Evolution of Mobile UA: AI Lessons From Gamelight - Valentina Bailly & Julia Scurtul.

[Music] [Music] [Music] so I'd like to welcome our next session to the stage we have a uh two people I think coming to the stage so I would like to welcome to talk about the evolution of mobile UA with AI lessons from gam light uh Valentina and Julia from Gamel so if you're there please come on up hello dear all I'm very excited and super nervous to be honest to be here

today on this stage because it's my first let's say experience on uh giving a speech on the conference like that so super privileged and super humbled to do this uh I will start with a short intro uh of myself and then I will pass the mic to Valentina and she will carry on uh and about after five slides I'll be back uh so I'm Julia account manager at gameel light with five years of experience in uh gaming industry uh worked at Unity with stop gross Partners before and now I'm super excited to be part of game light and uh on the stage quickly I want to call Valentina it's um uh our lead of BD team and also my dear colleague and friend Valentina the mic is yours okay great um the green button right is the one cing okay so actually it's not working but it's okay yeah can we do something um sorry it's it's a little technical problem it's good good oh yes

so we are back to us uh here you can see pictures here you can see in Real uh so I'm responsible for business development team of game light I've joined the company like around four months ago before I've been in Mobile advertising industry already for 10 years give or take I've been always in the rewarded uh app Discovery field and yeah so today we are going to kick in to speak about our products and to speak about our technology our company is AI powered and we do something that sets us aside from any other rewarded platform which is now present on the market so let me just tell you a little bit about what we are doing and how we are doing that so we are in Game Discovery platform we have our own audience and our own audience those are Gamers to who we are recommending games but what makes us different is that we are recommending only maximum four games per user the nobody's doing it on the market and for

us it's very important that we know exactly to who this app will be presented we need to find the best the most the one who will be driving RW of the user the one who will be um as well like driving rpu the one who is going to stick will be the most loyal user to the game and that's where we are using AI targeting AI algorithms so what actually we've done on the fun note of that the company exists on the market for about a year we are very very young we are not that many we are like about 10 people during the last year right now we are a little bit more at the same time what we've done during this year we've been ranked third for some categories by upser index which is crucial for the small and yeah for the company who been in the market for one year another thing is that we as well um like yeah we are rich you on the ad

spand around like a little bit over 11% for the match free titles by singular and what is the most important is that our AI tool is considered to be the best targeting tool in the industry by. Awards so we are going to speak about this part so how do we do that how we are using AI technology as the publisher ourself as the one who has the data access to all the data of the users that giving their consent we are able to track a lot of things we able to get the data about um app usage data we are able to get the behavior we are able to get demographics and many many other points obviously it's a massive data which we're getting in real time and uh what is the most important like it's not possible to do that manually so that's where our algorithm is King in so in the real time we are able to analyze it all and we are able to select the best we

are able to create the Persona to whom we are going to be pushing exact game so we have only one shot as you know because we need to be showing only four titles maximum so we need to find the best one and that's where algorithm is working it's analyzing it all in the real time and it's picking only humans who will be yeah who will be there who will be attentive so uh the thing is that it demonst like zero human activation because algorithm is working on itself another thing is self-adapting and self- adapting point is the most important why because the thing is that we need to have some data from the uh publisher themselves so sometimes they are not giving enough of data for us to for algorithm to start working straight ahead so there are like some kind of like a delay with the result but the thing is that this algorithm that we came up with it's self-optimizing and

it's learning from the data as fast as it's getting like as much data is getting it is getting um and it will be like self adopting and it will be finding the results so it's self optimization and we are able like obviously with the eye to look at the Historical and the real time data and basically we are just finding the best possible user and with that we are getting the most amazing results and that's my colleague who is going to speak about this and about the case studies here we go uh thank you Valentina Valentina is a BD real BD and I'm account manager she basically brings me the partners and then I'm working with outcomes with all the challenges that we might possibly have and one of the challenges that I see is that um uh Partners very often are using the same strategy uh for AI algorithm versus non- Ai and UI strategy has to be adjusted depending on the type

of the algorithm ad ad network is having and the typical scenario that um I'm seeing is our partners come to us uh for example a new partner comes and uh they say we know what we want we want women 35 Plus for example and sometimes I feel like it's like sounds like Tinder and I say look try younger ones they also can work well you know and this is how it works with algorithm today um because it can show you the new audiences and it exactly knows that for example this guy 18 plus might really enjoy playing your game and spend money there and maybe that lady of 35 Plus is completely not relevant for you because she doesn't like your types of games and um it's um kind of a dilemma for UA managers to be data driven or AI powered and uh yes UA managers are data driven they know their

audiences and uh which is true uh but a algorithm today what it does it just shows you uh a new possible audience and which is important to be open-minded and to take this into consideration because AI uh Power it means actually 100% data driven uh it just analyzes uh data B better than a human being and or non AI algorithm uh because first of all AI Gathering lot of data including playtime Behavior demographics device manufacturer and this reach data set fits our algorithm which uh is designed to maximize RW retention uh rpu basically uh the next um the next uh challenge that we are facing is overcoming this resistance to CH to change uh for Ed net n works with the AI

algorithm it's first of all to persuade UA managers that um try to try AI targeting uh once they are persuaded then it's a big challenge for UA managers themselves to go back to their leadership and decision makers and to persuade them um not to limit their campaigns and to basically rely and test AI targeting uh which leads us to the next uh slide of case study because case study usually helps in this process of persu persuading how your uh how how the competitors did it and uh first um um case study we want to uh share with you it's um go games um our partners uh they uh uh they advertise with us their title RPG and strategy one bloodline Heroes of lethos and uh they um chose two GEOS to start with which was us and UK and uh

again we persuaded them to use AI targeting which also took some time um and resistance but they trusted us uh with their goals of Rose day 724 they uh as a result uh we managed to overachieve their targets and we reached 43% and Ros D 30 you can see was overachieved by more than 30% which is is which was amazing for them and persuaded them to stick to AI targeting uh the next case study actually we won a platinum award and best mobile marketing award from uh for this case um it was um top grossing um RPG mobile game publisher uh so they again they chose us uh we they allocated 3 million budgets uh for for a month uh and received

150 ,000 installs uh and again uh with their goals of Ros D 712 they reached 15.5 and rpu incre they saw the rpu increase of 400 more than 400% for day seven and uh day 30 as well uh basically uh what we saw what we see right now that for even popular and saturated games uh it's um really working well with our algorithm with AI algorithm uh in general and it helps really our uh Partners to overachieve their rpu uh we're going to the next one lessons and conclusions and basically this is our last slide today um first one is become AI power just tested uh this is what we always say when we have our partners coming uh to us uh just test it and uh those who decided to go

with manual targeting and then finally decided to test AI targeting they see at least 40% uplift of theosis and um second one um it's very important to understand whether uh a traffic provider has AI or non AI algorithm because many say they have an AI algorithm but in reality uh what it it might involve uh AI translation or creative tools but uh on its core it can be just um uh use manual targeting or no targeting at all and they can focus Solly on the uh highest beers um and there are some questions that would help you to identify where your traffic provider has got AI targeting or not one first one is um how long is this learning phase because with AI algorithm it shouldn't take weeks uh what we see we see

immediate results with our partners they can achieve their uh kis um after the first closed cohort uh secondly also very important is the budget uh the budgets uh for the learning phase should be really minimum and um uh besides this another one uh question to ask is um uh how how are there m uh ads you know are they personalized or not how does this personalization work because uh for us um each user we have just one shot as Tanya said uh Valentina said we really tend to there is no the same ads that we show to two users it could be only in case if they have the same app usage data the same interest and stuff like this which is not possible uh each user will will uh have their own

personalized ad four ads that they will see uh and uh as a result they will play the game because they this is something that they like and um actually at uh advertisers they will get uh loyal users who like playing with high LTV and we see that um for example the rate of tutorial complete is about 96 98 uh% on our platform which also confirms that um it really works uh the last question um also to ask how does optimization work AI should uh help to make optimization easy um when you make change in your campaign it you have to see the immediate result immediate impact and uh this is what we see also the uh setting up the uh campaign should take few minutes um it shouldn't take like half an hour or something and for UA managers it it's uh it gives opportunity to focus on

something strategic um basically AI makes our life easier and helps us to focus on the important stuff and I want to finish with this uh slide become AI powered and just test it thank [Applause] you thank you very much now um we are uh probably have time for maybe one question we should take one question does anyone have a question for our like friends now I know um oh there's one at the back we'll take one question here and then we'll we'll probably have to move on but thank you very much for that presentation hi R from Carrie first uh just what type of AI are you employing can you give us some idea what type of models are being used underneath so that's a technical question or is it like data collection that we are taking no no what type of models are they deep learning models are they more old school models what type of models are they simple feedback

optimization Loops uh yeah um that's that's a great question that to be honest I am a account manager yeah maybe with the technical stuff I'm not so but I can definitely follow up if you can come to me I I will take your emails and I will follow up on your question I'm really not ready to answer it thanks great thank you great thank you well tell what let's let's move on thank you very much for the presentation that's very interesting and if you do have any other questions of course you can look up Gamel light on meet to mat and and ask you there so thank you thanks [Applause] sir