Unlocking Attribution with TikTok and AppsFlyer

Hey folks, and welcome to the Deconstructor of Fun podcast. I'm your host, Misha Kachkoff, and in this episode we're going to talk about attribution and why attribution matters. So, in mobile marketing, attribution assigns credit to a specific marketing channel or campaign for driving app install or different type of user engagement. It involves linking different data points such as ad spend and user actions to understand the impact of mobile of mobile ads across the entire conversion funnel from the point the person sees the ad to them tapping, to them installing, and finally performing different types of action inside that app or inside that game. Now, this process is crucial for identifying relevant user cohorts, analyzing responses uh of different creatives, and optimizing ad spend. Accurate attribution is essential for calculating cost per conversion, refining game

design, and improving campaign performance, making it a fundamental aspect of mobile ad ecosystem and modern free-to-play game design and marketing. So, in this episode, I'll be joined by by two very senior attribution uh and measurement professionals. Definitely not where I'm at at my at my best, but um I hope you enjoy the podcast and then I'm I hope just like myself you'll be able to learn something new about the the challenges of attribution in the uh the current privacy environment, as well as different different ways to to overcome these challenges and and the forward look on on how the mobile marketing will look like. I had a little bit of a malfunction of a microphone apparently in this episode, but you are not listening to Deconstructor of Fun for me, you're listening it for the guests. So, uh I hope you'll be able to nevertheless enjoy our magnificent guests despite my microphone malfunction. If you like this episode, if you want to get any much

appreciated as always, and like, uh share, give us a few stars on the uh whatever podcast app you're using or YouTube. That is always always much appreciated. So, without further ado, ladies and gentlemen, mobile attribution. This podcast is brought to you by AppsFlyer. In today's digital world, understanding your app's journey from discovery to download is more than just insightful, it's essential. Enter AppsFlyer, the leader in mobile attribution and marketing analytics that allows you to measure the full potential of your marketing efforts, making every ad dollar work smarter. With AppsFlyer, that's your new reality. Dive deep into data-driven insights that reveal exactly where your users come from and how they interact with your app. AppsFlyer, where your app's potential meets performance. Welcome to the podcast, and let's kick

off with the introductions. So, in the the blue corner, no no, this is not a fight. Uh so, first we have we have Steve Lockwood, head of client measurement at TikTok. You Steve, you previously held marketing and analytics roles at companies like eBay and Facebook. So, Steve, welcome to the podcast. Thank you, Misha. Good to see you again. Hi, everyone. Um looking forward to getting to the discussion. And I think in the other the recorder, you got Roy and I, uh VP of product at AppsFlyer. Roy, you've been at AppsFlyer for 7 years, and we met at last week. It was last week. We met last week in New York at this Google event, and I was talking to you about like to your boss about boosting doing this episode you getting from AppsFlyer, and Google told me they pulled me on side. So, and they said, "Roy is the wunderkind of AppsFlyer. Like, this guy knows everything." Like, this guy? I'm like, "Attribution?" Like, "Everything about attribution." So, I right away

came at you in the bar and like, "Roy, you got to come in on the podcast." So, Roy, really happy to have you on the podcast. Thank you so much. It's really good to be here. So, let's let's jump in right away and and Roy, let's start with you. And kind of going back, you know, 7 years at AppsFlyer, you probably seen it all, but how has attribution evolved, let's say, during the last, you know, well, 7 years that I've been that you've been there from like that that peak um new way times to what we're seeing now where a lot of companies are almost flying blind or not flying blind, like you tell me. How has the market how has the attribution really evolved? Woo, how long do I have? Well, we got let's let's do 5 minutes. Let's like go through all the issues. All right. All right. All right. Uh fair enough. Um I think actually I I I think there's a it's an interesting story. I think MMPs traditionally, so MMPs, I hope the audience know um the

measurement uh partner, these refers to the attribution provider specifically in the mobile ecosystem, which uh AppsFlyer, the company that I've uh been working for, um is one of them, the leading one. Um and within this uh period of time, so the way I see it and the way we see it, um attribution providers uh used to be, let's call it, um connectors between, say, clicks and installs. Right? And then provide a data, match the click and install, um and provide this data. Now, over time, regardless to um things that we're going to touch on within this episode like ATT, etc., um regardless to that, um the vision the company uh the mission has gotten a lot bigger. It's no longer just a call it a commodity of of of connecting clicks and and and installs. Um it's about providing the market

marketing picture, understanding ads efficacy, you know, name it in different terms, understanding what every dollar does for you. So, at at some point early on, we've also added um cost reporting to connect with attribution. Then later on, we've also added uh fraud enablement and fraud detection and fraud blocking, etc. And then obviously came together um ATT um um which really disrupted the industry. Obviously, this was a major major earthquake within the mobile ecosystem uh um marketing measurement, but not just measurement, also targeting and and optimization, signaling, etc. All of that. Um and then the ecosystem really adapted, and us at AppsFlyer, and I've been part of my role in the past 7 years, I've been leading um our iOS solution focusing on SKAdNetwork together with AppsFlyer attribution, and then the world of

attribution, the world of um the marketing picture, as I said, really evolved into um again what we call is a single source of truth in terms of last click attribution, which is combining um output of SKAdNetwork together with MMP attribution, consented data, um Apple search ads, etc. There are so many different inputs into providing again what we call the marketing picture. That was the shortest I could I I could do. Got it. So, everything so so so MMPs, so attribution providers have a bigger role now because they have to cover much more ground to give developers and publishers that holistic picture. And then one more question for I know you should ask Steve is like, what do you think about, you know, all the other pundits talking about the end of fingerprinting? I know that's a that's a topic that that um that that you like to, you know, address maybe sometimes. Uh we've been very uh

uh insisting on specifically AppsFlyer, we do not fingerprint. Um and we are very much aligned with Apple and every single one of their policy, and we are actually the biggest endorser of every new uh whether it's enforcement or whether it's a a new product that can help um marketing measurement. Um with regards to whether Apple and and fingerprinting, I don't know uh which type of players out there are doing that, I believe they've already uh did a very well very good job at at disrupting this and and not allowing actual fingerprinting. And I think from there and on, they will continue with improving privacy, um but I think there's a too much focus on something that really barely exist today in the market. Got it. Okay. Uh so, Steve, on to you. What what is what do you do at at TikTok? I mean,

you're head of client measurement, but TikTok doesn't have its own attribution platform cuz it's not an MMP. Like, what is what is your role in in this ecosystem of attribution? So, the goal of my team is to help our advertisers and agencies get the most bang for their buck from the from the campaign activity on TikTok. So, get the best ROIs, the best brand lift, the you know, the best contributions to to customer experience. Uh and so, I've got a team of um measurement partners that use a combination of first-party built solutions, uh things like the brand lift solution and the conversion lift solution, which is measurement based around incrementality. Uh we also do have things like media mix modeling data pipelines that we share with agencies and advertisers. We've built our own pixels for tracking. got event API connections. We've got a my SDK in production for uh for app-based uh businesses, uh which is, you know, uh the majority of uh the

gaming businesses that advertise on TikTok. And we supplement that with uh a network of uh partnerships that we've established over the last few years. So, uh mobile measurement partners are a big part of that partnership ecosystem for uh for our advertisers. Uh we've also got partnerships with uh research and measurement vendors, people like Nielsen, Kantar, uh LiveRamp, Skai. Uh so, we've got uh third-party measurement solutions where we don't necessarily have access to the data required to be able to uh do the measurement ourselves. Um and also, like for anything, you know, cross-channel measurement-related, uh anyone that wants to do multi-touch attribution across multiple channels or media mix modeling, uh it's not really our place to do that. So, we enable that for clients uh through uh establishing partnerships. They've got broader access to data. Uh they've got their own um uh proprietary information for things like, you know, in-store sales, for example, that uh a lot of the businesses

want to connect to their uh their online campaigns. Uh and we really execute, uh you know, alongside the solutions team uh to really build that kind of uh support for the not only the marketers at the clients, but also their own internal analytics teams and data science functions, um you know, alongside the first-party solutions and the third-party partnerships. So, the way I understood in short is you make TikTok um a preferable channel for advertisers to efficiently reach their target audience. Uh essentially, we demonstrate the opportunity that is available on TikTok uh through uh measuring measuring effectiveness in both brand outcomes and performance outcomes. So, so one thing that I wanted to touch a point on is something that So, we were at the uh the TikTok events during GDC, and you had a you know, a really interesting presentation, which is surprising because you were talking about attribution. And uh and it's not always the the most sexiest topic in in the building. But, you know, you you had a

great football analogies there. So, so that kept the the audience engaged. But, but the kind of like the big point in your presentation was around the last click. And the way I understood it is that you kind of referred to that the last click attribution model is becoming obsolete in And the way you uh you explained it was even a dummy like myself would understand is uh the way the last click is is like who touched the ball last to score. And so, you showed a video of like was it Manchester United playing where the ball was moving was it a different team? not Manchester United. Red team. They were playing football or soccer as the American unit is. And there was a lot of ball passes, a lot of play was being made until finally somebody scored. It wasn't Thierry Henry, but somebody scored a goal. So, that's how far I understand football. Anyways, somebody scored a goal in the end, and they then kind of

put it into a map of all these different channels, whether it's the YouTubes and the TikToks and the and the Instagrams. It's like all these ads were being seen on different platforms. They were all affecting the end result where somebody would download it from, let's say, another game using AppLovin, for example. So, according to this, AppLovin did all the job. But, if you look at the big picture, there was actually many stages going through this. So, uh my question kind of like is is like is this is correct depiction? And do you believe that the uh the last click model is becoming obsolete in today's media landscape? To understand the context behind uh why I believe last click is becoming obsolete, uh uh going back to how it actually came about in uh to begin with, it was actually uh it was created before mobile marketing was was even possible. It's before, you know, uh the first smartphone was uh was launched. Uh basically, in the early 2000s, when digital advertising started

to become more popular, uh for the first time, there was the ability to buy uh or buy marketing uh based on clicks. Uh it used to be that you could only buy audiences. You buy on a CPM basis. So, you know, the cost per thousand people reached. Uh now, you know, all of a sudden, you could buy buy clicks. Uh and that was primarily on uh paid search and uh display channels. So, so banner ads, uh MRECs, things like things like this. Uh and so, because you could buy on clicks, people started measuring the impact based on clicks. Uh so, this is really where last click attribution was born. Uh so, uh and it was incredibly popular at the time. Uh so, you'd pay a certain amount for a click, and you'd see that you'd make a certain amount in return uh via a conversion. Uh very, very easy way to establish an ROI. Uh and at the time, it, you know, it made a decent amount of sense to measure digital advertising in that context uh when it was primarily search and uh and display.

I think the issue is that, obviously, the digital advertising ecosystem has evolved so significantly since uh the early 2000s. Uh you know, the smartphones have launched. We've got mobile advertising. We've got video advertising online. Uh there's social media. There's entertainment channels, short-form video uh channels like TikTok and uh and and the uh YouTube. Um I think the challenge is that many marketers still retain the last click measurement approach uh to digital advertising, uh even though there's a much more complex uh digital advertising ecosystem. Uh and by doing that, what last click attribution is actually doing is it is it is allocating 100% of the credit for a conversion back to the last known touchpoint. Uh and this is where the football analogy came into play to say where, uh you know, the goal that I that I ran through, it was a it was a team goal. There was six players involved. There was multiple touches. Uh and the guy that scored the goal, uh when you look at the one-page report afterwards, all you'll see is his name.

Like, you won't see the build-up. And um the same thing is happening in marketing where, uh you know, there is uh exposure on certain channels, clicks on other channels, app downloads on other channels. Uh they're all connected uh from a last click perspective back to one channel. Uh and what it's doing is it's disregarding, you know, everything that happened in the build-up, everything that happened in the consumer journey. Uh this is not to say that in itself it's a bad thing. Uh but, I think the uh the misuse of last click attribution is where businesses are using it as its primary source of uh truth when it comes to uh digital advertising success. Um when it's used in conjunction with things like experimentation and incrementality measurement uh and from the longer term uh media mix modeling, uh and where attribution, and particularly last click attribution, is primarily used for things like AB testing, campaign optimization, digital channel optimization in collaboration

with the other measurement uh techniques which are uh establishing the roles of channels and not just the overall ROI of channels, that's where it can be effective. But, I think the the challenge is that uh there are still hundreds of thousands of businesses around the world that are only using last click, and that's all they're using to make decisions. And using that in isolation, uh it gives a bit of a distorted distorted view of reality in some in a lot of cases. And um that can lead to um inefficient investment decisions across different channels. And so, that's why I feel it's becoming obsolete, not necessarily that um you know, it's not misleading anyone. It's doing what it's saying it's doing. Uh but, it has to be supported by other solutions to give the full picture of uh marketing success given the complexity of the ecosystem uh that we've got available now. Thank you for uh for balancing it out, very black and white pictures of of everything. So, uh so, Roy, can I go into you like when when you're working

with multiple different clients, uh do you see them in like do you encounter them moving away from last click attribution model? Or or what's the uh what's the trend? The way I see it um and and is that last click is actually I I I I 100% agree with a lot of the things um that he was saying. Um and yet, last click is the most prevailing method. And even if we see new tools which are being released to the market, whether it's SKAdNetwork or also the attribution API uh uh by Privacy Sandbox, that is Android, um they're still relying on last click. So, even even as the future the the current um uh methods as well as future methods which are very likely to be adopted are using last click. Um that is not to say that I 100% agree that um the full picture needs to be comprised not only of

uh last click. But, but um I don't see advertisers steering away from um last click. And I'll give a a few examples. So, again, at AppSite here, we're committed to measurement, to marketing picture. Notice how I I mentioned marketing picture. I didn't mention last click marketing picture. And and even prior to ATT, we've released a first to the market by by measurement vendors uh incrementality product. That is to run experiments on your retargeting campaigns to measure incremental effect and not according to last click. Um there are a couple I've learned a lot from this. And this is, by the way, still an active product, and we also uh developed it further for also for user acquisition, and marketers are using it. Um as a complement to last click attribution. Um there are what I've learned is there are a couple of things that might be a

headwind for uh different methods other than uh last click. Um different methods um today are usually referred to um as incremental testing as well as media mix modeling. And what I've seen is one it's not two times but 10 times harder. It's 10 times harder to actually develop the solution to being able to run it, to being able to optimize based on that. Um so that's one one thing. Right? Um actually there are three. The second thing is trust. Even a third-party unbiased party like us needs to win the advertiser's trust in the data. And what advertisers want to do, they want to look into the data.

And then a lot of the times these methods really find it hard, especially media mix modeling, which is by definition modeling solution. So what's the difference? How can someone validate that my model is true versus something that I just made up versus on based on random coefficient? You would need the advertiser trust. It's really hard to actually validate a media mix modeling or model uh uh for that instance. The third thing that I've actually encountered on encountered on why advertisers uh sometimes refrain from different methods other than last click is actually the most interesting one and and I call it conflict of interest. It's within the best interest of advertisers to get more attributions. Because when a specific ad network gets attribution, it's not just the ad network. It's actually the paid acquisition user acquisition uh uh manager within

the same company that now can show that budget was well spent. And if going back to the example of soccer um on how uh the whole team helped uh uh with scoring a goal actually in many cases it shows that none of the team was involved. Sometimes in incremental testing it shows that although maybe a user saw five different ads and maybe the user ended up clicking it shows that actually if the user was not going was wasn't uh being showed with every single one of the ads and even if the user didn't wouldn't have clicked on the ad, the user anyway would have downloaded the app or converted to the product purchase etc. that this is the desired outcome. So this is something marketers don't want to see. They don't like it. So

sometimes these methods are the unwanted truth on on marketing efforts. Right? So these are the three main reasons which I see are are kind of a headwind for these different methods um and I and I agree that these go side by side with last click attribution, but as I mentioned before I don't see last click going away. This is the the prevailing methods also for for measurement, for signaling, for for attributing, for in feeding into the entire ecosystem as it works today. To summarize it like yes, uh last click is still very important and sometimes it is like in the way you're seeing when looking at a lot of clients, so it's a little bit contradicting to Steve's. Uh it is incredibly important, sometimes the only thing that matters, but nevertheless it's important to use media mix modeling and incremental testing. Can you a little bit describe what is the difference between media mix modeling? So the way I understand it is like you use different media channels,

not only user acquisition, but you might be using I don't know radio ads and and billboards and and TV ads. And then incremental testing is uh the way I understand it, so you so at least like I'm just saying so you can correct me. The way I'm understanding incremental uh attribution or incremental testing is that you're starting to you you have an uh a platform, a data platform where you can see what is the incremental uplift when you're using these different media channels to um to your overall growth of the product. But what's you know, what's the what's the difference and how wrong I am with my understanding? Kick-ass live ops need kick-ass tools. That's where Heroic Labs comes in. They just released a massive update to their Satori live ops platform. And having done my research, I can say with confidence that it's second to none. In a single product, you get everything you need including event capture, player segmentation, feature flags, remote configuration, AB testing, experimentation, live events calendar,

messaging, retention graphing, analytics and so much more. If you're serious about running live services in a game, get serious about the tools you use. Connect with Heroic Labs and make some serious cheese with your live ops. Yeah, I think that the term incremental is kind of being um was taken for this specific method because in the end, again, what advertiser advertising spend aims to do generally speaking is measure incremental effect. Right? What I'm the CEO of a company. If I have $100 in purchases and I'm not spending any dollars on ads. Now I want to start spending dollars on ads, right? I would want uh um every I would want to account for every dollar spent, how much incremental revenue it is bringing, correct?

What actually happens is that let's say I had $100 in in in revenue without any ads, but now I add ads on top of my $100, what actually happens is that let's say non-paid non-attributed revenue goes to 80 and attributed revenue goes to 40. So all in all, again, your supposed revenue has gone up but 20% of the previous uh revenue was generated is now attributed to paid ads. But what you want to measure in incremental the incremental that I mentioned is just $20, not 40. Right? And then this incremental testing is here to try to understand what is to tell apart between the 20 and the 40. Right? Then uh uh differently from incremental testing or not differently, but a different method, this is this is done through experimentation, a lot of what's known

as AB testing, um control and test etc. Um in different methods it's it's it's much wider than that. But then um media mix modeling is actually different method that goes and says, "You know what? Let's see. You've spent this amount of dollars on different ad networks, different channels, different things. Why don't you just input all of that data into my black box machine and then I will output what I believe each additional ad spend or decrease in ad spend actually uh uh resulted in your desired outcome. And this is how I try to attribute with this model that takes in a lot of ad spend data across multiple dimensions and then output how I I I attribute into each and every one of these dimensions. Got it. I See, I'm so stupid. I will need to listen to this podcast probably three times afterwards and

and try to to grasp it. Uh Steve, would you like to add something to like we know we Roy had a had a you know, good point countering your your um your last click attribution point. Would you like to uh you know, come come back with uh with your perspective? Um yeah, I think there's uh yeah, uh Roy raised some uh some interesting points and some interesting challenges as well. Like particularly with uh individuals' KPIs and their own kind of vested interest in demonstrating the value of uh uh the campaigns that they're executing. I think uh we do see that quite a lot actually. It's um it's an interesting challenge. Um but what seems to be happening within a lot of businesses that we're working with is that kind of perception uh is very much within finance teams. Uh but the different perception around wanting to uh evolve the the measurement approaches uh that's primarily been driven by marketing teams. And I think that uh

there's like an interesting kind of like uh internal conflict kind of situation going on with a lot of uh businesses when it comes to uh how to effectively spend uh the marketing uh the marketing dollars. Uh and uh and last click is kind of like you say, getting used to defend actions uh and then other parts of the business are uh looking at alternative solutions to understand how to how to kind of like grow the success of um you know, the marketing investment. It's kind of like um you know, is marketing viewed as an overhead or is it in view as investment? I think the people that view as an overhead essentially rely on last click to demonstrate the reason why they've put that much money into marketing. Uh I think the businesses that view it as investment or a growth driver uh are looking into things like incrementality experimentation to really prove that the business impact rather than, you know, justifying their own their own KPIs. Um and I think that

you know, when you say it's here to stay, uh I agree that attribution is here to stay in one form or another. I I think that now that the the tools available don't need us to rely on last click. I think, you know, AppsFlyer and all that MMPs, you've got future attribution as well as last click attribution, you've got data-driven approaches, you've got multi-touch attribution. I think it will become more the norm that people will start to use the advanced attribution solutions uh and to replace last click. Um And like you say, they will start to calibrate those solutions against things like incrementality. That's actually a really important point. When you when you raise the incrementality compliant attribution, we think about it exactly the same way. I think when it if you look at the three main versions of measurement for marketing, incrementality, attribution, media mix modeling, uh no one needs to be in a situation where they need to pick one. You know,

having approaches in one or more of all of those solutions is actually going to be beneficial. And the the calibration of solutions against each other is where the thought comes into it and uh this is where a lot of businesses find it difficult because they don't have the presence of big analytics teams internally. So, they're having to rely on third parties and that's where the kind of trust comes in. It's like, are they telling us the truth? Have they got a vested interest in, you know, promoting certain channels over others? And it used to be before uh you know, before clicks were even a thing where it was just traditional marketing, the clients and the advertisers themselves, they did a lot of the analysis and a lot of the measurement themselves internally because they had much bigger analytical teams. And so, when last click attribution came in, there was suddenly rapid reporting, a focus on automation. There wasn't as much of the perceived need to go through the calibration processes and

the and the detailed analytic processes internally. But I think a lot of businesses starting to bring those teams back and focus on what can they do to supplement the third-party measurement and the channel measurement. So, you then kind of got like a a combination of different um different areas of expertise measuring slightly different things and the advertisers themselves are doing their own calibration across all the solutions and making their best judgment about um you know, how to how to create the overall kind of source of truth. So, when when we kind of pull back and and talk about like effectively evaluating the full funnel impact of marketing channels um to to really get a clear picture of your marketing performance, is is that that's when you would be using, you know, not only last click attribution, but also media mix modeling, incremental testing. Roy, is that the kind of approach to to use as many as many tools as possible, right? Yeah,

I think in the end and and and again, I want to emphasize here AppsFlyer's specific commitment to these attribution methods. We've been trying very hard to actually onboard and push towards not only last click as a measure as a measurement methods. And and for that reason, I I I was familiar with what headwinds other measurement solutions are are are in the market. I also want to add another note that I think different ad networks actually have different perspective um just by the nature of who they are. So, for example, you have interest-based versus intent-based. So, you have all search, for example, which is very intent-based and they love last click, right? So, you have um any search, you know, I I don't need to name it, you know, Google, Apple, um any any search platform. And then we we have intent interest-based uh which could be TikTok or others that

actually expose. So, I think with regards to the user, this is a very good point that with regard to the soccer example, um where a user gets exposed to a brand to a product within an interest-based type of ad networks and then say later on the user go on goes on to search for that product and ends up clicking on an ad of that product um and then wins the last click. So, I think this is where, you know, media mix modeling and incremental incrementality experience really show um that the actual contributor the actual uh um ad efficacy is to be attributed to the one exposed to the brand. This is this is where, you know, incremental testing kicks in. Um this is a very important distinction

with regards to the different types of ad networks. Also, Steve, something that you had in your in your presentation, if I remember correctly, was the um uh looking at views as the last consistent currency in the measurement economy. And and kind of like discuss Can you discuss how they how the views drive real business outcomes compared to clicks because we've been, you know, really talking about clicks, but now you we've moved already to search and and and other other elements. Uh the point I was really trying to make there was not necessarily that views is the best way to create a foundation for measurement, but it is going to be the the area that is always going to be consistent. Uh I think that, you know, any form of advertising, you know, 50 years ago when it was only really TV and print advertising and a little bit of outdoor, uh you know, the one consistent factor was if you wanted an ad to be successful, people have to see it. Uh and you know, it doesn't

matter how advertising evolves, whether you can uh click on it, download an app, interact with it, engage, like, comment, share, uh a lot of these things that are used as kind of like metrics for success in digital advertising, the one common factor is still if you want an ad to be successful, people have to see it. And so, if that's the consistent factor, that should be the foundation of all future measurement approaches. And things like last click disregarded the impact of views. And so, views are starting to now come back into attribution. And I think that, you know, whatever new measurement methodologies that might come up in the future, I think they'll be built around traditional principles around uh having to include exposure as part as a foundational component of of measurement. And then of course, when we're talking about views, we we need to talk about TikTok, which generates a lot of views.

So, what is what is TikTok's unique role in this advertising ecosystem? Like, what makes the TikTok's ads more effective? Is there a specific audience and like why should you know, let's in this in this particular case, why should game publishers choose TikTok? Not not necessarily over others, but but in addition to others. I mean, our perception is that advertising on TikTok gives a different dimension, a different connection to the to the consumer. If you look at the success of the two biggest advertising channels built up their USPs around connectivity, their identity graph, their first-party data. Uh they knew everything about the people that were using their platforms and you know, they created very robust targeting options and they were really a you know, a leader in digital space in terms of incrementality measurement because they could measure against their own identity. Um Google have obviously been

successful, you know, from a last click perspective with their search channels, but also the you know, the connectivity that they've had across different parts of their advertising businesses. So, you know, they've got Google Cloud, Google Maps, Android, they they Gmail, YouTube. Like, there's there's a huge amount of integration across different kinds of experiences and that has helped them, you know, get the kind of the complete picture of of user journey based on well, primarily cookies, but now with the uh their Android data. I think for TikTok, it's a slightly different USP. So, the advertising on TikTok, we believe most naturally sits the sits within the environment more than any other channel. So, TikTok is an entertainment channel, it's short-form video. We've got content creators, we've got individual consumers producing videos, sending them out. Some go viral, some you know, get hundreds of millions of

views. And people consume a huge amount of the created content. Uh the way that advertisers are approaching TikTok is they're getting in the minds of the creators. So, they're creating ads that are in themselves entertaining pieces of content. Uh and you know, short-form video with the level of creative flexibility that TikTok offers, there's a real kind of natural environment for entertaining ads. And so, what can the impact of entertaining ads be? Uh well, you know, they they tend to resonate with consumers in a in a deeper way. It's you know, we see success across the full funnel, whether it's, you know, initially building awareness through to driving conversions and then even driving loyalty for consumers. I think because advertising on TikTok is you know, the strongest experience of all of the channels available. That then feeds through into different kinds of success with with customers. It's

It kind of like it's no longer about just acquiring people. It's about building a relationship with them because brands you know, when they're using TikTok effectively leveraging all of the possible solutions. So, you know, the organic content combined with the paid advertising combined potentially with the commerce side of the platform in certain markets. That kind of like complete complete set of like consumer interaction and consumer engagement. We believe is really feeding through into you know, the success that we're seeing for advertisers. And then in terms of the audience demographics, like TikTok is usually you know, referred to as a as a network with a lot of younger audience or you know, not not kids but but yeah, younger audience compared to male-dominant YouTube and and the Facebook for the elderly. Is Is Is that a Is that correct

assumption? The demographic of the TikTok audience is actually a lot older than a lot of people would would think. It's We've got a pretty even distribution against the nationally representative population in the majority of markets that we actually operate in. We're not yet in a position where we can release a demographic profile of the audience. We're working with third parties to have that validated. Uh And you know, that information will be released soon in our in our key markets. But the presence of you know, 30 to 45-year-olds for example is a lot more significant than a lot of a lot of people realize. So, you know, we we do have a lot of people that are 18 to 25 using the platform. It's I think we over index in that group. But we don't under index in the older groups. Good good insight because there's always the misconceptions that you would hear and

kind of feel would be worthy. Step into the future of game monetization with Xsolla, the premier partner for game developers seeking to expand their reach and revenue. Xsolla provides a robust suite of tools designed to enhance user acquisition, manage subscriptions, and streamline payment solutions worldwide. Unlock new markets and maximize profitability through their customized, secure payment architecture and dedicated support team. Xsolla, powering game developers to achieve global success. Um so, really kind of moving on to you and and and zooming out from focusing on the specific channels and really talking about multi-channel strategies. Like, what are the sort of proven measurement principles that companies should embrace when when future-proofing their measurement strategies? I mean, it seems like it's becoming ever more

challenging when when you read the news but nevertheless, there are some companies that are excelling at scaling even even today with with new product launches. I can't say there's something that's future-proof. Um but the constant is change, right? And what's what I do think some companies are doing much better than others is adapting, adapting fast. Um and take us for example as as providing that service also with regards to adaptation, there needs to be a lot of it and us giving that service, AppsFlyer that is. Um so, you know, we talked about last click. Even just within last click, there is so much to do and to be done in the future and this keeps evolving. Um so, there's The way we see it is that within even if we focus within last click for a moment and I think we kind of touched on that

but I would want to elaborate. Um One, there's the element of signals. Even within last click, how deterministic can you do it versus how probabilistic can you do it? Then there's the level of confidence in the last click in the clicks reporting. Different ad networks may have different incentives to either report clicks that have never happened or kind of abuse it. We talked about view, impression versus click. All right? So, the confidence of what actually happened. And then the third element is also related to this is the intent. So, within and and and the ecosystem has evolved so significantly in the past few years, not only on the privacy manners but also on the ad formats. You have You have music, you have podcasts, you have video, you have an ad

which is an image. You have clicks, you have playable ads. You have so many different types of ads that actually show different intents of the user and intent is a very important tool within the stack of evaluating whether a conversion should be attributed. So, this is just within last click and within different ad types, etc. Now if we go into privacy changes, etc. So, we've had so many of them, ATT, different GDPR, CCPA. Now we have DMA which doesn't directly impact advertiser but it affects the gatekeeper which then in turn affect advertisers. And and and also as I mentioned, we we also have We know we have the future of what's going to happen with Google Android and privacy sandbox. So, adapting to all of that even within the same

method is extremely hard and as I mentioned, you also use different methods within last click to kind of form a marketing picture. Then stepping outside of last click as we mentioned, you also have incremental testing and you also have media mix modeling which by themselves also evolve significantly with the ecosystem and a lot of them requires cooperation between different parties between in in the market. You you're talking a lot about all these different channels and and and that all do it and I was just wondering like how important is to um to you know, when you're when you're looking at your impact across all these different channels, like how important is this sort of a cross-channel comparison because all the channels are so different and how are you even able to compare them efficiently? Of course, there's ROAS and and CPI and all those sort of numbers but what what other efficient ways there are to compare these channels because we're talking clicks, we're talking views.

All of these kind of play together. Perfect. That's a great question. I think us as humans we need a number. Five different numbers is not helpful to anyone. Um and we need a final number and we need it to be the one that is more the most accurate and the most the closest to truth. But but we need one number to to kind of rely on and there's a lot of There's something within this whole In this whole session, we were mostly talking about measurement and attribution and I want to mention a critical part of attribution and that is signaling back to either the attributed or not the attributed media buying partner um the ad network that is because in the end, optimization of these channels even if you don't get exactly the right

number as an advertiser if you're able to feed into TikTok, to feed into any of your partners which you help them optimize better, target better, show give users a better ad experience, give end users a better ad experience. This really feeds into a good marketing flywheel. Let's call it this way, right? So, you have measurement but an important an extremely important, sometimes overlooked part of a measurement solution is actually signaling back to the the the different ad networks. And then Steve, like something that you had in in your presentation was was this engaged view through attribution. So, EVTA. How is that different than traditional attribution models that we've been discussing? So, our engaged view

attribution model is attributing conversions where the user views the ad for 6 seconds or longer. Um they It has to be the self attribution network for TikTok enabled. And it really gives uh What what we're trying to do with this with this approach is give advertisers options. We're we're viewing it as a proxy for you know, high quality attention on an ad. You know, there's a lot of talk about attention as a metric and you know, our first iteration of that of this is it's to use that 6-second point uh which a lot of the uh advertising competitors that we've got already got these kind of things. Uh and we're trying to introduce the idea of um Okay. If people don't buy into the fact that someone can watch something for a split second and then we attribute an action against it, it's um you know, that that's fine that people are you know, obviously a little bit

wary about that kind of metric. So, we're trying to introduce the idea of we know who this person um spent a significant amount of time engaging with the ad and 6 seconds in a mobile context is um you know, that's a pretty pretty long time to be engaging with an ad if you look at you know, the overall stats across channels with with skippable formats. Uh and we've introduced it as a as a new reporting metric for anyone that is um is enabled with a self-attribution network. And um you know, I think it's it's hopefully going to be maybe the first step into view-based attribution for some clients. So, I think that you know, some advertisers that are uh that have been stuck in last click and you know, have concerns about standard uh uh like one-day view attribution windows and want to see you know, view-based attribution connected to some level of qualified attention. Uh that's the that's the purpose behind

EVTA solution. Okay. From And do you look at EVTA as well through AppsFlyer platform? Yeah, actually I I I'm glad you've brought it up. I I I think I said AppsFlyer, we also again, I was talking about the advancement of of last click and again, view and engaged view is obviously a part of it. And a few months ago, we've launched the beta of what we call enriched engagement type uh which is which is driving the industry forward exactly to that. To say there different type of ads. And the industry cannot stay only within a click versus view, but there are also other types of ads which needs to be recognized when attributing, which needs to be recognized when optimizing towards, right? And and I've mentioned these type of formats before. So, engaged click, engaged view, engaged um and then also just regular click or or view, etc. Um so, these are different engagement

types and and we've seen um many ad networks which are happy to join um to join this this beta and we are really looking forward for the entire ecosystem to follow uh these these footsteps because we believe this will be the next step for for um the entire mobile ecosystem. All right. So, super important. And another thing that is super important is of course robust data connections and and capturing all audience actions in a very secure way. So, Roy, you you look you you know, you talk to numerous different companies. Uh what are the sort of steps to to get there? To have these secure actions or actions secured? What what what do you mean with regards to getting these actions secured? Oh, sorry. So, so to capture all the audience actions securely. Yeah, I talked a little bit about confidence of

whether the reported engagement was actually the the engagement that took place by the by the end user. on which ad networks rely on. Um so, which partner reported what and how do we count on what has been reported? Some different ad networks again have more trust than than other in what they report, but we're also looking into expanding into kind of verifying uh uh the engagement that took place by the different ad networks. Steve, any anything on on your part because you're probably, you know, answering a lot of these questions of of um of capturing all the actions cuz there's a lot of them on TikTok. I think in terms of like the evolution that we see the market going through when it comes to capturing signals is is heading towards much more

privacy-safe environments for for capturing information. So, uh you know, it up until relatively recently, the primary source of capturing signals was either via pixels on websites or SDKs in mobile apps. Right now, people are adopting you know, events or conversion APIs in a much more in a much broader way both web events and app events. Um I think the adoption of APIs is is going to increase significantly. I think that's being accelerated by things like upcoming cookie deprecation, although you know, the cookie deprecation I think is having a knock-on effect on you know, app-based advertisers as well in terms of thinking about how to you know, evolve the um the foundation of signals. Uh and you know, in the next 12, 18, 24 months uh we expect to see an increase in the usage of specific privacy-enhanced technologies for measurement. So, things

like clean room, trusted execution environments, the secure multi-party computation. It's all these things where we're not going to be changing the measurement methodologies necessarily, but we're going to be evolving the way that those signals are captured, those events are captured into a you know, into new privacy-safe environments that gives uh you know, any advertising more flexibility to um have more visibility of what's going on in the measurement. It's you know, these these privacy-enhanced tech solutions will be secure environments where they can feed in more data to match back to the channel level data and get maybe more more insight from their incrementality and their attribution. Um how quickly that will evolve, we're not entirely sure. There's a when it comes to things like privacy-enhanced tech, there is quite a significant burden on the advertisers themselves in terms of you know, a lot of setup to be done on their side. Um but I think ultimately, that's where

we're going to start to head to when it comes when it comes to signals. I think the most important part of of this conversation is is talking about the success stories. So, of course, there has to be a business impact for investing into state-of-the-art attribution and then being on you know, on top of your game here. So, could you talk about some of the success cases where companies have made this investment uh into measurement and through that seeing boost in their ad performance. So, perhaps Steve, if we if we start with you, you don't have to name the company. You can just say what size they were approximately, but it would be interesting to hear some of the success cases. Uh I mean, I think one of the one of the biggest success cases, uh I won't name the business cuz we don't have it published as a case study. It was a um It was a business based in the UK. An online online retailer that had both a web presence and an app presence. And probably about 18 months ago, uh they were very stuck in kind of last click measurement. Um

And they were getting frustrated with their growth rates and um you know, we were speaking to about what could they do to you know, have a put a different lens on the way that their their marketing mix was performing generally. And they went kind of through the process of like establishing foundations for all the different principles of measurement that we've been talking about today. So, they started with a more robust signal setup, you know, adopted event API and conversion APIs. Um and they made the last app SDKs. Uh they changed their MMP. I won't say from who to who, but they they wanted to have a different view of the MMP view of attribution and leverage more view-through attribution in addition to click-through attribution. Uh they also onboarded a third-party multi-touch attribution provider. We actually had a kind of a hybrid multi-touch and media mix modeling um approach to their methodology. And they started doing a new series of incrementality testing on all of the

channels in their mix. So, the conversion solution we've got on TikTok, they were also doing conversion lift incrementality testing on many of the other channels where they had that available. And they really went through the process of consolidating all of that information, feeding it through into what their version of the source of truth was. So, what Roy was saying about you need to have one number ultimately. They kind of viewed it as they landed at a source of truth, but they derived that source of truth based on all the multiple inputs that they were doing across the different measurement solutions. And the calibrated outcome was their source of truth. Uh and they kind of projected decay curves for each of the channels. And so, they established what they estimated to be the point of diminishing return. And then they rebalanced their media mix. And you know, based on the new forms of measurement that they were doing, calibrating across different multiple solutions. Um and the success was like from a campaign perspective, you know, CPAs went down, ROIs went up. Uh but what they actually saw as a business was

their overall revenue growth across the business accelerated to pretty significantly. And they attributed a decent proportion of the revenue growth that they had in the 6 months following implementation of the the new measurement strategy specifically back to deploying that strategy. Um and it's you know, we see a lot of cases where you know, small businesses have quite accelerated success from quite small changes in their measurement. This is a relatively big business already and they've become an even bigger business as a result and there's um and it wasn't really you know one smart campaign it's not like they did a great creative they really went through a rigorous process of redefining how they want to think about measurement and uh and the the benefits were significant and this is all done in you know uh kind of three or four month window they went from they went from zero to 100 in a very very short space of time. Roy what kind of cases have you seen with companies really

investing into their attribution and through that improving their ad performance? Yes yes so I think one of the things that we've been uh uh seeing in the ecosystem is kind of relying on um both uh SK network and then also on on MP attribution and we've seen companies that uh thought that um SK network really covers um a lot or everything and and we've been preaching that obviously um it doesn't not necessarily just because of uh web to app but also different ad networks are not necessarily signing all the ads with with SK network uh signature and there could be different publishers which are not supporting of SK network and there could be bugs and issues which are not transparent because of the nature of SK network. Um and and one of the things that you know um I don't want to promote too much but one of the things that we've promoting we've been promoting uh in the past two years is our uh single source of truth solution which kind of combines uh

uh ad fire attribution together with um SK network into forming a better marketing picture. Um and what we've seen with um multiple advertisers that adopted and started relying on this number is that they see a completely different picture. Um not not just that they see more attribution for all channels but also for specific channels over the others and they were able to drop a CPI or eCPA from these channels that actually made uh really important budget decisions based on that. So they moved to joining the data of SK network together with ad fire attribution um and were able to see which channels are actually performing better than others and then changing budgeting decisions. I wanted to end up this conversation on on kind of like where do you see the future

going with the measurement landscapes and how different companies can prepare for these changes. So everyone we we start with you and and with Steve. Yeah. Um great so I I I think again with regards to everything that happens all privacy really the key is in adapting. Um whether it's um ATT Apple SK network right so we've seen SK network 4.0 hasn't been going well with the ecosystem. Um who knows maybe there'll be fixes for um scan five and then we also have uh sandbox which we know is for a fact is going to come and I think um actually it's better to start sooner than later. I don't know if and when Google advertising ID is going to be deprecated um but I do know that starting to work with sandbox attribution today can benefit advertisers not just in being ready but there are actually some advantages already today that are

better than the tools that we currently have as an ecosystem. Then on top of it we we talked about this so there are different uh attribution methods which are provided by the operating systems which are iOS and Android um and then we also discussed uniting joining the data with MMP attribution and depending on the advertiser whether they're mobile first only or also have other platforms such as connected TV PC web etc. really connecting the dots there. This is crucial to understanding um the user journey what the user has done and what really should be attributed for uh the conversion. This is uh the second thing. Um and then again kind of really re-emphasizing on the importance of having all the data one place one number um you cannot rely on a few numbers when you're making uh marketing decisions. You need one source of truth.

Um yeah. Roy do you see any any effect from DMA because one aspect of the DMA was of course that that um that you should be able to get more data from the gatekeepers. Is that is that does that have any influence or what do you think? I think a really good and important interesting case with DMA that we're already seeing taking place is the different iOS App Stores which can open a gate towards different measurement different uh uh distribution channel for either these advertisers or also different ad networks that might be able to run on that inventory. Um so this is something very interesting and imminent that is coming up from DMA and yes I agree there are also other opportunities which are maybe not as ripe uh but they could come. I don't want to elaborate because this would be very speculative. Um but but yes DMA at its essence is very

essence is very promising although it's also in some cases limiting in the case for example we've seen uh with uh Google asking for advertisers to now send them uh users consent in order to uh to know if they can use that um end users data or not. So it kind of has its uh pros and cons DMA but generally speaking I see it we see it as a very very good thing that is happening to the ecosystem and and for app developers. Steve uh how do you see different trends uh in the measurement landscapes? I I have thoughts in three areas so uh technical developments operational developments and then maybe strategic developments. So on the technical side uh something that we've spoken quite a bit about today and Roy was just alluding to you know the usage of um uh privacy enhanced solutions like the privacy sandbox I think they'll start to become the foundation of attribution uh and a

contributor to the things like experimentation incrementality uh and I think that adoption uh of these kind of solutions will accelerate over the next couple of years. I think that um even for app based businesses I think the cookie deprecation is going to start to accelerate that thinking across the across the whole market uh and accelerate the adoption. Uh on the operational side of things uh I think there will be an acceptance that no one solution can be relied on when it comes to measurement. Um and uh you know this like this idea of you know calibration across multiple solutions uh in the search of one source of truth uh I think that kind of concept will get a lot more broadly adopted and so there'll be the there'll be the realization that um you know you can't rely on last click alone you can't rely on incrementality alone uh you can't rely on media mix modeling alone. Uh really the those multiple inputs feeding through into one overall

source of truth really is the way to uh get the best understanding of uh you know how the uh how the marketing mix is performing. Uh and then strategically I I I think there'll be a more there'll be a longer term viewpoint in terms of the kind of measurement metrics that that we're looking at. So uh right now in the in the measurement world uh a lot of things are very very short term. We're looking at campaign impact we're looking at campaign brand lift we're looking at CPAs we're looking at ROI. Uh a lot of advertisers I speak to now they're wanting to understand okay what's the longer term contribution to brand equity rather than just campaign brand lift and what's the contribution to customer lifetime value rather than the campaign CPA and ROI and rather than looking at engagement what's the contribution to customer experience that they're getting from their marketing. Uh I think that uh measurement is start going to start to integrate more into overall business

success measures uh a lot more closely rather than marketing success measures because I think it coming back to what I was saying uh initially around uh you know is marketing an overhead or is it an investment uh I think that people are starting to see it more as an investment and more of a growth more of a growth driver and I think that the kind of measurement metrics that businesses are looking at uh are starting to connect more to business growth rather than uh campaign efficiency. Really really good advice for for for future um and something that that companies can work on. So thank you so much for for all the information. I have to download and process this probably listen to this podcast for the for the couple of times again definitely not my forte on the on the attribution side. Uh so so thank you for for all this information and as as always your LinkedIn profiles will be you can connect to both Steve and Roy from below ask any questions of

course stick talk actually you don't have to search very far to to find these companies and and and the people uh from these companies. So thank you again for for joining the podcast and um um hoping to see you in the uh the future events. Thanks Misha. Thank you. Thanks for having us. Thank you. You did it. You made it to the end of the episode. As a fan of the show it would help us out if you'd subscribe and leave us a review on the podcast service of your choice. Those stars and likes, while a small action on your part, mean a lot to us and help to grow the show. More importantly, are you a member of the Deconstructor of Fun Slack group? If you have five years or more of games industry experience, go to deconstructoroffun.com/slack and apply to join. Join the games industry best professional community filled with peers always willing to lend a hand, or subscribe to our newsletter to get all the latest insights from the Deconstructor of Fun content creators. Thanks for listening.