MMPs and Attribution: Grow Your Mobile Game Now!

in today's Mobile gaming Market user acquisition and marketing are critical to a game success as part of an effective UA strategy install attribution and the ability to measure ADC performance is absolutely necessary these capabilities and more today are delivered by a kind of vendor called an MMP today we discuss first what is an MMP secondly how do you determine which MMP partner to work with third how has that business evolved and fourth how will it evolve in joining me today to talk about this topic in depth we have first Brian Murphy who is the head of gaming and new product strategy at appsflyer one of the most prominent MMP vendors in Market second Josh chandley who is a co-founder and coo at Wildcard games he runs user acquisition at a studio and is one of the sharpest and deepest people in the mobile mobile games industry and finally we have James oakair who founded user acquisition company bubble eye and who has also created an open s MMP platform

called open attribution you really can't get a better set of people in the entire games industry to talk to about mmps we are going deep and we will provide you insights you can't get anywhere else coming at you right now so stay tuned hey guys welcome thanks so much for joining us I thought we could just jump right into this discussion about mmps mobile measurement partners and start with the very basic question for a lot of product managers or Executives out there who kind of know what an MMP is but kind of don't what is an MMP what is it used for Brian since you are from maybe you could start and then I'd love to hear a little bit of additional thoughts from James or Josh sure thank you JK and guys thanks for joining as well the mobile measurement MMP or mobile measurement partner that phrase and acronym has been used for I want to

say maybe like a decade now people might remember ha offers in SL tune and they ultimately were like the Genesis for developing tool sets like this and at its core at least initially they did two things one was they recognized that an ecosystem of app developers and game developers in this particular context had to work with several ad platforms to achieve the level of scale they needed to gain what they call positive Roi or Ras all right and so integrating that many platforms and I'm just going to use the ones we're all most familiar with which is Google meta snap Tik Tok Amazon appen iron Source Unity Etc that requires an incredible amount of intensive resource allocation to plug into a mobile application directly so ordinarily you'd have to integrate all of their sdks together and that was incredibly problematic so companies like appsflyer and others that are out there in the space would pre-integrated all of those ad platforms into a single what we

call SDK or software development Kip and then give you access to all those Partners all at once from the server so it removed an incredible amount of engineering efforts so that's one thing beyond that in the core business was the attribution where in a very complex ecosystem you can imagine Facebook Google snap Tik Tok they're all going to want to crame credit for driving conversions because they get paid for it so somebody ultimately has to be the authority within the room of all these major platforms determining who gets credit for what across large window s of time with several different touch points clicks Impressions views Etc um bigger engagements smaller engagements and then utilize some more level of methodology to determine who gets credit for what and generally what apps flyer or somebody else might use as the last click model but it's more complex that in a multi-touch environment So eventually somebody actually had to say our methodology trumps everybody else's beliefs and this platform ends up getting the credit for it and abs fer and

others that are in the space have been given that responsibility and then finally at its core Beyond just the install there's a way to orchestrate inapp analytics using events to determine the value of these campaigns so not only can somebody like appsflyer tell you who actually drove the conversion But ultimately what value did it drive Beyond just the install itself if a lot of our customers as you can imagine want to see Revenue events attached to that to show um the ROI of the campaign or Raz so you could see very clearly hey look if I SP spent a million dollars this month on Facebook I got back $1.5 million in Roi or roaz um and therefore that's a positive campaign to continue to invest there maybe the other platforms weren't as successful and those are row as negative and they might deprioritize other platforms at the end of the day if you're orchestrating this properly you could actually spend to the end you only are limited really by how many users are there on the face of the Earth that you could actually advertise towards provided you're doing this properly right so maybe a simple way of

explaining this to the audience I I think they got it but essentially that on the attribution side you know who's respon like you might be running multiple ad campaigns who's responsible for specific installs and then on the performance Side based upon the ad campaigns that you ran how did they perform was it profitable not profitable in in very basic terms and maybe shifting to you Josh like you as a customer of an MMP user of an MMP would you agree with that definition is there anything you would add yeah absolutely I think the first of all James Brian JK thanks all for being here this is cool and yes so wildcard is an is an Absol customer I think that the simplest way of maybe restating it is let's say you're a new studio just soft launch a game you start with UA on app1 unity and iron Source you get one install that day without an MMP all three of those networks will bill you for one install and you will pay for three installs so that is the fun fundamental reason that

we need to use mmps and that is the fundamental reason from that point on when we get to data and revenue and all of these things it is all built on that fundamental incredibly high value proposition value proposition thing that they do which is they they mediate attribution and I think Brian kind of made a point of like we can all argue about how that attribution occurs and I don't think there is a perfect solution because just like you know credit card fraud we're largely now on last click and so oh my God every inap ad Auto clicks isn't that a surprising thing there is no perfect solution but it it provides a framework that we can all build upon and it accomplishes that fundamental goal of I don't buy one install and pay for three um everything after that is gravy and just the the way that mmps have evolved over time right and James as a I think partner and founder of bubble eye right which was an agency and is somebody

who's actually working on an open- Source MMP project any additional thoughts or context one thing I always think of like as a cause root cause for this is why does this not exist on the digital side in on the web and that's because there you have direct links you can call just call an https URL stick anything you want in the URL and the receiving website gets all that data 100% of the time whereas in Mobile when you install a package from a Play Store you don't get to stick any additional information in there so that is always a big break and that is where you know this need for an SDK on both sides you know that you have somebody on the publishing side and somebody on the advertising side that both have code running inside of the two apps is why there this exists in Mobile and does not exist in the digital side and I think that what Brian said and Josh mentioned is mmps have also been changing you know what they do in our Marketplace over the past few years and theyve more and more

Central as far as like a data Hub yeah now I haven't run UA for a while but back around like let let's call it you know three four years ago or so there there was a lot of talk about idfa deprecation and how with idfa deprecation that has the potential to kill mmps now mmps are still around people are still doing stuff but could you guys talk about the current kind of state of the market with respect to mmps even with idfa deprecation and how has the role or you know what mmps do how has that changed from then and kind of like those dire warnings of MMP death to today yeah I I could start I I think there was an initial belief from a lot of people publicly that measurement providers would cease to kind of adhere to the attribution and and I don't want to say predict it but analyze it give it back to the app developer and on some

level there was some real validity to that ultimately they would be more of reporting an analytic solution that you know they would be utilized for rather than doing the attribution I think number one there was a a very hard belief that the deprecation of idfa which was certainly real would persist to the same level of degree that it did four years ago and I don't think that's necessarily true I think the data that we're seeing is telling us that there's more ID idfa dual consent than there had been previously and I think that's becoming more pervasive over the market and so therefore a lot of L data at least there was initially to to has been gained back and so I think it's not as dire as everybody made it out to be and I think users just over time have just become more accepting of the fact that if they you know if they just accept it's not really going to impact them all that much whether or not they want their data to be seen or not seen so I just think you know over time that that the saturation of the loss of those have kind of come back to life to some deg or another at the same time there's a

couple of like I would say peripheral solutions that have come to Market number one k ad Network as as we as people understand it was really the mechanism for the attribution now it was really Apple dictating who was getting credit versus who wasn't it was a very ambitious project but one that was fraught with a lot of technicalities that I think people like apple didn't truly understand and it caused a lot of up people within the marketplace however the major thing that a lot of people predicted was that if scan is assuming everything then what if somebody's able to actually project when somebody didn't didn't consent so probalistic attribution if somebody's able to do it outside of scan it was going to duplicate everything almost the way that Josh just described it before both sides are going to take credit for this and ultimately Solutions either internal or external were going to be needed to determine what was scan attributed versus non-scan attributed and there was a lot of like really meaningful inative ways of doing so and so therefore

Solutions like that to really see what the campaigns look like rather than scan saying it's all falling to scan were necessary to prove marketing campaigns in Roi and so now there now Solutions like that have been deployed in place to overcome that challenge and now people are getting a much clearer picture of what their marketing campaigns look like two other things that I would mention the ad platforms themselves are trying to do their own problemistic attribution so truly another way to gain back some of the measurement that was Lo C by scan and the introduction of iOS 14 how that works with apple is really at the behest of the the the media platform themselves but um or even the app developer in many cases the ability to gain back measurement that was lost to idfa deprecation has come back regressively and then finally conversions apis so you applications that are taking users away from the mobile World towards the web world and then bringing them back into the mobile World utilizing potentially connections directly with the ad platforms themselves Facebook Google

snap Tik Tok Etc from the web or building landing pages and capturing first-party data and Signal there have become another route where they could offset the loss of idfa so there have been a couple of different solutions and some people even think this is going backwards in our world maybe five to 10 years have crept up to offset the changes that Apple forced to upon everybody else when they made this change got it yeah I think there's I Tak taking us like even more towards the Layman actually I think so Brian's answer was super in depth and really crushed the scan piece I think a huge number of networks especially inapp ad networks are honestly are working even more simply and that is that most of these networks whether you're talking about App 11 Unity or liftoff Maloo dsps or even offer walls or whatever they're attributing the same way they were pre

idfa apocalypse or whatever probabilistic attribution which is just we now we just call it fingerprinting because it got like a bad WAP and there was a big marketing campaign behind it and stuff but like what is that core service the appsflyer and adjust and singular provide they they figure out who who caused that install and even with what we used to call deterministic we knew it wasn't perfect because there's been like a rise and click fraud and lots of different things and so what they used to do was collect huge number of signals ranging from device ID to like IP address and other most sure IP address like lots of different signals and the the feature with the most deterministic power was on iOS it was called idfa it turns out if you take away the best data point things still worked but to be honest like Brian and abs and adjust knew that

beforehand because beforehand one in five iOS devices didn't have idfa so this is already being a pretty tuned model and one of the reasons that it seems to have fundamentally worked is it's not like if they can't attribute it you get no data what happens is if you get 100 installs that day you got 100 installs that day attribution is just goinging to figure out where they came from so if those 100 installs generated a lifetime revenue of $1,000 and you spent 500 bucks you did good and what as marketers what our UA teams need is enough aggregate accuracy to know which networks to reward with higher budgets and punish with lower budgets the scale that would be required for that 99% confidence interval where fingerprinting has like massive impacts it is so high that it's outside of the realm of what truly matters for most

advertisers if you're spending like 20 grand a day on appen and 20 grand a day on malocco or whatever it's unlikely that it's like an unsolvable problem that's my perspective at least James do do you disagree no I think I agree with you guys' points about the idea phase and stuff I think that you know I'll defer to Brian's expertise on scan stuff it's super complicated having you know dealt with that a lot at I I think it's pretty complicated I think that fingerprinting is honestly I think it's here to stay something that I've been very curious about in my own stuff is so if you let a developer host their own attribution Services it means that they can send all the traffic to their website which will prevent ad blockers pie hole things like that from blocking traffic but also means that you know it's harder for Apple to have some kind of argument that you shouldn't be able to track your own users because of course inside of a game you're going to track your your users um and if all the

traffic is going to your own you know the domain that your game owns then it's you know what's the argument for Apple to block that traffic right it becomes much more clear that they're they are just acting as a monopoly would it would it be fair to speak a little bit I mean we kind of touched upon this attribution issue but in light of the current situation with idfa deprecation uh we don't need to go too deeply into it Brian but could you speak to how most mmps do the attribution currently and what some of the challenges are so so the the goal the Holy Grail hopefully even Josh's Point even deterministic attribution isn't 100% deterministic but let's be honest here if you get a dual consent conversion a dual consent opt-in user and you get an idfa out of that that's the goal and if you get that great in lie of that probalistic attribution however you want to call it would be to Josh's Point earlier absolutely correct in that look if we don't have an idfa we might have some level of data on this

user from previous times you know with a an internal ID or something like that is there other signal that we might have Beyond just the idfa itself so again like he mentioned IP user agent device setting there's several different components that somebody could actually build into a probabilistic methodology to determine whether or not somebody is who they are and where you can attribute the campaign towards so if you don't have deterministic attribution which has been compromised as we're all talking about right now over the past few years there's other alternative ways of doing it that go beyond um the identifier itself and so the hope is that the accuracy level when you're utilizing something like that can get close to where it was before idfa really took hold to everybody and so in many cases you're utilizing again your own identifier methodology your own inputs as well as the complexities of a multi-touch ecosystem and determining you know where do I offer credit to and who gets credit for what in Li of going to something like organic when you paid

for it you don't want it to be unattributed because now you've wasted an install that's nonorganic and it's actually not only impacting you financially it's also hurting your data because it's telling you what you're bidding on isn't working when in fact it actually is and therefore the bidding algorithms and all the data science you're putting behind the scenes to be successful have now been compromised because of an unclear ecosystem yeah I mean I actually would love to talk about one one core thing that Brian said there which is that deterministic attribution has been fundamentally compromised over the last I mean even 10 years kind start had a lot while ago and part of that is with the last click methodology which is I still think it's the best like default way to do everything and it's imperfect but then one of the reasons why this privacy movement hasn't hurt many networks as we thought is because like the idfa itself was weaponized so it being taken away isn't that important

because the way like when I run UA on App 11 or or Unity what we do is we give app Unity these these beautiful impression and click links from appsflyer and they're able to do things like say well Unity you got an impression a week ago app or App 11 you got an impression a minute ago and a click a minute ago then they open the device we know who who to give the install to go ahead Josh do you think there there is some cheating though with last click where Sans could potentially force a last click are you are you telling me that my Apple of 99% CTR isn't real is that what you're telling me yeah go ahead you mentioned it right before the call the problem of people Auto clicking not people yeah yeah but not not people is the thing yeah not people Auto clicking yeah the other thing is okay so we have these impression links in these click links and the the MMP can

mediate who caused that install cuz okay fine last click is imperfect but you look at time right if someone clicked a week ago and then there was an impression one minute beforehand they can start to do some things and there's like rules and ways to follow that some networks don't follow those rules what do you guys think of media mix modeling or mix media modeling the mmm I just was meeting last week I met the founder of PMC which is a python library for doing medium mix modeling Brian you're nodding I'm sure you know them and I've always kind of thought of it as kind of a little bit of like black magic not very positive but actually talking to him made me kind of change my mind that maybe there really is something there and it's something that is the future a little bit the the the hard part with media mix modeling first of all media media mix modeling has been a methodology it's existed for decades upon decades upon decades I mean it's really how television and their models from exactly goes as far as a stor shelf like in a

safe w i mean that's like what we're talking about here so it's an incredibly deep level of analysis that needs to be done to be successful the challenge that we saw particularly in gaming when we were building incrementality or modeling is that it's not a one-size fits-all solution by any means because the level of depth that somebody like wild card would need to know in their own properties versus somebody like blizzard for example are completely two different worlds they operate in even though they're both gaming companies and they're two different genres um the way they generate revenue and the way they determine value from advertising are totally different so building a model that has the level of depth necessary per customer and then applying that across 12,000 people 12,000 customers like that's a humongous Venture and in many cases companies have not succeeded at it as a third-party software solution because of that reality I I believe you know there it can't be for everybody I will say this data science teams external can actually have great impacts

on app developers or advertisers in general for media mix modeling but specific to them and ultimately would be like a bridge to the data science team they might not have at that particular time and they can build medium mix models that can be relatively accurate and successful but to apply it like I said as a thirdparty solution as a software technology across a wide base of customers the way I described it is very challenging and that's what I've kind of seen over time I do believe that has a place in the Market along with last click I think a lot of ways you can complement one another but I don't think either one is basically like the here all be all like the way Josh described like last click is very valuable it's still not 100% perfect no question about it I I'm gonna just say this is maybe over the top but I I think Mmm like tries to measure the wrong thing for most companies where it it is it tries to measure exactly the impact of marketing and that almost misses the

point because like first of all Brian's right it requires Mega scale it requires like mega mega scale and we already do have a lot of things around that but also like app launched eight goddamn new campaign types this year what are you going to do how how do you physically run test all eight of those models out and measure the impact and then stop marketing like it doesn't work you're actually doing is you're sacrificing performance for accuracy and so when marketers need to pick between like well do you want maximum accuracy or do you want maximum agility all the way down to minimum viable accuracy is like if you can just drunkenly Meander towards a billion dollar valuation that's totally fine all we need is enough accuracy to be able to know who to punish with lower budgets and who to reward with higher budgets then even then is term of like it's just a

measurement piece how do you how do you train them how do you train apple or Unity with an you don't pass back Revenue that way you know so the the the MMP is still like Mission critical that's my perspective at least I that's a little bit too extreme well maybe we can now just talk about the business case and so whether someone in the audience might be a new mobile game developer or for smaller Studios with limited budgets like if you were to characterize how much it would cost and then what can these kind of newer Studios or smaller Studios expect if they were to turn on an MMP how would you guys characterize that are there like and you don't have to give exact numbers but if you can ballpark how the pricing is what the specific costs are on on an aggregate or on a unitary basis think go ahead Josh yeah okay I can I can do that first and speak super broadly I'm gonna be mean to Brian for a second I I would actually say avoid integrating MMP until you need to if you're super and you're in the tech launch you can run Google ads with

Firebase you can run meta ads with the meta SDK you you don't need to get into an annual contract yet or even like a short-term contract because especially if you're running UA in at less expensive GEOS to to Tech launch and MMP at low scale is going to probably end up charging you like three and a half to 5 cents per install as they rightly should your scale is Tiny But if you're paying 20 cents per install that can increase your Tech launch cost by 25% on the marketing front and when all you're doing is drenching your product and users and seeing what falls over the value ad there is is questionable it's as you scale where the MMP is one of the the kind of service providers where as you scale the value tends to snowball from my perspective and so delaying your decision until you have the exact right partner for you it it tends to be the op strategy in what I've seen cuz when you're on one one MMP everyone

will tell you they're trying to sell you a different MMP it's easy to switch but it's not easy to switch nobody ever switches I people hard it is a big deal really only X Studios that are growing and launching a lot of new games have that chance because they can do a lot of new integration and Josh are you like you kind of broadly characterize the cost for an MMP on a per install basis can you speak more broadly about the different pricing models you're seeing in the industry by mmps and then rough ranges if you could provide any rough ranges yeah I mean so the it's another one of those Services where any range I provide is going to be either accurate or unhelpful because it changes so dramatically first numbers you said were pretty accurate right three and a half five cents at low scale you're going to spend like three and a half 5 cents per install at in a million dnu if you're paying that then

good for RS fer that's a really good deal for them but for the the kind of Base core of what you're talking about where we're talking about attribution cost Revenue those kind of features we're not talking about fraud we're not talking about there's an am literally on your slack with a 60-minute SLA who will help you in an instance it's a problem you're not talking about any of the super Advanced Data features just the core like measure your marketing yeah um at like a small to moderate scale you should be expecting to pay about you know one to 2% of your CPI maybe in basically a CPI fixed cost and a per Revenue per like cost so basically there's probably going to be some percent of cost being charged so if you send a million dollars there's

a fee and then some amount of CPI fee like if you get one install there's a fee and that's usually the fairest model um because if it's just CPI Forex is getting a deal if it's just Revenue hyper casual is getting a deal and so how do you how do you kind of right size a package and it really and especially when the the the cogs like there really are cost of good sold for media for mmps so so it it ends up being something like that most of the time but for like a brand if you need to ask the question what is an MMP cost I'm assuming it's going to be a very small game publisher like three and a half to 5 cents per install and then as Services grow perhaps some small some small fee on cost and revenue is usually about what it costs and the reason that you would start to consider that is because your sections are run sustained UA and you don't just need to know if

your servers fall over you need to pass that go ahead yeah yeah so if you're running one ad Network you don't need an MMP as you're scaling up you're running multiple ad networks you need to know what campaigns what ad networks are doing the best for you which ad Network you're running right so like Josh said if you're running if you have Firebase and you're using Google ads you don't need an MMP you don't have to use an MMP right and if you want to buy on mgrl that's your first ad Network cuz you're in Thailand and you want to buy installs from integral you have to get an MMP there's no there's no way there's levels of data extraction I would say too that prevents somebody that might prevent somebody from working with just one alone because in some cases some of the enriched data that comes back to them could be restricted that the advertiser unfortunately does not get but by no means am I suggesting that if they run Facebook alone they still need a measurement provider we certainly have a lot of people that we know that are obviously as you can imagine small Developers that run only on Facebook or Google one or the other and can pull it

off with either one of their sdks and do okay with it they're not getting everything they would want out of it but it's good enough to start from however the moment that you start to introduce a second third tertiary ad Network because you're scaling that's where the problems really start to get pervasive and the ultimately a measurement provider is needed I mean there there's only been one scenario guys where I've seen a company do it on their own and operate this themselves and that was King who was King Crush of course and they did this for about three or four years and then after certainly during around the time of iOS 14 and with all the innovations that they needed to to work on and these are things like fraud like Josh just mentioned before they weren't they didn't have a deep leing technology they didn't have an audience's tool to retarget from they just felt like a lot of that software Innovation they'd have to develop and build from a time- to Market perspective was not possible and from a cost perspective internally it was going to cost more to build it than it would be to have somebody else do it for them and then leverage a third party so timately they did an RFP Etc and then

apps fer ended up taking that business and then ultimately is now King is their customer is our customer so I've seen the challenges of trying to do it internally it's not easy it doesn't mean it can't be done but I will say that you know the challenges that exist at a scaling business the way that Josh just described it become vexing and so ultimately a third party becomes a part of the situation right and getting back to this concept around value to the studio that's using an MMP so certainly the attribution stuff Josh Brian you guys both mentioned fraud and just to clarify that for our audience like let's say you're running an ad Network and there's some fraud you an MMP can help say hey these campaigns that I ran there's a bunch of fraudulent installs detection go ahead go ahead no please your scale is better well I'd say that the self-reporting networks the sanss like the Facebooks and the Googles generally are pretty fraud free but once you extend beyond that then fraud becomes are they where Organics go yeah no I'm I'm I we generally feel that

that traffic because it's owned and operated traffic it sits within their own ecosystem is pretty safe so it's you know generally we don't run into much fraud at all with them again nothing's exact science but that's generally where we come out but when you extend it beyond that to the some of the networks we even described earlier I don't want to name names but fraud does become a reality because now they're tapping into in inventory sources that are not their own fraud will become a a reality because of that so bot dete bot detection click flooding uh install hijacking all this sort of stuff plays itself out in the programmatic world um and so therefore fraud detection if you're leveraging networks such as those are probably something you're going to have to evaluate and determine whether or not it makes sense from like a value chain perspective so that's kind of where we're going with fraud there right so I would say that like to me when you're first breaking into to needing an MMP there are and I've seen a bunch of Studios now do this there's only ever been three kind of things that I've heard in different kind of flavors over and over again fraud is like a big boy

problem like a king problem and it's like incredibly pervasive because you're buying the same inventory and bidding on it in eight different ways and it's like how do you how does that's that's that is the perfect definition of like an environment for fraud to make be profitable to do if you're spending a half a m a month you just this is very unlikely to be a big issue and so the three things that I kind of hear is like layering in multiple ad networks we've talked about that the other thing that people get caught up on and don't realize straight away training data for the ad networks ad networks now rely hugely on your MMP to train for instance like how how you going to run a vo or a Ras campaign if the ad network doesn't have Revenue posting back to it so if you want to run super targeted campaigns that aren't broad you need an MMP then the final thing is especially where like let's say you're a scalein company when

you scale UA what you're actually deciding very simply is are we ready to bet the company on this game or this moment or whatever you should really look at that a lot of ways and make sure you're very clear on that and what that means is you're going to need to do internal data analysis projects where that attribution information is like a requirement for the deliverable you need to be able to understand exactly your LTV curve who it came from from who to scale with all of those things so if you're going to try to go in scale you really need to have been on MMP for at least like B Brian James like good sample size mature to three months out at least to make that kind of a call especially if you're building your LTV models from scratch because you're a new studio you can't buy off of like three days mature data unless you're hyper casual there you go the other place I would say that where does the value chain lie again now JK is one of the one of the I would say

one of the areas where they felt a lot of people felt like particularly gaming developers and in the marketing discipline felt like there was an opportunity in a gap in analysis at the creative level that they were going to that they could gain back potentially in terms of efficiency that they were losing at the bottom of the funnel in regards to the the loss of idfa so so now idfa and then you know whenever Google sandbox comes along who knows when that'll be but Android it's in it right I'd say that basically there's ways to gain back efficiencies to determine at the creative level what's working versus what's not working sooner and you're probably seeing not just generative AI solutions for Creative being developed which is on the production side but also an analytics tools that as well that are basically allowing and helping app developers and marketers determine hey look what of my creative is working best sooner rather than later and so ultimately they're able to see advertising campaigns working much faster I think a lot of

thirdparty solutions and even measurement providers will be providing a lot more services around creative for optimization than they have in the past and I think that's another extension of the value chain of of a measurement Provider from where it started to where it is today and I think that's another natural extension to where you know an MMP actually can build tools and Technologies based on its own data to show value Improvement there as well got it so Josh when you mentioned Advanced Data Services as being additional value from mmps are there other with the AI stuff and the automation I'm a little skeptical personally because it wasn't so long ago when AI was first becoming like a buzzword we thought we could automate away the UA manager and there were some mmps that were offering like literally we'll change your bids for you in real time and it'll be great and that ended up being a lot harder than it it it we thought it turns out a lot of it is like negotiation with your account manager understand understanding integration understanding training

periods understanding you know events that have happened subub bidding whatever it happens to be and then as we think about creative we're now hearing the same thing where it's like okay so there's two things there's one that's definitely working really really well and that is instead of making art with artists making art with robots that that's like Works super well concepting reviewing outputs and business results and creating entire new videos or entire new things and iterating with computer vision Shout Out pool day AI right I feel very uncertain that it will work what it helps you do is increase your shot on goals but then even then hey pool day is actually a success case for the first option because it's still humans like putting things together and bulk uploading and doing stuff it's not like a robot like saying this one worked for that reason I'm going to make a 100 more like it iterate okay the 100 more like

it the the the 10 sub ones performed best and this kind of that had they had this feature to them let's do 100 more like those 10 and just like automating through we haven't yet seen that service full stack like work another one that's working on that is blackbox we haven't seen that like truly scale yet either I think I've seen a lot of interesting service offerings in that kind of area but as someone that's hyper bullish my previous company bubble eyee we did a lot of this you know we did all the sub Pub bidding that you mentioned and all this kind of stuff and it was always something that we realized was that you need a human in the loop and these are decisions that involve a lot of money a great example of like so there are some some new Apper services that we're excited about I'm not going to talk about yet I have not physically used them so I can't voucher them yet but alarming and detecting what's happening below the surface quickly and using forms of AI to do that is incredibly

interesting let's talk about like how we buy you may automate something on I want to hit this day seven row ass increase or decrease performance well what are you doing you're assuming that post day seven you will mature out to this and that sounds super smart right well who's the biggest inap ad Network in the world right now is app L they're like stomping on everybody until a few months ago their best campaign for inapp ads was the day seven Rass campaign so it optimized on day seven Ras it turns out you can't release new campaign types without impacting the old campaign types so they released the day28 model where now you set a day28 Target and it's super deep and what it ended up doing from our Viewpoint was all of our competitors and people that Randy 28 so I had to buy all of the users that matured the best post day seven so even

though we were buying on the same day seven Target our post D7 curve was flattening down because all of the best maturing users were no longer in our campaign we were doing the exact same thing exact same steps exact same rules and because we detected that a human said basically oh no there's there's a problem here the ability to do that is is is very very valuable and so it's not even about to my my point of view it's not even the the the value proposition in my mind isn't like going from a 45% day 28 Target to a 44% it's deflect it's detecting the subtleties of when things are changing in ways that are hard to detect and that is that is where some of the really cool stuff is happening right now in the AI but then zooming all the way down to like the JK's kind of like use case of like we're going to add an MMP why do we want to pay for this better data you need to measure a Ras curve and

you need to understand what type of traffic is performing the best you you need to be able to to just grab that data and see it mature but then a more advanced use case is the ability to have your MMP like literally host some of your data and you can just pull it when you want it allowing your data teams to discover what they need and go back and grab it versus make a discovery make a change wait to regather data big use cases certainly one of the things that I think a lot of companies have figured out or realize is that we can also be sort of an extension of somebody's own ETL environment in that data can be pulled from a storage environment like an AWS S3 bucket or something like that pulled directly in real time with all the raw data attached to it and then make determinations based on that in their own either data science Warehouse or bi system and then ultimately go back to bid and optimize really really quickly so I think in a

lot of ways now some people don't even use the apps flyer AI in fact many don't the data is just stored um locally in some other instance and pulled to a bi system and then uh done in a way that's much more efficient and truly are you are a backend infrastructure almost like I would say like a five Tran uh that's coming from the measurement provider itself so in a lot of ways more were built in the data engineering environment in that regard and I think a lot of companies are looking at things the same way because as a company scales the interface itself is not critical to them it's how it gets into their bi system efficiently and could be done something in real time rather than it being viewed on an ongoing basis and James you could I could I uh bring I have a question actually it was something we talked about uh before the call and I really want to make sure we get to it because I think it's a really interesting topic I'm going to try to re ask the question that Josh asked so Josh maybe you can cover what I miss what is happening with the more and more networks becoming self-attributed and what is the future of mmps if more networks be self iOS there a lot less

becoming self-attributed there's only one now and I I think that's probably the future as well for mean only only one when I run when I run meta I get zero installs probabilistically I I've run publicistic on uh Facebook and it's it's so on Android on iOS on iOS if you have the Facebook SDK in your app Facebook will do their own probalistic matching oh yeah yeah yeah they will and inst scan says it's complete you can't use that for like marketing decisions that's for training data it's not for marketing decisions Google does the same thing if you buy based off of that you're going to lose a lot of money because they're just so bullish I see that I would imagine Apple wants to move that direction as well right they have an MMP they do often require people put sdks in the advertising I guess let's take a step back and talk about like

this the stories of like Sans and I'll say it in kind of a cynical way Brian so what is a non sand well it's like it's this pure beautiful thing like mopub was with this pure beautiful ad mediator where it had no real loyalties where we're just providing impression and click links and we're using those signals to determine who is most likely to have caused the install and then we can actually Tre that install out and that works really really well is it perfect obviously not cuz it's depending on last click and so clickthrough rate is just going up and up and up over time the problem with that though is if you're an ad Network you really want to find ways to convince a mobile measurement partner that you have a higher share of the installs because then your perceived row ass will increase and Apper has or App 11 has literally no negotiating power here because if app stopped supporting Apper tomorrow for

everybody everybody could still buy that traffic like I'm an ad monetized game a game publisher if Apple disappeared tomorrow I to lose Revenue obviously because they're fantastic ad Network and they're super effective but there would still be a thousand different ways to buy my inventory however with YouTube or with blue app Facebook there's only one way to buy that inventory and so I'm sorry Brian but like Facebook and meta can just go to Apper and say well look you got two options option one you let me tell you when I at an install I will pass you the idfa and I don't give a if I got a click a year ago within some bounds sure and you know got a click 10 seconds ago I you tell your customer that I got the install option two is we drop support and your advertis to pick between using apps

flyer or adjust or singular and using meta for a scale platform what do you think your your your customers are going to pick the ability to scale or like a replaceable measurement solution I would choose the former yeah and so all of a sudden well not over over many years is a more honest answer more and more networks that with their own walled Gardens started to demand essentially self- attribution where they would just say we're going to pass the idfa we're going to pass the gate I remember when I first started to scale on Facebook I didn't notice I had to look back but like Unity was one of our major partners and when we scaled on Facebook just installs just steps down um on unity and nothing else had changed we just add Facebook and it was like at the time I was young enough where I didn't understand the why but looking back now I would have solved that problem differently now we have

this new problem where what does Brian do what do mmps do when all of a sudden that your client is running UA on Snapchat Google and meta and all three are saying yeah I caused the install here's the idfa that this poor helpless creature is being bullied by these Mega gigantic ad networks what do you do well Apple solved this problem for us on iOS they instituted a privacy policy where idfa was removed and so all of a sudden this mechanism of passing an idfa from meta from Google to an MMP and saying we caused this install was cut off but one network was given get out of jail free kind of Hall Pass they gave one network the ability to keep passing back idfa Apple search ads seems unfair yeah

it does so how do you solve for self attribution being too easy to achieve you make it a platform play so when you think about Goog Google privacy sandbox Where Do We believe this is going it's we're just removing Sands back down from like um like a ton to just a one again that's the biggest single change and kind of the transient weird issue is all the Sands that were passing back and and basically did not use any kind of impression or clicks or anything else to attribute they were just passing back idfa they lost their attribution mechanism if I run iOS meta UA today in my MMP unless I turn sot on single s of Truth on I will see a big bump in Organics and then three or four days later in my scan dashboard I'll see scan push back for me and saying actually those Organics that we couldn't attribute cuz there was no freaking

clicks was Apple's telling us is from from Apple one thing I'll say guys is that some of the major srns and this started with SNAP and Tik Tok has actually gone to meta and ultimately I can think it's fair to assume like the other srns which really are only one or two others left that I didn't mention have begun to do what we call Advanced srn Integrations in that a lot of them now are passing back much l speciic so they're starting to B basically enable clicks and Impressions like they haven't before to do a deeper integration in terms of like a a match there each one of those is so which networks are doing this and what Brian's talking about right now is the future yeah and so the ones that I just mentioned James are the ones that are doing it today but I do think each one of the srns that are out there will probably replicate this to some degree or another and then the ones that are coming on that will eventually become srns too I would say Reddit is an opportunity you know Amazon will probably get there and then even Discord its early stages I think Discord might have the infrastructure in place over time to become a self-reporting network

but I believe each one of them will do a much deeper integration the way I just described it to get around the challenge that Josh just totally described which is the mechanism for us to actually do the attribution was the idfa and now that's been compromised by scan Apple search ads now is winning all this unfairly to actually make it so if I don't have the idfa at you know at my disposal can I work and find a way to do a probalistic match using that level of campaign detail to show a better result and so I think we are certainly noticing in the market that that's having a major Improvement and it's good to know Josh that you actually are very bullish on that result because we are certainly seeing it and we think that'll become a larger part of the ecosystem that it had in the past maybe we could uh kind of shift gears a little bit James you created this open source project open attribution and given the discussion we've already had about value to customer customers and things of that nature maybe you could walk through what

was the motivation behind open attribution having an open MMP of some kind and what is the current sort of state of that project yeah my original motivation was meeting lots of small Indie devs who Wen probably not wanting they didn't you know talking to them about uh buying ads and telling them like oh you have to use an MMP they would say well I already have Firebase you know I already have product analytics I have you know something like amplitude why can't I use this to do my attribution I already have tracking in in my game and then telling them no it's actually different it it does an extra step that is not done by any of those analytics platforms and there was often a lot of kind of push back from that and it always made me think there's room there for somebody to build something that is a little bit easier to enter for an indie Game Dev who wants to do things a little bit more Hands-On and then another thing was this slow Creep by Apple to kind of manipulate where the traffic from your game is is going right that was something that made me think

not just that and this is something I think Brian I think apps could do this right is that you allow people to use their domain for the tracking right so then from app from Apple or Google's point of view they can't mess with what traffic is leaving your app because they can't tell the difference between which kind of API calls you're making right are you just tracking something in the game so those were kind of the two things that got me interested in it and then I just was looking for a new project to start working working on and I started working on it it turned into you know how many people work at appsfire right and it's you know a few people working a part-time open source so it's definitely a much different scale of a project and stuff right now it's you know we're still working on the sdks is probably the main thing that we're currently trying to get done so building um a web SDK for the landing page building the iOS and uh Android sdks stuff like that um and then hopefully in the future we'll start working at some point with one some ad Network to do some kind of uh integration so we

have at least one ad Network to get started with I don't know how we would ever get someone like Facebook or Google to play but you know we'll see where things go you can sponsor your project anyway so we're running a little bit long there's actually a number of other topics I would love to cover with you but maybe we could end with just future Trends and so there's a lot of emerging stuff I think one of the great things about the games industry is just how much complexity and how many changes that we see when was at Gamescom earlier this year one of the comments that a few folks made to me was that they felt like the role of mmps would change that there's going to be a search for additional value and growth from mmps and so one of the hypotheses was that there will be a shift towards actually analytics and having your MMP become your analytics provider and then there may be a shift into other areas what what kind of changes and potential services or Technologies or things that you think would be the future of how things change

for mmps and Brian since you are working for one maybe we start with you sure I think you know operating is sort of a backend data Pipeline and then ultimately the corresponding analytics that you're delivering is something that I think you're on the right track with there I think a lot of companies will start to make more of those Investments we discussed before like creative is an area where if a measurement provider has the requisite amount of data it can provide optimiz to its customers at that level at the top of the funnel maybe more investment in that in other directions would be some a place where the customers might have lost some level of row as because of idfa being challenged but maybe they can make that back at the creative level and ultimately I think in some ways it's sort of cleaning up the ad Market itself like creative is working for the right reasons rather than the wrong reasons I think there's a lot of opportunity for analysis there that would be pretty interesting beyond that ging specifically you know with the massive amount of consolidation we saw take

place over the past couple of years driven either by a reduction in VC investment after a number of years of of really heavy V VC investment as well as the Privacy challenges that people have been going through whether that be from Apple or from Facebook or other wild Gardens frankly created obviously this massive you know m&a activity that we saw by crazy that went like crazy you know take two buy Zinga and even Sega buying Roo and I'm basically scratching the surface I mean I'm missing one or two where gosh was it super play that went for a billion dollars not too long ago and these are all reflections of a challenge in operating as an independent company over a longer period of time but what was the acquisition strategy you know predicated upon and I always feel it's like three p pillars you're buying companies for users you're buying companies for Content but maybe most importantly you're buying companies for first-party data first- party data is an incredibly valuable signal that can get back to multiple parties and help you

inform decisions going forward sharing that data is incredibly challenging now that we have all this scrutiny in the marketplace so um I think mechanisms to utilizing first- party data valuable first-party data whatever it might be it could be an ID it could be Revenue data it could be fouille completion the first-party data sets that somebody might utilize to show improvements for optimizations as they're running advertising campaigns across any Media or technology platform they might leverage you know as a partner and so but how do they share that data securely without exposing any user level data and so we really believe like tool sets and mechanisms in that vein ABS flyer supports its data clean room in the same way but there that's by bone means the Only Solution that's out there in the market to basically share that data the way I just described it and I think there's a big opportunity to ultimately securely share that level of data that's incredibly valuable that a lot of the investment was made towards in the first place and do it in a way that exists in

a Marketplace and an ecosystem that's only going to get more scrw scrutiny and so I think that's an area where I think that like game developers could get a lot of value out of it I think we're really at its infancy right now but if if you're asking me where I think like there's some really interesting technologies that are out there to support something like this but do it in a much better way than we have in the past right and Brian does that mean that mmps at greater scale with greater amounts of data can then better optimize for their customers relative to mmps at smaller scale I'd say that's probably true the more data you have the larger your data set the more quickly your models improve innovate and get better and so whenever somebody is a at a much larger scale than maybe the other one they're just at a major advantage it doesn't mean that somebody can't build a very Innovative technology that can compete just as well but more often than not you make the bet on the company that has a much larger data set because the data will be more accurate and ultimately the models train faster and that's just a predication of what AI in general Josh what do you think that's

you James is kind of interesting kind of follow up for you with the scale clay kind of question the interesting things I think coming up are going to be there's that paper by Microsoft about on device attribution which is a cool concept so that would be that the device stores a history of Impressions and clicks from ad networks and then when an install happens the install query can can check that history of Impressions and clicks and then makes it own attribution and then calls out so then in that case there is no attribution being done like the cloud it's all done on device I thought that was a cool idea I also think that like data clean rooms and stuff like that is is kind of where things are going I also think my idea I said earlier about branded um you know that that somebody it doesn't always need to be going to apps.com even if it's appsfire customers right it could be going to you know Candy crush.com which I think prevents other third parties from interfering with that traffic um so those I think are some of

the things that are coming up in the when people are trying to like predict the future when there's like a period of change it's often to think about it from purely your perspective and so when you think about apps or an MMP as like just like an analytics platform that's from a very publisher centered focus and we tend to forget sometimes I to forget sometimes that um mmps are not just controlling my data set they're controlling my ad Partners data sets too and to Brian's point it feels like mmps are becoming like what I wrot down my paper was a like a first-party data optimization Hub which is like a very kind of ambiguous thing to say but there are things that are really really hard for us to do and the number number of things that we need to be able to do with our first party data that includes attribution get so

long that even for King even for 10 cent it gets a little bit ridiculous like I do not envy James trying to understand how to support pre-install traffic or CTV traffic or desktop to Mobile Traffic these are really big problems and then you get into like other weird Niche use cases where it's like okay so maybe I add Maloo in a quarter or six months maybe someone's think I'm thinking about that well they have new technology on the ad Network side where you can pass back your first-party data from new installs and it pre-train their UA models there are all these different things you can do with your data that involves passing back that to the ad networks and then passing it back to you the level of complexity of owning that is getting larger and larger to the point where small mmps are having trouble now small specialists in the

space are having trouble now so my kind of expectation is just the number of use cases gets so large that Publishers just lean in even more on the because yeah I get it the the original use case was attribution but what they've actually become now is this mesh they're the sticky glue that like attaches you to all of your growth partners and I that's where the value ad really is now attribution is a part of that but post install events all the different things that they're doing are just a part of this larger first party data like optimization kind of Hub platform kind of play that I see happening all right guys well I think we can end on that but maybe if you guys have one last message to the audience a little bit more information about what you guys do or any last message we can end with that so if you guys have anything feel free to chime in right now if uh you know anybody got laid off

at adjust or you know is looking for something to do we're open source open attribution you can find us on GitHub you want to Fork it you want to make your own version of it it's fine you want to host it you can do anything you want with it just looking for other people to join so no nothing for me thanks for having me same here I a lot of people I've talked to in the past so that if they want to reach out to me Brian Murphy brian. Murphy abs.com and I'm the head of gaming and I primarily work in North America but if you want to talk about anything you let me know okay one fin thing Brian do you have any comments on the adjust layoffs what do I think I think number one it's it's a really unfortunate thing James just mentioned it we you you might know I have a lot of friends that worked at App 11 and and adjust for that matter so to see a an unfortunate downsizing to this degree is unsettling in a very challenged ecosystem in the first place where we're seeing layoffs and downsizing take place every other week whether it be gaming Developers or service providers like an adjust for example I think that

ultimately adjust was purchased for a few different reasons and I think it's there now to serve really app Len's growth as a a media platform and to Josh's Point earlier they're doing incredibly well look at the stock price every single day although it had a hiccup yesterday it's you can't argue with their success I don't necessarily think that a just as a commercial product for third parties is ultimately going to be their major investment going forward and maybe they feel like adjust as a tool set to help them improve their media platform is a much better investment and in many many cases what might happen is they feel like the amount of service that needs to be required to operate it for their main purpose is not necessarily uh what it used to be and so I think this is a very strategic decision and one made off of real hard dollars and figures and I know it wasn't easy to make it but I think in a lot of ways Apple EV makes very tough decisions but at the betterment of the company itself it's success and I think you know it's probably the right decision the way they see adjust as

being that their partner and their company that they Acquired and you know who's to say that's wrong but the good news is is like to James's Point there's opportunities out there if anybody needs any help you know Phil reach out to me like we know a million different people in the market and you know again you have my email address or reach out to me on LinkedIn if I have something available for you feel free all right so James Brian Josh thanks so much for all of your insights and feedback and with that for our audience we will catch you next time thanks guys all right thanks guys yeah [Music] [Music] w [Music]