well actually a lot but let me simplify that into just three things. Um many people talk about

scan but sometimes I believe what matters more than scan and it's more ROI you know impactful is the AT consent itself. So you know simply the users more users who opt in the more advertising ids you get which kind of gives you better visibility through your MMP and your advertising platforms can perform better. Yeah. So think of scan always as a plan B. So always plan a should be at constant. So here I I want to actually share with you some of these you know proven ways to boost the ATD optin rates. So why is it matters is I've seen some of the studios doing it well and who have raised their AT constant rate from almost like 10 15% to even 40 to 50% maximum. Okay, which are dramatic. So some of the things that you can definitely do is first like using like pre-prompts like explain to the gamers why you're asking you can even tell them that oh this is like important

for us to keep developing the great products so we need this data. So you know how will it improve their experience and so on. So a clear friendly pre-prompt build trust and transparency and I've seen some of the smaller studios just directly asking users can you please please opt in for at which work like that and second is definitely you should customize the at prompt uh you know test it out and lastly uh some it's kind of a kind of a trick that you can also uh ask the users who have already unconstented to consent later on so it's actually possible for you to revert the settings from you know opt out to opt in uh when you go into the settings of each like application you can actually uh turn it back on. So maybe time to time uh you can actually just show them you know in-game uh messages asking them to turn on the a uh turn on the AT back and deep link them into the setting page of the uh that application. So definitely scan is something that is

really important but before that I want to give out this very practical advice of uh optimizing uh the AT constant rate first. Right. Yeah. And yeah it sounds like going from 10 to 40 to 50% sounds pretty pretty huge. So yeah I I and I I've seen some of those tricks. I I do think they're simple tricks but but seem to be fairly effective. Um so maybe moving on now uh I I do think one area or one measurement strategy that seems to be a little bit controversial just because of a number of folks who you know some people say you you should you should understand MM others are like well you need massive scale and this and that. So can you give us your perspective? So how are mobile studios practically in in real practice implementing MM is that actually happening uh based upon you know your your real world customers what are you seeing maybe before you know talking about some of the real steps that you

can take to implement the MM uh you need to understand that MM is kind of uh not easy for the smaller studios especially since that it's all about you know coding stuff and using um some of these open source models out there like Robin or Meridian by Google. So definitely uh you need to be having at least like two to three uh people dedicated to data science or data analysis. So if you're large enough to afford the MM as part of your measurement stack then I would certainly advise you to try it out because many studios out there uh have benefited from using MM as a as a complimentary measure to whatever they're having. So I definitely recommend it. So some of the practical steps that you could take is certainly uh if you have two to three people at least a full-time or uh the part-time uh BI uh team. Yeah. And if you are a casual midcourse studio um certainly you can explore that and the first step could be selecting the model. So uh fortunately uh some of the bigger ones like Meta or Google they actually are

providing the open source models that you can easily leverage to run the MM models by yourself. Uh so one is the Robin of Meta. The second is Meridian by Google. So they actually have uh slightly different philosophies because um Meta's model is more on the frequentist model whereas the Google's model is more in the basian model. These are the statistical terms but you can choose whatever that fits you well. Uh and second is you need to choose your variables but it's kind of a preset thing. Uh first is the marketing variables which are most likely be the cost data or the impression count on a daily basis uh per channel, country and OS. And certainly you need to choose the target variable which you are analyzing eventually could be a revenue could be a downloads uh could be even like some other things uh like a purchase count. So you need to set them uh right and lastly something that we call as context

variables. So context variables are like the external variables outside of the app itself or the market activities themselves that might affect the target variable and this includes seasonality. This can also include the app store rankings or your competitor's app store rankings or some of these external variables that might affect the target variable outside of these marketing activities. And then uh step three four five step three calibrate the model by running the lift studies uh like meta lift study or the Google's conversion lift to calibrate the model. This this part is really necessary because the models themselves do not guarantee uh the accuracy unless you actually calibrate them with the ground truth the guaranteed truth which are the uh incrementality. You can check the model accuracy by using some of the statistical metrics like R squared MATE or MCMC conversions. Um and lastly after all these uh things that your data science team must have gone through you can finally run the model almost like uh

like every week or every months and the collaboration that I mentioned in the third step can be done in every like three to six months. So lastly uh you know important thing is the interpretation of the MM results. So we've been done some of these um case studies with Google and some of the casual gaming studios out there like Amma notes or Aloha factory and most of these cases there's a huge gap between what you will see in the MMP dashboard versus what you will see from your MM models. So don't panic because they're born differently. They need to interpret they need to be interpreted in a different ways. So one of the things that you can do is you can actually you know use the budget optimization uh feature of MM to really just you know really find out those channels that you should increase the budget or to decrease the budget from. So you can actually uh reallocate some of your budget using the MM's budget optimization feature. And second, you need to definitely attend to those

channels that show the biggest discrepancy uh between last touch attribution versus the MM. So those who show the biggest discrepancy are the ones that you might want to take a deeper look into and think about the reasons why this channel is so much overstated under last touch point model. When it comes to MM, it shows kind of less contribution. Okay. Yeah. So, sounds really complicated, but I think the one of the main things I took away is that clearly like if you need to have two to three dedicated data science as well potentially BI folks on staff, it does seem like you do need at least some level of scale to basically justify three additional hires for example. And um yeah, I I thought we could now shift towards an area that that I think is very exciting for me because I think this is kind of a newer kind of frontier but or may maybe it's not so new but maybe we could talk about this because you know

we have been talking about UA but kind of pivoting over to monetization and how does monetization and the data that we get from monetization also feed into UA there's been some discussion about how they're related. I wanted to get your sense uh you know is in terms of how your customers are actually using some of this data and from what you're seeing in the industry how are they related um how does this monetization data feed back into UA you know there's a lot of things to say about this and I'm also equally excited about you know talking about this topic uh we we I I can definitely tell u you know the audience and um you that monetization and u ua are just tightly tightly equivalated to each other uh and more so in a future and actually this is already happening in a modern you know DSP world take1 for example uh they

don't allow users to run tier OS campaigns without using their mediation solution M1 max and you know potentially the reason for this is that these platforms have access to symmetry uh first-party data and bidding information from their mediation systems which could heavily influence the user acquisition algorithm and ultim ultimately the results themselves and and some of the numbers out there back this up. So I I recently read an article uh by Gamebiz Consulting and they say that ML1 and Unity both perform four times better when using their own mediation platforms. So these stats are just based on their own like gamebased consulting clients. So it's definitely worth noting. Um so you can definitely tell that these uh data uh mediation side data are getting used for uh UA algorithms. And another interesting thing that we're seeing is that more and more platforms are using ad impressions uh to for their target optimization campaigns. And one interesting example

is definitely Meta. You know Meta surprisingly uh didn't actually have this kind of setting to allow users to optimize for ad impressions. But from January 1st this year, they kind of like unlocked this feature for Android app promotion campaigns that they can use the ad impressions at the target optimization event. Uh which has been already done for many like DSPs like M1 or um you know other uh like Meloco or Liftoff or other uh appear DSPs out there. But the fact that someone like Meta is opening this up is actually really crucial uh kind of fact that IAA is being regarded as very important signal to even the bigger platforms and those who are not following this trend I believe will uh follow this trend really soon. Okay. And it seems like you also so I I guess that that'll be an ongoing area that that it it would be exciting to see um kind of new things happening in. But I also wanted to ask you guys about or

you about Airflux. So maybe you could also talk about Airflux and then how does Airflux leverage user data. Well, just to introduce about Airflow uh which is a product uh you know developed by Airbridge team uh is an AI powered ad monetization engine which kind of helps um casual gaming companies optimize interstitial ad timing, frequency and the cooldown as well as the rewarded videos payout amount to really help them maximize the user LTV. So we have been working with leading casual gaming studios in the world and have seen great results like ranging from like 10 to 50% of the user LTV uplift. So one of the reasons why we built this tool in the first place is we believe that the eCPMs of these publishers out there are getting like a little bit slowing down. So we know that there's a lot of things that happened in the course of last few years that the length of the you know videos um got much longer from like 30

seconds to sometimes like 60 seconds. And we also understand that more of these publishers are heavily heavily monetizing through IAA. thinking about the hyper casual gaming studios or hybrid casual gaming studio midcore light core they all monetize through IAA in the countries like T2T3 countries and even with the entire user base and with all these frequency increase uh we're still seeing that the CAT LTB ratio reaching at one is really hard to achieve and also people are doing a lot of these segmentations and AB testing when it comes to monetization and I've seen much of these work has been done pretty manual ually hand job or using the Google spreadsheets to just like adjust some of these variables which is really time consuming and inefficient. So then we decided to build this kind of agentic service uh airflux to help them auto uh help them predict the LTV segment users um auto segment the users into many different buckets

run the AB testing and use the reinforcement learning to keep enhancing these ad policies so that they can reach the optimal uh level of frequency at timing for inter interstitial ads and also pay out rewards for the rewarded videos. Okay. So, I there's a lot there. Could could you maybe walk us through exactly like the the flow of how Airflux works. Um, and then you know basically what it does uh and you might be repeating yourself a little bit but but sorry I'm a little bit slow here. And then you know how quickly it it adapts and the kind of optimizations that are happening through some of these like AI initiatives whether it's a Gentic or otherwise. I can just like go back to the genesis of this project uh this product. Um we were working with a casual gaming company with a very uh I was a successful title with 10 million MAU uh that are played by like people around like 30 40 different countries T1

T2 T3 all well mixed. So when we were actually working with them, what odd thing that we realized is that uh for every users from every country, the first interstitial ad timing is always the same. But when you think of it, it's a little bit weird because you know those users that are playing this game from Philippines will certainly differ from those users that are playing games in France. So we did kind of a little bit of testing that we segment the users into many different like groups. Uh and then for the European users uh we kind of tested uh whether if we delay the at first interstitial at timing just by one or two or three minutes how does this affect the retention curve as well as the LTV and surprisingly for users and friends uh you as you might expect just by delaying the first ad interstitial timing by 3 minutes it greatly greatly increased the retention rate as well as the AIP conversion rate which is really interesting. Because interstitials, they actually are lucrative. On the

other side, they're really disturbing the users gameplay that those who actually see this early might not be as engaged as we want and do not make purchases even. And this kind of had a compound effect on the retention and the IP purchase. So then we thought that oh if this is the case, what if we can automate this? What if we can optimize this using the auto segmentation, AB testing and reinforcement learning all combined in one piece? Maybe we could really, you know, alleviate the burden of the ad monetization managers having to do manual segmentation, AB testing and do it again. Yeah. So this kind of it kind of sounds like you guys are just using AI to do a lot of multivariate testing uh for this specific application which is like I IAA for for inapp ads but then ultimately you're you're probably targeting towards how do you maximize LTV for example. Yeah, that's that's okay. I got it. And then um so so

it sounds like so how new is this product and how like it sounds like there are a couple of use cases or case studies that have been fairly successful. How are how are customers kind of reacting to this now or how how how much has it been rolled out to date? So we've been definitely testing with like 20 different customers even before we launched the product but the actual launch date was just uh six weeks ago. So uh very interestingly ever since we launched this product we've been posting some of these things on LinkedIn some of our case studies our observations of how this could really you know doesn't really negatively impact the retention while raising the LTV and some of the case studies have resonated. So we're actually receiving like hundreds of these uh incoming demo requests from all around all around the world right now. So initial response from the community is great and some of the case studies that I can just share with you is um something that I really like. Uh it's about a very cautious um midcore gaming

company. Uh so they have experienced that if they show the interstitial ad too early, they knew that certain users from certain countries were certainly going to drop off. So they didn't want this to happen. So they rather took a very conservative approach of showing the first interstitial ads on the 10th stage. So users has to pass all the way uh from one to nine and then once they reach the 10th stage and clear it then they'll see the first interstitial ad. So although this strategy might make sense uh to certain customers who are you know higher probability IP users or countries like coming from countries especially in T1 countries in Europe but most of the other users the the the problem is that 80 85% of the users are not those kind of users so they were actually u missing out the huge revenue opportunity so they kind of implemented airflux and just like let us take from there that every kind of uh each like

ending of the stage they kind of like let the inference API of the airflox to take control of whether to show the ass to the users or not and by after doing that we have uh saw we have seen a surprising result at the 54% LTV uplift because there were obviously a lot of you know revenue that is like on the table that is left behind uh so we actually did the segmentation based on the predictive LTV as well as see some of the generate attributes like country OS and some other like playtime and segment the users for those users with higher IP IP probability of course like we didn't show them. We we stay patient until like 10th stage or ninth stage. For those users who are not likely to make IP purchases were just like pretty much used to watching the ads, we just show them anyway from the beginning which kind of resulted in this kind of huge leap of 54%. Okay. And so it sounds like um even though you're an MMP, sounds like you're venturing into an area around well I guess that that is sort of the the job of an MMP to help help optimize to not

only collect data and you know manage attribution data but also help your customers optimize for revenue and things of that nature. Uh seems like however that a lot of the ways in which people think about MMPs is from the perspective of user acquisition. Mhm. Um you're now you're you've now have a product that's helping optimize ad add uh sort of inventory or the the ways in which ads appear to the player. You're also getting data about those ads. We mentioned before, you know, the potential of um of of of having, you know, like data from from ad monetization impact UA. And when you think about AI and how it's being applied to Airflux, could you imagine a scenario now given all the things that you have talked about where AI could potentially co-optimize both together? Um I'm wondering how you would think

about that. It seems like there's an opportunity there, but um wondering what you think about that. Yeah, for sure. So we we try just just to clarify, we try to build Airflux as a independent solution from Airbridge. So although it's coming from the same company, we want to build the airflux as a just MMP agnostic product. So no matter which like MMP solution you're using, you're free to use like test it out. Okay. Uh so when it comes to uh you know UA and monetization, we believe uh just by optimizing the LTV, you can actually create a compound effect both on the LTV side and the tech side altogether. The reason why is that, you know, many of these casual gaming companies are having low profitability. Meaning that with with just like with every install, they may be just like earning just one or two cents out of it and very thin margin uh that they're actually they're kind of on the fence right now. And when it comes to, you know, fundamental UA equation, it's all about LTV ratio or the rorowass. So if your rorowes is positive

say above like 100% or like even somewhere between 100% and 142% uh by you know counting in this u commission fees you can scale and you can spend more. So LTV uplift I believe is tightly tied to your UA strategies because as LTV increases that means your rorowes improves uh which kind of shorten the payback and suddenly channels or campaigns that previously didn't uh look profitable are worth testing again. So for example I had a good talk with Ricardo of MAF uh at the previous like uh MAU Vegas last week and what he's saying is that once you can spend just 0.01 01 USD more into the CPI, you can actually get so many more installs because it's all about bidding. It's kind of an auction if you want to if you are willing to pay more. Yes, you can get more quality traffic. So, this kind of silus difference can make a lot of impact on your UA. And lastly, what's

interesting is we have seen that the inflow of the new users generally tend to lead to higher eCPM. The reason why is you know a lot of these biders they assess the users if the new users comes in they don't have much information on these users or the less information on these users. So they tend to uh you know give the higher bidding uh higher bidding prices to these new users which kind of in turns uh leads to the compound effect of your entire game's eCPM goes up. So I think like just just in summary I believe better LTV better UA and better eCPM and it kind of like goes in a one big cycle uh which is virtual to each other. Got it. And I I get the point that you're saying which is to the one point which is that even for a slightly higher LTV because obviously the app stores are an auction based market. you can get dramatically more. It's not it's not a linear relationship, right? You can get dramatically more UA scale and then the

more UA scale you get, what you're saying is also on the ad monetization side, there's a scale effect to the number of users that you have that are watching ads and so you get a higher eCPM with some level of scale. So, it's it is a potentially virtuous cycle. But going back to like uh the practicality of how a studio would think about implementing something a a product of some kind or maybe you guys are going to create a product that helps to co-optimize using AI or a Gentic or something both of these things could you talk a little bit more about how that could potentially work? Yeah so this is kind of the futuristic uh you know future for thinking not that much realized. So once we actually tap into this monetization, we realize the same things could happen in the UA or even the CRM side because you know at the end of the day it's all about segmenting the users doing the AB testing and if there's an incrementality there you just do it again and again at scale. So many of these things uh the

basic principles can be applied back to the UA. Uh for example like uh let's just think about the segmentation. How do you segment is just infinite possibilities of how you segment the users. You can actually just send the audience of uh your you know IP users who have made at least one time a purchase last 30 days. You can just send uh those users who have like seen the ads many kind of you know seed audience can be made and out of that seed audience you can make a lot of these look back uh sorry lookalike audiences uh corresponding lookike audiences. So the targeting wise there's this infinite um options that you can test out almost like a segmentation. Creative wise, there's a lot of layers that you can test it out. For example, like it could be UGC, it could be playables, it can be video, uh, or it can be display. And even under UGC, when you when you kind of decode it, there's a lot of hooks, there's a lot of main message, there kind of CTA. So, you can just think of it as a when you vectorize all of these variables. It's just the combinations of

all these interventions here and there. And uh right now what's happening is just users they just like by instinct or by intuition they just kind of uh distribute their budget across all of these different layers of variables and just see how it goes and that's it. And the and the feedback cycle is so so because it's human done. I've even seen one gaming studio just allocating one full-time marketers just to turn it off on and off these creatives on and off every day which is great. But I I think this uh eventually can be uh to a certain degree be automated by using the AI. The same kind of principles that we're leveraging for Airflow. Yeah. Yeah. It it's I I think you're absolutely right about that just in the in the sense that like to your point there are so many features, knobs, variations, segments, all of these things which a human being can't keep all of those variables in their mind, right? And it's going to be difficult to see the relationships

and especially when you're when you're also taking two different parts like UA and and ad monetization as an example and understanding how they are related which may not be entirely visible to a a human being. And so you take that together and it does sound like there there should be something an AI tool of some kind that looks at these things and kind of co-manages them together with with an understanding of of both I believe. So uh the reason why is that when we talk about these agentic services we only take about talk about the chatbot the chat UI that you can you know type in something and get the response back but actually in reality agentic the definition itself is replacing some of the human workflows that are pretty manual or hands-on that are not automated to the extent where you can just leave it there to the AI. So I think one of the things that we can

definitely dramatically change in upcoming three to five years is some of these manual tasks both on the UA and monetization side like sementation, AV testing, hold out test like all of these reinforcement learning all combined can definitely make this loop much smoother, automated and grow up to scale. Yeah. And I I mean just the way that LMs work, right? I mean you've got gradient descent which is basically optimizing against you know millions of different features right and so uh it's yeah to your point I think that um maybe the next thing we can talk about is like we've got all this complexity already right in today's market we've got all this complexity we've got AI which we've been talking about which potentially can manage a lot of this complexity um some specific things that AI can do um and even within the last two to three years we've seen I mean all of these things all the things that we talked about right w with respect to scan uh the the sand stuff with men's amm with respect to like cappy all these things that are

happening um so just to kind of like pause for a moment and you know we've talked about a lot of these things but if we were to if I were to ask you to look into the future the next two to three years what what do you expect data like how do you think data strategies evolve from here. I I don't have a very clear answer to that. But if I am to, you know, see what's going to happen is I think it's just pretty much the same as what we're experiencing right now. A lot of fragmented data points or a lot of fragmented data sources, a lot of methodologies. So I can definitely say that uh more and more uh the advice of me just giving out to these bigger studios is to invest into measurement stack because nothing's going to solve the uh this measurement gap easily. So definitely invest in MMP, scan, try out web to app and also like uh just casually or regularly running lift studies with meta and Google or

some of the bigger platforms out there. Marketing mix modeling as kind of a high level overview on your budget planning and sometimes you can even use it as a budget optimizer uh just to decrease and increase the budget that is allocated across different channel sources. Yes, you can use the MM for that. So I believe all of these things combined and even with even some there's some other things that I haven't even mentioned today like data clean rooms or private data marketplaces. Uh so I I believe that just mixing uh just three to four of these tools uh in a very adequate but still like ROI positive manner I think that will be the holy grail of uh this measurement stack in the future. Unfortunately but fortunately I I think that's that's going to be the case. Yeah. Right. And I I think maybe one of the lessons learned is that and I think I mentioned earlier with every new kind of major change there are winners and losers generally and kind of what we've seen you know you kind of touched upon app having a advantage through data. Um,

I I would submit that like some of these Chinese Forex guys saw an advantage by using specific strategies and things of that nature. And and so I think ultimately whether you're a marketer or an executive or product manager, um I think the main message or the main thing that that that I think that we need to be more cognizant of is how will we gain that alpha? how will we gain that competitive edge? And it's becoming increasingly clear that using data in specific ways um whether it's um with more accuracy or to your point on signal, right? Uh will potentially give some game studio a pretty dramatic advantage. And so the next question I have for you Roy is that given that what types of maybe new tools or

platforms or signals or something do you foresee emerging to give studios that extra competitive edge? Got it. So maybe uh you know I'm I'm coming from Airflux so I should definitely reiterate myself on this point that increasing the LTV especially for the IIA revenue uh it's just like have a so much bigger compound effect on everything. So I would definitely suggest uh you know just investing more into this ad monetization and also we can think about ASO. ASO is often something something that is pretty undermined but with ASO you can have like better conversion rates and better LTV altogether. So I I would definitely recommend checking out tools like you know our tool like Airflow or Press Play by feature which is kind of the um ASO automation tool uh that is really helping casual games like test out. So it's kind of like leveraging the same principles as a verifox which is interesting that we're all going to the same direction. So once you increase the

LTV uh you have a lot of these options to explore for example like you can diversify your UA channels like maybe using some of the rewarded uh ad platforms out there because once you have a higher LTV you can reward uh the gamers better and there's like 20 30 different reward uh ad channels out there and you can boost your app store ranking with the long-term CP CPA campaigns with that. you can also leverage some of these, you know, playable ads. And second is definitely once you increase the LTV by touching touch touch basing the ad modation ASO, I you know, I've been listening to a lot of your podcast recently and certainly the uh creative production is where the AI benefits the most. Uh so some of the tools like out there like inimo AI like turning game plays easily into videos or play which is kind of playable maker uh using the AI just just quickly uh create these playables for you a lot of these tool or like UGC generators like pulled the ML you can leverage them to

certainly explode in your um creative production and lastly uh I believe that there should be some agent tech services coming out in the UA space right now I I'm seeing uh some of the ones like uh reforged labs. So once these like UA agents take in place and also automate the budget reallocation or the assessment on each channel campaign creative or country OS uh this kind of whole loop is complete the admonition ASO which kind of helps you increase the LTV more creatives in place um and certainly more opportunities with more creatives and lastly kind of a gentic services just running UA kind of combination uh optimizing them on and on. All right, Roy. Well, so just to close or or to end the conversation today, I wanted to ask you about kind of practical advice that studios should be thinking about. So, I've got three rapidfire questions for you. And the first question is, I'd like to know your number one practical piece of advice

today for a studio trying to optimize their user acquisition. Well, for the bigger studios, invest more into measurement stack. Um, for the smaller ones, try out Cappy. Okay. And question two, uh, your number one practical advice for a studio trying to improve their monetization strategy. Smarter segmentation, more AB testing all the time. Okay. And maybe also to use Airflux. Yep. And then because you did, I I have to admit. Yeah. Okay. And uh number three uh a warning or pitfall to avoid when setting up or evolving measurement today. Well, although last touch point is the default and the most basic stuff uh just like biggest pitfall is blindly trusting last touch attribution and especially when it comes to scan you need to understand that scan has a much longer look back window uh like uh 30 days and also you need to understand

that it's also the last touch point attribution. So although it's great that you use less touch point, just don't uh you know blindly trust it. Okay. All right. Well, that wraps up our discussion for today on datadriven strategies reshaping mobile growth post. Huge thank you Roy. Thank you so much for sharing your practical insights and helping us sort of navigate exactly how to turn data into competitive advantage. For anyone out there interested in a deeper dive into these strategies or exploring more, check out Airbridge's website or connect directly. Can they connect directly with you? I'll I'll leave kind of like contact information and information in the show notes. Um, but as always, please tune in to Game Makers and for our audience, we will catch you next time. Thanks again, Roy. Thank you, JK. [Music]

[Music]