ABOUT THE PANELISTS: Sharath Kowligi | Niels Beenan | Cameron Thom | Krystel Bitar

ABOUT ROCKETSHIP HQ: Website | LinkedIn  | Twitter | YouTube


KEY HIGHLIGHTS

👟 Generative AI is increasingly pivotal in gaming, enhancing both creative processes and user acquisition strategies.

🏇🏽 The gaming investment climate has become more challenging, with a higher bar set for securing funding.

🎤 Adapt strategies to maintain user motivation and reduce churn.

🎳 Demonstrating robust prototypes is now critical for attracting investors, reflecting a shift from ideation to tangible proofs of concept.

🎻 The discussion highlighted a trend towards hybrid and casual gaming experiences, which are better able to adapt to current market conditions.

🚀 Changes in player behaviors post-pandemic and new privacy regulations are significantly influencing game development and marketing strategies.

⛱ Panelists emphasized the importance of adapting to these changes to leverage opportunities in a rapidly evolving gaming landscape.

🧭 Future strategies in the gaming industry will likely focus on integrating AI more deeply into development pipelines and exploiting new monetization avenues.

FULL TRANSCRIPT BELOW

Shamanth Rao

Hello everyone. Welcome to this webinar on Gaming UA Strategies for a Gen-AI World 2024 and beyond. I’m excited to have a superstar panel with us today. Please join me in welcoming Niels Beenan, Strategic Partnerships Director at Singular, Sharath Kowligi, Marketing Manager at DieNo Games, Krystel Bitar, Product Manager at Sensor Tower, and Cameron Thom, Chief Business Development Officer at Adikteev. We have a superstar panel, and I’m your host, Shamanth Rao, founder and CEO at Rocketship HQ. 

Before we get started, a quick note about your hosts, Rocketship HQ: We’re partners with TikTok, Meta, and Google and have managed over a hundred million in ad spend for some of the biggest developers in the world including some of the biggest games out there. 

To start things off, this is a very tough fundraising environment —not the easiest fundraising environment in a very long time. What are you seeing as regards to funding and investing? Sharath, if you want to take that?

Sharath Kowligi

I’m usually on both sides of this conversation when we’re prototyping, of course, always interested in talking to investors whose values align with us. But then on the other side, as an angel, because I’ve been in the gaming industry for so long, things are different.

Things are definitely different than they were during the pandemic. Everyone’s numbers were looking great, and it seemed like the gaming industry could do no wrong. We had so many new titles coming out of pretty much every single development hub. Now, it’s very different.

It’s not just because of changes around tracking or anything like that; it’s just a tougher environment. People are back to work, and there are not as many people playing. So specifically, when it comes to raising the bar, it’s higher. Good games are still getting funded, and good teams are still able to pull funding, but making that case has become harder.

I see very few people being able to raise significant sums with just a PowerPoint presentation, a slideshow, or a PDF. I see a lot more prototypes. I have an entire phone full of prototypes right now that are maxed out on good builds that are rising, but I don’t think it’s dried up.

I think there are still a lot of people sitting on dry powder. A couple of things that were perhaps a bit unfairly written off a couple of years ago are now slowly coming back because of the hype cycle changes. And then we have Gen AI, which is, you know, half the topic for today.

There are a lot of startups that have just added two vowels to their pitch deck and suddenly they’re a lot more interesting. But even then, the ones who are just a PowerPoint, who are just an API of ChatGPT, are going to struggle to raise. But folks who have a genuine offering, and there are a lot of very cool things that are coming out.

There are a few that are genuinely going to make people’s lives better. I think those folks are going to pick up funding for sure. So that’s me, but I think that you know, maybe Cameron has some insight on this stuff from his vantage point.

Shamanth Rao

Cam, can you tell us how you’re seeing all of this impact growth and growth strategies for the folks you work with?

Cameron Thom

It’s definitely a challenging time for gaming companies and tech firms in general. With some recent unicorns not panning out, VCs have shifted their focus to profitability.

So they now seek to prioritize profitable growth. But considering the economic context, they’re being more conservative. So securing funding, as you said, was trickier overall. And there’s no indication that that trend will improve in the foreseeable future. I’d say, personally, on the structure of gaming teams, there’s less availability of large UA budgets.

So, the efficiency of teams working together and shifting the focus to retention has been a key message shared at many of the conferences to start the year so far.

Shamanth Rao

Niels, you also talk to a number of developers. What are you hearing? What are you seeing about how the macro environment is impacting growth strategy?

Niels Beenan

I agree with a lot that’s been said here already. In my role, I’m quite lucky because I also work directly with some of the gaming VCs, and I always ask them the question, “Hey, what is the outlook for the next 6 to 12 months?” The reason that I do this is because we have a growth product that is specifically designed for startups and scale-ups to help them track the ROI of their activity. And we always want to make sure that we have all the products in there that these guys need and also that the VCs actually need to be able to see, “Hey, is my investment actually paying off?” And I think there are two things here.

One, what we’re seeing is if you have people that really understand the economies of games—people that have that experience, having worked at developers before, knowing what it takes to make a successful hit—these people have, of course, an advantage. They know how to pitch, they know how to prove a concept, whereas the ones that are more experimental or new to market, I think if you compare that to like two years ago, they’re really going to struggle.

So yes, of course, it’s trickier for more experienced people; that actually might be an opportunity right now. But also, what I want to highlight is, if you look at recent investments in PC and console, for example, I also like to pick up on Web3. Immutable on Polygon now has a 100 million fund today at GDC, which is encouraging. Mobile games are definitely difficult. PC, console, Web3, I see some signs of light there.

Shamanth Rao

To discuss the two vowels Sharath mentioned, how is the advent of generative AI impacting the kinds or genres of games that are taking off? Are you seeing any macro trends in terms of genres, sub-genres, or monetization models that are taking off or not? Krystal, you certainly see data from across very different verticals and genres. I’m curious what you’re seeing.

Krystel Bitar

I’d say it’s a bit early for us to draw direct lines between the trends we’re seeing and the direct impact of GenAI, or AI, at least at a macro level. It’s tough to see that there’s a direct correlation with AI picking up.

We know that AI is impacting many aspects, but most notably, some companies are trying to develop using AI, and they’re also trying to use AI to drive in-game content. However, the notable impact is mostly coming on the UA front because we are already seeing creators being generated with AI.

We’re already seeing AI helping with ideation. But it’s not easy to say that’s definitely an AI correlation. I can give you a quick snapshot of the market. We’ve already released our State of Games report. But I can also give you my point of view on where I think AI will be driving some of these trends moving forward.If that’s useful.

Shamanth Rao

Yes, please go ahead.

Krystel Bitar

Interestingly, this year has been quite an interesting one because you see some sort of recalibration in the industry. Downloads have returned to pre-pandemic levels, but we’ve seen a decline of 10 percent in downloads broadly.

But some genres are way more resilient than others. Hypercasuals took the biggest hit, but casuals are quite resilient. And you’ll see that across both downloads and revenue and you definitely will see the pockets where basically, I mean, Monopoly Go dethroning Coin Master and Coin Looters still skyrocketing in the download space but also on the revenue front. It did decline, not much by 2 percent roughly, but the figures are a bit different.

Downloads overall have suffered, but revenue—you see that APAC, for example, has declined more disproportionately versus others. And more established powerhouses in the RPG, strategy, and shooters—they’re still dominating the revenue space, but they did decline as well by 8%. Again, the casual gaming sector is increasing, and we’re now at 30 percent in total revenue overall.

So you’re seeing a trend that’s driving towards more successful monetization of casual genres in general. And that’s quite interesting when you start digging again. What’s interesting also, what’s picking up, is we’re seeing a lot of like these combination of successful genres that are, like, call them hybrid formulas internally.

Hybrid casual is still a phenomenon. It’s still successful. They’re robust. We’re seeing new successful titles but also like these mixes of what is an appealing formula like Monopoly Go again is a successful one. It’s a casino mechanic, a casual progression. Block Blast is a hyper-casual model but with casual gameplay.

You see a lot of these, like trial and experimentation with different formulas, where genres are no longer pretty much boxed. And this, I feel, will continue in the future of how we look at the gaming industry. We need to redefine how we think about genres in general. On the AI front, where I see it bringing success…

I mean, I think it’s two big buckets basically: the UA one where we already know what to talk about, and a lot of the predictions that we’ve had with the privacy ATT, the deprecation of a lot of our signals and measurement, have impacted everyone. Let’s start thinking about brand.

Let’s start thinking about retention. Cameron just mentioned it’s quite an interesting theme across the board. Everyone wants to be profitable and have a long-term plan for the business. You see this in UA implementations for successful brands. On the UA side, you’re seeing that there’s a lot of thinking behind creating a campaign in the more traditional sense of a brand.

There’s a lot of thinking behind how my brand is and how much it’s resonating with Gen Zers. You can see this theme across all the top wrestling games. Basically, and it’s quite interesting because that was not a conversation you could have with casual gaming companies a long time ago.

There were a few giants that are doing this, but the rest is, that you can’t measure this, so it’s hard to think about it. But this has proven to drive a lot of in-game retention. There’s a lot of, we’re starting to see the start of re-engagement campaigns coming, and that’s also an interesting bit because, also in the casual genre, you would not have had this conversation two years ago.

Re-engagement campaigns have driven an all-time high in downloads for Gardenscapes this year, I think. One of their main creatives was a comeback creator for returned users. So it’s interesting because this is more of the execution of what we predicted would happen by the loss of signals and everything that’s coming to place.

It signals a shift in how gaming companies implement. I believe Gen-AI could help a lot in this space because you could remove a big bucket of ideation, and a lot of iterations on creative implementation. The second bucket, and I’ll be brief, is around LiveOps, basically LiveOps and in-game events.

Again, the top 10 grossing games in revenue have all doubled down on LiveOps in the past year. This is quite interesting because it’s apparently linked to every spike of these LiveOps and in-game events to a spike in revenue, and a spike in downloads. We have solid retention numbers over time, particularly for some of the big ones.

And levels are quite expensive to maintain. And I believe that in general, we’d see, we could see gen AI supporting a lot in moderation, in managing, improving community management, player support in general. And that’s something that could persist in 2024.

Shamanth Rao

There’s still a lot of activity, and certainly, folks could check out the state of mobile report that you referenced for more detail, Krystel.

Just to switch gears a bit but still staying with the two vowels that Sarath you referenced. As you said, you’re on the investing side of the table but also on the game development side of the table. What are some of the ways you’re seeing gaming studios use generative AI for game dev processes? That could be UA-related but also other than UA.

Sharath Kowligi 

I think a lot of it has to do with the team’s culture, which is kind of strange. On the marketing side, UA has always been, you know, you could even do ads that have nothing to do with the main game for many years now, and UA teams have been quite open to doing those kinds of experiments. Game teams…

It really depends on who the founding team is, who the key people are on art and visuals, and even narrative. Like, for example, at Dino, because I work with two art directors who are extremely enthusiastic about AI and don’t have 20 junior folks to outsource to or delegate to, there’s a lot of AI that’s used for early production. It’s a lot. 

And then at the same time, there are other teams that I chat with regularly where they do have these large teams. And then there’s a decision to go okay, AI is what UA people use to make their ads a bit faster without working with tons and tons of people. But then in the game, we’re always going to just use our stuff.

So I’m seeing almost like two polarities develop when it comes to what happens within the game. Of course, my preference is that we are able to use these tools in a good, responsible way. Because we were able to work on core prototypes when in the past we were only really able to work on one, and that’s because our team decided, yeah, we’re gonna use AI, we’re gonna build the basic template, and then we’re gonna make a lot more based on this. And on the UA side, of course, the biggest challenge has been to find creatives that still resonate.

We can now make 50 creatives in the same amount of time that we could have made, let’s say, only five. And even that would have been very expensive. That has really changed the game on the UA side because there’s no doubt that creativity is the biggest lever we have left. And there’s no doubt that AI lets us make more creatives.

What has been, I think, an interesting challenge, at least for us, is picking the right tools. Because there are so many out there. And picking the tools that will give us the kind of quality of output that we want because we don’t want to create gibberish and those kinds of things. I’m still looking for a copy that will beat the very basic plain now exclamation point, but it’s still very exciting to see what’s coming out of something like Midjourney or At Labs or Pika; they are competing with very well-produced, very expensive creatives.

We’ve worked with some very good advertising teams in the past, but the stuff that comes out of the AI is now competing well with the product that comes out of good agencies as well. And that’s very exciting for me. But of course, it depends on which and. It still doesn’t solve the problem of the CPI needing to be below a certain threshold; otherwise, it doesn’t make sense for the game.

Those problems remain, but we don’t need 25 people working on something. That, I think, makes it very likely that the next huge hit games are going to come from very small teams that could really be anywhere in the world. And that’s been true for gaming for a very long time. It felt like a year ago that that was maybe not going to be true anymore. And now once again, it’s true. So I’m very excited about that.

Shamanth Rao

You talked about the volume of creative output, really the testing velocity and how creative iterations just become so much easier. Cam, can you speak to how you see creative testing and creative iteration play out among the folks you’re talking to and among your clients as well?

Cameron Thom

Absolutely. On one hand, generative AI increases the pace of iteration. But on the other, you know, data sources being crippled with SKAN, the process should take longer. That being said, marketers are looking at performance proxies, kind of like CTR, to start optimizing.

And on that front, I know TikTok is a fantastic channel to test creative since it’s very binary. The creative will either pick up and spend or it’ll be crushed and never spend any money.

Shamanth Rao

Niels, what are you seeing?

Niels Beenan

I think our customers are ready to examine their creative data as we combine cost data with attribution data, so we already see, “Hey, like what is the ROI of a specific creative or specific format or a specific idea.” What I find interesting right now is that of course, we can see how a creative is performing, but what AI does is also like put in the “why” is it performing. What Cam is referring to is, of course, iteration testing, specific messages, testing specific colors, for example, like very teensy tiny things that like back in the day teams had to do but now actually like can be taken over by like an AI tool.

That is definitely going to reduce some workload. It’s going to help a lot of developers, and I think small developers as well, really find that one-hit creative. One word of caution here, though, is that, of course, we’re talking about iteration.

So an AI tool is going to be looking at, “Hey, like what has historically been performing and how can that then perform like in the future with this specific format.” But if you look for example at completely new ideas, and everyone hates them, but look at fake ads for example, that is not something that an AI could have pooped out.

So I think if you look at iteration, if you look at like the testing velocity of creative, definitely that’s something that AI can do but actually coming up with an idea or completing a new concept that’s going to be changing like the market, I think here we’re going to still be relying on agencies.

You’ll be relying on creative folks, so I don’t see much happening or changing there.

Sharath Kowligi 

I am seeing some movement in designers and market research, some of the areas that Niel’s talked about. We’ve always been able to get a lot of market research, like reams and reams of surveys, because of the number of people that we can reach.

But then going through the surveys, going through those videos of playtests, that’s been always a bit tricky, and we are seeing, at least you know, with the teams that I work with at DieNo, there’s definitely a step change in how we’ve been able to utilize reports from, say, Salston or even a playtest cloud, all of that. It’s been much easier to find those patterns and identify user motivations and clusters of user motivations. It’s not the same thing as getting the fake ad made.

But if you can start seeing those clusters very quickly as to which game genre is overlapping heavily with the game that you’re trying to build, you get there a lot faster, especially if you’re trying to build something new. I know that our design team has been having a ball.

With those custom GPT applications, it’s certainly helping with narrative design as well. I would not be surprised if dozens and dozens of new narrative games came out in the next couple of quarters because suddenly, one person can output the same as an entire writing team, and it often does take our teams to come up with a good narrative game. And yeah, I think we’re going to see some serious disruption there—in a good way.

Shamanth Rao

Niels, to build on what you said, we can understand why something’s working. I think that’s been something I’m seeing a lot of impact around. So, for instance, we actually have a YouTube video about how we basically took 20 video scripts that we had produced. We put the performance metrics, which is the IPM for each of these 20 video scripts, uploaded this as a PDF with all of these, and said these are the 20 scripts. This is the performance metric. Now identify what’s doing well and why, what’s doing badly, and why, and give me more ads similar to what’s doing well.

And that’s been very, very effective in ways that we didn’t certainly anticipate. There’s a bit of a stereotype that chat GPT if you just give it a stupid prompt, it gives you stupid output. But if you give it an informed prompt like this, I think it can be hugely impactful. 

Sharath, I know you briefly mentioned custom GPTs, that’s another video we’ve made about this, but we have a custom GPT that’s trained on 150 scripts and much the same way that we just talked about. And because the custom GPT really knows what is a good script, it does a shockingly good job of coming up with something that’s pretty much close to being ready to use.

So it’s just something I continue to be astonished by. Speaking of scripts, a big area where we work with scripts is, and certainly, a lot of developers I work with work with scripts around UGCs, and live actor-based ads. It’s a bit of a different beast compared to just gameplay ads.

It’s become easier than ever to make these with AI tools. Certainly, I talked just now about the scripting aspect, but really the development of some of these ads has also become easier. What are some of the opportunities you’re seeing with UGCs and live actor-based ads for games? Krystel, if you want to take that.

Krystel Bitar

You can no longer dismiss UGCs, particularly since they’ve also been one of the common denominators of successful games in the past year and the last two months. The issue with UGCs is that you need to be relevant today, and this relevance changes week over week.

There are common denominators. For example, you could see some themes that recurring across the successful creatives, like their direct CTAs, very cheat tutorials for more complex games, and propped audio. You could see all of these common themes that you could just adapt your videos towards.

But the theme itself is following hypes, and hypes change very quickly. So I do see that custom GPTs could help detect what is trending today in a particular market. Like today, it’s what’s your job interviews on TikTok. And I would guarantee that next month it’s not going to be the same.

So these are the few things that could help the creative team just be on top of things without a massive research team that is helping them. However, that also could be a bit tricky because you need to worry a bit about your brand safety, in line with where would you want your brand direction to go, or whether is it reaching the right audience.

An interesting thing that we observed last year is that casual and hyper-casual games are finding a lot of success on TikTok. They’re spending a lot of money on TikTok basically compared to earlier during the year. But our audience tools are showing that core gamers are more likely to see an ad on TikTok.

Core games do not have short-form video content that is snappy, short, abrupt, and about success versus failure. Potentially, these games could start dabbling in this. But it’s too complex because you have this complex game mechanic, this very long story. Like, what do you pick?

Where do you start evolving these creative concepts? Custom GPTs can potentially help a lot with ideation and creation.

Shamanth Rao

It’s interesting that you talked about what’s trending, and that’s certainly a cornerstone of high-performing creative and UGCs.

Something that we are also seeing is just as important if not more, is what isn’t trending. And by that I mean, just leaning on player motivations. So can we show transformation? Can we show progression, you know, can we show relaxation, create a relax? So I think something I’m noticing is really taking a lot of gameplay-related user motivations translating them to UGC.

That’s certainly something we’ve seen work well. And, of course, the trends continue to work, although you suddenly have to stay on top of them from time to time. Sharath, I’m switching over to you. What are you seeing in terms of UGCs and live actor-based ads, what’s doing well, and what’s changing with AI-based tools?

Sharath Kowligi

I think the thing that’s really changed for me is the pace of getting the raw footage from somebody who’s an influencer and turning that into an ad because the AI tools are able to do captioning so quickly. So with UGC, they tend to do like influencers, especially those who have a certain amount of serious following, they will do their best to follow the brief, but they’re also, of course, conscious of their brand. 

So there will always be a little bit of deviation from the basic script that you gave, especially if it’s, you know, of course, I call it a UGC, but if you have someone who’s really big, you would still get a UGC from them, and those have been very successful in the past for previous games that I’ve worked on.

So they’re going to make some slight changes. You’re going to have to invest a couple of days into turning that into a proper ad—into meeting the requirements for Facebook, TikTok, and everybody else. Or right now, you could just upload it into a number of AI tools, either 11 or Ad Labs or any of these. And you’re done in five minutes. 

And sure, as a UA manager or a media buyer or someone who’s invested all that energy into negotiating this deal with the influencer. That takes up a lot of time. So then by the time you get the MP4 file if you feel like you’re only halfway done, that can be that can be a long week.

But now, with the tools at our disposal, you’re 90 percent done. The last mile doesn’t have to go to two or three other people on your own team. Okay, now we have to turn it into this thing. It’ll be pretty quick. And then, of course, for the A-listers, there’s always back-and-forth.

It’s like hey, this is what the ad looks like. Now do you approve? And it’s very rare that you get an hey immediate thumbs up emoji and you’re done. It’s like ah can you change these other minor things? So there’s always at least one iteration for the bigger. And this was normal in the good old days of advertising as well.

It is what it is. That part is not going to change. But the part that is going to change is the tooling. So I think it’s gotten much faster in a way where I’m kind of almost waiting for like the other shoe to drop going Oh now it’s going to do something to slow all of this down because of other good reasons that I haven’t thought of. But I’m hoping that that won’t be the case.

Shamanth Rao

And I would add to what you said in terms of just adding some of the ways we’ve been using AI for UGC. One is localization. It’s been easier than ever. It’s shockingly effective to localize English to other languages. Text to voice-based videos.

I think users have just gotten used to text-to-voice videos. footage. They know it’s an AI. They know it’s not a human, and it’s just working well. AI presenters. I see again when I say I see this is just based on the performance I’m seeing. It’s not close to prime time but people are getting used to it.

I think the thing that AI still can’t do is, you know, we recently made ads that involve people dressed up as cops, one person being a cop, and another sitting in a car being intercepted by a cop. Yeah, I still can’t do that. You could do somewhat elaborate graphic design to make sure the creative background reflects that.

But that’s not ready yet. But I think there’s just a lot of iterative work that’s Easier than ever. Something we are also excited about is face swaps. I’ve been testing this around very promising and very exciting and definitely excited to continue to test this out.

To switch gears from the letter acronym to a four-letter one SKAN 4. 0 for the longest time worked in theory but not in practice. Now it looks like it’s actually working in practice. It was out a long time ago. It’s just seeing increasing adoption. In ways that you know a lot of people are comfortable plugging into because even the scan three was working before can for is just something a lot of people are comfortable with.

How are you seeing the advent of SKAN 4 impact UA strategy and planning? Niels, if you want to take that.

Niels Beenan

SKAN is not everyone’s favorite topic. I’m going to work with many customers who are more privacy-focused, and many of them are early adopters.

One thing we need to point out is that your success on SKAN4 also depends, of course, on how well-versed you are in SKAN in the first place. So, there are two givens here. What we’re looking at right now is the adoption dashboard to see, hey, who are those ad networks that are actually increasing in terms of SKAN4 adaptation?

We see some programmatic partners now really like making the ranks there. This data is available, of course, on a website. Look for the SKAN4 adoption dashboard, and you’ll see it there. Now, what I just want to point out here is if you look at a partner like Meta, for example, that has been doing very well on the SKAN, then we see that in the ROI index as well, 2023.

Meta is still a bit sluggish now in terms of SKAN 4 adaptation. And we do expect that to accelerate in the next couple of months. But more generally speaking if you want to be successful with the sense and if you want to actually benefit from what SKAN 4 actually has to offer which is an essence of deeper data granularity right?

It’s going to make your life easier. You’re going to have more optimization levers to pull. Then right now is to look at Hey like am I actually set up correctly for SKAN like how is SKAN playing out for me in the past two years? So I think we’re still very early days. This is going to be a question of Hey how big are these ad networks really going to be ramping up?

How ready are they going to be? But once the industry is up to date. I think we’re going to see a lot of benefits there.

Shamanth Rao

We’ve been turning on SKAN 4 for a lot of advertisers lately. Certainly, we’re looking forward to seeing more of the results of, speaking of privacy, you know all of us I would like to think love probabilistic attribution, so especially in gaming, where probabilistic networks do drive somewhat disproportionate impact and results.

It’s been a bit of an elephant in the room that fingerprinting is not and that’s always been a bit of a debate. So how do you see this now in terms of the impact on performance among games? And how do you see this changing going forward? Sharath, if you want to drive this.

Sharath Kowligi

It’s a bit hard to see a scenario where this kind of attribution is possible on iOS for some of the smaller networks like smaller in comparison to Facebook and Google. And then people not using it. So I don’t think that’s going to happen. 

I think for as long as it’s available people are going to UA people are going to try to max those out. It’s also been very hard to kind of predict when and if the platforms are going to stop, put a stop to it. Because they’ve they’ve said the same consistent thing for a long period of time and now it’s been a long period of time where nothing has really changed.

From a buying point of view, I think people are going to be using these networks for as long as they can deliver. Unlike some of the other networks now, they’re able to find higher-paying users. It’s hard for me to see a scenario where that won’t happen.

And it’s also equally possible that you know next Tuesday Apple will pull the plug on these things and no one knows. Yeah. It’s a high risk for anybody buying on these networks and I don’t envy anybody. They’re planning they’re quarterly planning on media budgets if something like this happens.

If something like this happens, it’s going to have a huge impact. The best way to mitigate this has been the things that Niels and Crystal were chatting about, which is that you know you have another game within your game. You have something that makes a niche game has a very broad appeal because that’s essentially what all the mini-games or fake ads do.

It’s to take something that’s really niche and make it very accessible to a large audience. So I think having a good set of fake ads, if you will, is necessary even if you don’t need them at the moment because you’ve got all these probabilistic networks giving you very large campaigns—six digits a day—possible on these networks for the right kind of genres, of course.

That’s where I’m on that. It’s it’s a bit wobbly. It’s hard to sleep peacefully if that’s the only network that you’re scaling on. That would be a problem I think.

Shamanth Rao

At least the way I see it, advertisers are in a less precarious position than even a year or two ago. And it’s a bit probabilistic because Now Facebook has its own probabilistic with AEM. The equivalent of probabilistic, they’re not gonna announce or claim that it is anything equivalent. Google always had Firebase. TikTok just had a SKAN equivalent come out.

With that, I think there’s just less of a dependence on any single channel, you know, whereas a year ago, I think a lot of people were running away from Facebook just because Facebook was SKAN only and the other channels had much clearer attribution.

And now I feel like It’s at least a little bit more stable. And if the platforms do crack down I don’t think the platforms could crack down on every single network out there. I think there’s going to be a question of degrees in there. So that’s what I’m saying. But Sharath, go ahead.

Sharath Kowligi

I agree with you. And I think one of the things that have changed To make everything a little less risky is even when it became really hard and it wrecked all the VO campaigns on Facebook if you had broad appeal in the game or you had a broad appeal creative you didn’t really need VO you would still be able to scale pretty well with MAI.

And then I have to say that AppStore Connect dashboards have gotten way better. So the custom product page. has allowed a certain amount of cohort reporting. The actual Connect itself now especially for Facebook and TikTok you can see a decent amount of tracking. So it is de-risked compared to a year ago but If you don’t have any ads revenue and if you’re only going to monetize with two or 3 percent of the people and those 3 percent within that 3 percent there’s another power per skew where a 0 to 25 percent of the entire user base has to subsidize 80 percent of the user base. We’re still dependent quite heavily on the equivalent of a VO campaign and finding those whales. That is a high risk. And yeah.

Shamanth Rao 

I would nitpick on a very small thing that you said about App Store Connect getting better. I think App Store Connect was terrible. It’s better than terrible now. It was a low bar, so it’s just got slightly more usable. I don’t think it’s in terms of usefulness. It’s still somewhat limited, just based on my experience. 

Neils, coming to you, among the advertisers you talk to and work with, what are you seeing regarding probabilistic? What is the outlook around probabilistic and how it might change things?

Niels Beenan

Well, if you take a step back and look at the user-level attribution itself, this is going to disappear.

So that’s what the deprecation of the wall of privacy sandbox deprecation of keys. And that in itself actually means that if you look at a performance marketeer or a BI person or an analyst You have different sets of data that you can be working with. Like you’ve got your tracker data you’ve got your scan SKAN you’ve got your deterministic data right?

For double opt-in users as well. You somehow need to rhyme that with what is actually paid and what is organic as well. And I think if you look at probabilistic attribution this is called a fingerprinting for what it is right? I think the elephant in the room here is that. If you were to apply a single method of attribution then you might actually come to the wrong performance outcomes.

Now again like at Singular we’ve been working with customers that are just a bit more privacy conscious right? And we also said like Hey like so what can we do to support an analyst or performance marketeer? And we said Hey why don’t we unify all of the data that’s available into like one specific dashboard?

So we can tell you Hey. What is your MMP saying? What are your networks saying? What is scan action reporting? And if you then see that there’s large discrepancies in between those three methodologies then of course like there are some work that you need to do. Now where I’m a bit worried is when I hear Hey Fingerprinting is the main driver of performance.

And I’m seeing case studies come out as well that are saying like look we are working with a network that is not popping up on any of the MMP fingerprinting we’ve been able to lower our CPIs and then propel that network into my top two or top three. Word of caution there. Like I would be as a BI person be a bit weirded out.

If my performance is not coming from, say, the Googles, Metas, Melocos, and Liftoffs of this world but rather from a very small partner, the second thing that I would also like modern analysts to do is in case you are running probabilistic campaigns and deduplicating activity.

So you’re taking the sense of Hey I’ve got fingerprinted signals here or MMP or tracker insults whatever you want to call them. And I’ve got scan activity here and automatically in all of those cases I’m actually going to favor those fingerprinted activities or tracker activities over your scan activities.

That is probably going to be hurting your TikTok activity. It’s going to be hurting your Google activity. And again then we go around and say is that then actually the performance You’re seeing is this correct? Am I investing in the in the right channels? Now we are not here to tell people like how they should be running their campaigns and listening to what you’re saying Hey in some use cases this is the only way that I can actually find those specific users right?

That’s not the role of the MMP here. But what I do want to add here is that if you have specific and different sets of data available, make sure that you track them for discrepancies and understand where the performance and signals are coming from.

Sharath Kowligi 

I couldn’t agree more.

Krystel Bitar

A small parenthesis on this is that this is a forward-looking session. I think what both Charles and Niels are saying is that this is a very risky strategy in general because it’s only going to get worse in terms of signals. If you’re not really training your accounts on either a metal or Google or tick-tock on the right signals and you’re trying to find those users, it’s only going to get worse in the future. So you can just fall into this short-term pitfall, basically.

Shamanth Rao 

And you know, I’ve definitely seen what I think Niels alluded to, which is that I think some of the networks show very strong performance, but they might be taking credit from other networks.

By the way, I have seen this happen with Google and Meta as well: They do take credit from other networks. And that’s definitely something we need to be very, very careful about. Neil, that ties into what you said about taking multiple sources, triaging, and not relying on one single source of truth if you will.

Niels Beenan

it’s a discrepancy. If your numbers don’t add up but if they don’t end up at say 50- 75% you really have a problem. And we do actually have a lot of developers that come away and said Hey like my numbers don’t make sense. And why is that? Well you’re talking a couple of percentage points. You know if you’re a very large developer that’s still you know a significant but then still for your performance operations for the selection of your channels, That is statistically like not relevant but like when it’s above say that 10 15 percent mark then I think you know you should really look at a BI person or a BI team and say Hey do the number a little bit.

Shamanth Rao

Switching to another dimension of privacy and limited data availability. Right. Crystal, you touched briefly earlier on retargeting and how you know I think Gardenscapes is a retargeting example. So, what are some of the ways you’re seeing, Cam, I’d love for you to take this. What are some of the ways you’re seeing marketers approach retargeting lately, given the constraints on iOS of having limited numbers of identifiers?

Cameron Thom

It’s funny how everybody considered retargeting dead and now it’s making a big comeback. A lot of these games are are not getting any younger. But with retargeting pre privacy it was limited mainly by budget as the the core focus was always UA. But now that the addressable audience is reduced there’s still a good amount of opt-in users to be reached through deterministic. 

Some companies are softening their stance. We touched on probabilistic before Apple ended up dropping the hammer on fingerprinting enforcement. But even without that, we’re seeing companies run quite profitably on iOS, shifting more of a budget share to retargeting.

Furthermore, we’ve developed a churn prediction model. This allows companies to start accurately predicting when a user is likely to leave an existing app, which creates the perfect opportunity to either reward them to incentivize them to remain or cross-promote them to another app to keep them within their ecosystem.

And I think, as we touched on the opportunities with generative AI, creating that quick response mechanism within a game to reward a user and increase retention is, you know, the sky’s the limit with creating that personalization journey for each individual user, I would say.

Shamanth Rao

So, there’s still an opportunity for retargeting. As you talked about, Cam, there are very many of these. 

We are fairly close to time. We talked quite a bit about AI tools that have been helpful and impactful in creative aspects. Are there ways in which AI tools are impactful for analysis and data related strategies? 

Sharath, what are you seeing and doing?

Sharath Kowligi

I think one of the biggest things that we’ve been using AI and AI tools Is for training custom gpts on our data schema Because we don’t actually have a full-fledged Data science team so in in many ways I’m the data scientist And we had four prototypes of very different types of games and even in teams where you know it could be a hundred man company with 10 full-time data people but the moment you go to a new game or you bring someone new you’re going to have a long onboarding process where a new person has to understand the schema understands exactly how certain things are nested understand a game.

It can be very time-consuming. And then, of course, there’s always this risk of one person in the company who knows everything about building something. It’ll be a fantastic database, and you’ll have a great dashboarding solution. And then they’ll leave for their dream job at CERN to find the meaning of the universe, and you’re kind of with this massive database that you don’t know how to use because it’s got like, you know, 50 joins to run for a basic-level query.

Those are Huge risks. And there are always really high entry barriers to scaling data usage across a team of any size. But what we’ve done right now, and I’m most happy about, is that we’ve fed each of our prototype schemas into different GPTs from the beginning of a prototype. Every single successful query gets fed into the GPT.

And then we have people with you know very minimal but some knowledge of SQL is still required. Cause you want to make sure that the thing that the GPT outputs is not garbage. Even people with very basic SQL knowledge can get a meaningful query out and can do analysis with it. That’s something that’s been very very effective.

I would say that and the consistent character output by mid-journey have been the two completely. Awesome unmitigated wins for us in terms of how much time it’s actually saved how easy it’s been. And and of course you know it can be exposed to everybody within the company. It’s not something that you need to do lots of levels of access or anything like that.

We’re currently using Google BigQuery and a chat GPT custom GPT. But you can set it up for pretty much any architecture, and you don’t have to worry too much about whether it’s Periscope, Tableau, or even Looker. You can just run those queries on anything. It’s very liberating, I have to say, where you don’t have to go through long schema discussions and all of those things.

It’s really cool how we can switch from game to game within one day, which would be almost impossible in any other scenario. It’s been very rewarding indeed. 

Shamanth Rao

Thank you for sharing. Since we are very close to our time and we are at our last question, this is going to be a rapid-fire go-around-the-room question. Please share one uncommon or interesting way in which you have seen or used AI tools rapid-fire. Name a tool that you’ve used that you found interesting or uncommon. Let’s go. 

Krystel, let’s start with you.

Krystel Bitar

I’ve actually built a customl to assess the housing market in the UK because I’m about to hire a house. Let’s see how it goes.

Shamanth Rao

 Excellent. Sharath.

Sharath Kowligi

I’ve been using one to tell Canadian jokes, so it’s all right. However, translating jokes is always very difficult. I first hit on the idea of doing Finnish jokes because most of the DieNo team is Finnish. Finnish jokes are notoriously difficult to understand via the standard Google Translate, but AI does a fantastic job of it. There’s a whole custom GPT that I built out just for the Finnish Donald Duck jokes. It’s hilarious, but I’ve enjoyed the heck out of it.

Shamanth Rao 

Amazing. Cam?

Cameron Thom

I’ll give a quick plug to my friend Alexei Chimenda who I think will be on your your show as well.

He created a company called Poolday. ai. He’s replicated himself as an avatar and uses that avatar for marketing purposes. It’s pretty cool stuff. I always secretly thought he was a robot, and now I’ve confirmed it. So, it’s cool to see what he’s doing, though.

Shamanth Rao

 Yes. He will be on the show I think, in a week or so. Niels,

Niels Beenan

For me, it’s a different example. A BMW campaign actually had AI art on cars. And that art was actually depending on the specific car and the specific user. And then actually looking at that car. So everyone would actually be seeing a different ad, and I found out really cool.

Shamanth Rao 

Excellent. And this is perhaps a good place to wrap up this webinar. Thank you everyone for being such amazing panelists. Thank you for the attendees. for the registrants. Thank you so much.