How Automated App Ads can Help You Achieve More for Less

uh finbar's going to join us and he's going to talk to us about how automated ads can help you achieve more for less hello fema uh hopefully you're connecting okay great good to see you um so um without further ado i'll hand over to you perfect thank you very much let's get this screen sharing going um one thing uh that that's you know let's look at silver linings and stuff it's very interesting uh way to be doing this but you know i'm happy i didn't have to get up at 6 a.m for a flight this morning so i said how it works how very true amazing is that um sharing is that look does that look alright for everyone that looks great thank you perfect amazing so firstly just to introduce myself my name is fimbro i'm a vertical strategy lead for gaming here at facebook

today i want to talk to you about an exciting new solution to help developers drive sustained user acquisition performance at scale on facebook which is automated app ads and the purpose of this today is really to give an overview to everyone but also to provide enough details so that if at the end of this it's something you want to start doing you have enough information to understand the context and what we're trying to do with this um i'm going to assume that most people here are already familiar with facebook's app ads uh so we'll be discussing with automated app ads why we built it how they differ from existing ads what to expect from us and then providing an overview of how to set yourself up for success and some of the pitfalls to potentially avoid um so we're gonna start with a high level overview and then we can move into some more detail as we as we go on so let's go um automated ipads is basically our next step towards enabling app advertisers of all sizes to effortlessly effortlessly effortlessly effortlessly effortlessly achieve the highest performance from their ads automated ipads is built with our powerful machine learning

and automated systems to maximize the power of your marketing we understand that for any developers your business growth depends on performance when it comes to your app ads campaigns every euro needs to have impact on both your app and on your growth as a business for app marketers sustained performance is dealt with co however performance can fluctuate and sustaining performance when you scale is often complex even on our own systems and products i think anyone who's run campaigns on facebook understands that to really keep it going it's it's okay to get success as you start scaling that becomes more complicated there's a lot more work involved and especially for for companies that may not necessarily have the resources uh to troll a lot of individuals at that and that's what we have in mind when we're talking about our automated app patents achieving top performance often requires advertisers to spend much of their time monitoring and tweaking multiple campaigns in addition having so many campaigns running at once can create audience overlap which can actually prevent budgets from being

allocated in the most efficient way and can negatively impact performance so again uh we'll jump into that more with some more details later as i'm sure you all know no matter the size of your business or your marketing team the effort it takes to sustain performance can make it more difficult to scale since facebook's introduction of app installation 2012 we've continued to build ad solutions and tools to further empower app business growth it's our goal to empower advertisers of all sizes to effortlessly achieve the highest performance of their ads with this in mind we've set three goals for our continued investment in our vision number one performance we want to drive more of the actions that you value number two scale we want to help you to achieve greater scale and sustain performance and number three efficiency we want to enable you to do this more efficiently from create from creating campaigns all the way up to managing overtime with the introduction of automated app

ads we've taken a huge step forward to making our vision a reality automated ipads is a way to run app install ads on facebook that enables you to achieve more with less and delivers at the same time better performance greater scale and more efficiency we built automated ipads to help your business grow through better performance in driving more of the results you value greater scale at sustained performance for your ads and streamlined campaign and creative management to make it more efficient than our current app advertising solutions and automated ipads is our next step to help solve for sustainability of performance and i think today automated ipads is really where we have a huge we're really excited about this is on the efficiency in the state and the sustainability point today automated ipads is available free app install objective and we'll roll out additional features shortly so when choosing between automated ipads and your business as usual facebook activity you can decide which is the best choice

for you your business's usual manual facebook app ads or automated ipads i will go over the differences between the two of those as well as considerations to help you decide what to use based on your business skills so what is the difference between automated app ads and the app installed as you're already familiar with essentially where automated ipads are concerned as mentioned less is more automated app ads oops automated app ads require fewer inputs have simpler campaign structures that are designed to reduce and even eliminate overlap and they rely heavily on machine learning to sustain and improve performance they also have the ability to test a large number of creative company combinations dynamically so as you can see from the the comparison on the left and right um manual app ads have multiple ad sets and multiple ads within a single campaign whereas automated paths it's one campaign one outset and one ad um and we'll go into a little more detail on this how does it work so first the delivery

system collects a lot of inputs from automated ipads sites so when you're creating manual ads you have a number of over 40 inputs whether it's creative bidding optimization or targeting for automated app ads we've reduced that and simplified it so that it's up to 15 inputs and and as you can see from the from the from the graphic here uh with manual app ads you dictate and and you select the targeting and the placement whereas with automated app ads um the machine learning will you don't select targeting replacements and machine learning will actually figure that out over time so um for automated ipads the idea is that the system starts delivering and it shows your activity in different placements at different times of day and over time the system then learns which of your assets in which placements end up performing the best and it starts delivering more of that so woody every subsequent impression the delivery learns the best way to deliver the ads and the system then can take any if you if you wish can take manual adjustments then into account further on

um as you can imagine on on automated ipads as well i think an important point is that uh it's just kind of constant tweaking and having all these controls at the start that can make actually sustain performance very difficult uh over time so the idea with automated ipads is it really is a it's a simpler interface it's simpler set and then really heavy heavily leveraging machine learning um to then drive that performance versus having people on the back end or or people on the teams tweaking things making and making changes so um the test of automated ipads is the right product for your user acquisition activity on facebook uh i think the first thing to do is set up a split test to properly a b test automated ipads versus your business as usual when they overlap if automated as op-ads is determined to be the winner then we suggest shifting your spend over to automated ipads it's recommended to try this in a few test markets first before shifting 100 to automated ipads and if automated ipads and b and business usual activity are delivering to the same audience business as usual activity may interfere with our

machine learning and with the learning phase to prevent automated ipads from achieving the best performance which is why we we suggest when testing it it should be a clean a b test versus just setting the boat up and seeing which ones and if your activity currently relies on playables dynamic language optimization or placement asset customization um even though these aren't supported by automated app ads we still recommend a b testing so that you can see what gains are possible from automated ipads without these features if on the other hand you're looking to drive mobile app engagements or you're using um dynamic ads then automated ipads probably isn't suitable for you at this time so uh let's talk about some of the best practices for automated ipads firstly how do we set it up so in order to create an automated ipad campaign you go into your ads manager as per usual and select app installs your as your objective uh this time when you've done that you'll see two

options one is to create a manual app ads campaign which is what you'll be doing to date um and the other option is create an automated ipad's campaign now at this point it's it's important to know just as a tip there's currently no way to filter by automated ad campaigns in your in your ads manager account so when you're creating it we highly recommend referencing automated app ads in the name of the campaign so that if you want you can sort and filter them for reporting yourself something as simple as putting aaa in the name of the campaign should allow you to do that uh in terms of targeting as discussed there are fewer options available to manual ad campaigns so the only targeting options available are country and the operating system and language is an optional targeting option that's also there you can't have two automated app ads that are targeting the same country language and operating system with the same optimization type running at the same time and we'll talk a little bit about what it is yeah shortly um it's

and it's basically you know in summary the whole point in that is to allow automated ipads to succeed um and to improve performance so we don't want we don't want multiple campaigns starting the same audience cannibalizing potential conversions it's also possible to target multiple countries per campaign however as mentioned since dynamic language optimization is not supported you should group your countries that have the same primary language or add additional language targeting uh this will help to ensure that people receive your ads in the right language if you need to test your targeting while this can't be done in automated app ads this can be done with manual app install campaigns so you if you need to do any testing you can run that with your manual ads and then incorporate any of those learnings into your automated ipad campaigns so once you've selected your targeting you'll choose between one of four optimization goals available app installs app events value or a new goal created specifically for automated ipads app installs with up

events while your optimization goal should be based on your kpis creating multiple campaigns in order to run a mix of optimization goals is a great way to diversify your activity app install ads with app events um so firstly currently the the only app event available for this is purchase events so we could call it app install with purchase events i want to talk briefly about it just to get because it's because it's new just to give you an idea of what it does and so if you have a specific cost per install and cost per event goals then this new optimization type enables you to balance app installs with app events in a single campaign one thing that's worth noting is that compared to mobile app institute mobile app install objective the app events with app the app installs with app events objective might deliver higher cost per installs but also um but lower cost per purchases and similarly if you're comparing it with app event optimization you might find higher cost per purchases

but actually lower cost per installs so we're really aiming for the sweet spot combining performance and scale and i highly encourage you to to test it out i'm going to leave this slide up for a little bit um to allow everyone to read it but effectively what we're doing here is providing guidance on which optim which automated ipad optimization goals you should think about using based on your current optimization targeting and kpis so choose the targeting choose the optimization based on on what you're currently doing today but as we already recommend a mix of strategies for manual app ads a mix of strategies for automated ipads is also worth exploring so in other words um if you're or if you're if you're exploring multiple uh if if multiple of these options apply to you then you can also test multiple options with automated ads again i just want to reiterate the restriction that no two automated ipads campaigns can be targeting the same country os language with the same optimization type so you can have multiple automated ipads

campaigns targeting the same country operating system and language as long as they have different optimization types so you could have one automated app ads campaign that's targeting that's looking for installs one that's looking for installs and purchases and one is looking graph events and one is looking for value that's fine next i want to speak about creative um which is really a very powerful feature of automated ipads and in fact given the the simpler interface creative is going to be the main lever that you'll have to drive performance each automated ad campaign allows you to provide up to 50 videos or images and will then combine those with five headlines and texts allowing for more than a thousand combinations of creative headline and text for the best performance we recommend or we suggest uploading a maximum number of assets and ensuring that the assets you provide contain a healthy mix of video static and dimensions to ensure maximum eligibility for most placements and we tend to see worse performance for automated ipads when activity doesn't

make full use of all the creative slots available over time you'll want to think about replacing your worst performing creative assets with new ones now by words performing that doesn't necessarily mean lowest volume because what we might find is the automated ipad system might find a pocket of your audience who respond really well to one variation the creatives you're using um so we don't want to necessarily if it's working if it's performing if you're happy with results we don't necessarily want to replace that just because it's not driving the volume so we're looking for more okay well what are the what are the the lower performing uh creative assets and subbing them out and introducing new creatives to refresh the activity similar similarly to manual ipads we recommend introducing new creatives or videos about two to four times a month in order to combat creative fatigue and keep things fresh so um when you think about your mix of 50 creative assets in addition to condition considering the dimensions as i mentioned there are fewer ways to approach this and really put creative at the center of your acquisition strategy

firstly um which team or concept is strong do different themes resonate better with different audiences and how do different visuals and copy work in a variety of markets so what we're saying here is you might find that for some games uh for some audiences the competitive elements of your game really appeal to them for other people maybe if maybe there's an element of customization that really appeals to them so by using the the pool of 50 create creative assets to mix the um the themes and and you know what you're saying about your game it will allow our system to find the right audience for those creative assets and having a strong diversity of creative themes and will give more flexibility than having 50 iterations of of the exact same same creative and again creative will be your main lever so it's really important to focus on the creative you're using and in fact i'd say that one of the the bigger shifts that we expect to see with automated ipads is because of the limitations on the

targeting and it's where we're we're leveraging machine learning to do that instead it should allow more time to really focus on creative and to build strong creative concepts see what's working on that site so now that we've set up an automated ipad's campaign how do we see if it's working what's the reporting look like the first thing to note is that reporting for automated ipads is simplified and only provides breakdowns based on what you can actually control a campaign creation so what this means is that for automated ipads since you can't select age gender and device or or placement these breakdowns are not available on the reporting side for automated ipads uh you will be able to do breakdowns by country creative and by any of the buy time options are available any non automated ipad campaigns in the account will retain existing reporting and again just just to reiterate this is where it's really important to have labeled your campaigns with automated ipads in the name if you want to actually filter them and find out okay well which of your campaigns are

currently running automated ipads the current way to do that is is by naming only and so if you want to review creative this can be done on an ad set level so you can review performance by reviewing the dynamic by selecting the dynamic creative asset breakdown before we wrap up i just want to spend a little time discussing some tips and some watch outs and i also want to ensure that everybody here has a very good understanding of what the learning phase is because the learning phase is really critical and a key factor to what we're what we're trying to do what automated ipads hopefully you'll be familiar with the learning phase and understand what i mean when i say the learning phase but in case anyone isn't effectively what the learning phase is is that when our is still exploring the best way to deliver your ad set actively trying out different audiences and placements and as the term implies it's not yet stabilized and it's still learning so ad sets typically exit learning phase as soon as they're stable their performance stabilizes which tends to be after approximately 50

conversions over a seven day period um the 50 is is a standard 51 and 49 there's not much difference between them so it's it's more of a guideline than a hard rule but i think the key piece is that what we find sometimes is that if people end up changing their activity or making decisions based on results that you're seeing while they can while the activity is still in the learning phase what ends up happening is that they're making activity before the system actually understands who the best audience is so we'd always encourage everyone even with manual ipads to make sure that your activity is exiting the learning phase first and after you've exited the learning phase then make a decision based on those results um with automated app ads the learning phase is the same but we won't actually flag if your system is if automated app ads are in the learning phase or not so i think it's important to to to incorporate all of these recommendations up front so firstly we want to make sure that your budget allows you to at least reach

50 conversions so whatever you're setting up and secondly making sure that you choose an optimization type that we expect will happen within at least 50 times within a seven day period so um in addition uh like that you might also find is that your budget varies depending on the country in the os that you're targeting so setting an appropriate budget and having conversion events that are frequent enough to hit that learning threshold the other thing to note is that editing any changes you make will reset the system and enter it back into the learning phase now that doesn't mean that we should never make any changes oftentimes the benefit of making those changes and outweighs the cost of going back into learning phase but what we would say is firstly wait until you've exited the learning phase before making changes and then secondly when you are going to make changes make all of your changes at once and so that means that you're not going repeatedly back into the learning phase

that's hopefully something that most people using manual app ads understand at this point if you're going to try automated app ads um it's just worth keeping in mind as well listed here are the edits that will reset the learning phase so and before we finish up and take a look at some of the testimonies and case studies i just want to share some final watch outs uh this is really a list of things to avoid when using automated app ads some of it you'll find is an overlap from or manual ipads recommendations as well so number one try and avoid small audiences this is a system based on machine learning it needs it needs information in order to succeed so the general guidance is for app event optimization and mai goals to have at least 4 million people in the target audience and for uh value optimization to have at least six people now i think some people ask is okay well if i if i can't control who the audience is how do i ensure that my how can i control the audience size i think what this really comes down to is a few things it's how you're

combining country and os targeting um if the country you're targeting doesn't have four million or six million people then consider bucketing that country with additional countries who speak same language similarly if the population is approximately at that level then maybe consider not not reducing it any further with potentially language targeting unless you need to do so uh campaign overlap again this is very related to to the um the learning phase piece what we're seeing there is with automated app ads we want to make sure that um if you're running automated ipads and targeting the same audience with business as usual activity that could negatively impact the performance of automated ipads so what we want to see here is um try and reduce overlap make sure that multiple activities aren't targeting the same audience uh make sure that your bidding allows you to to achieve what you're looking for i can't emphasize this enough but really make full use of the creative of all 50 creative assets available so another issue we've seen is when people

don't use all the creative assets when they're using small number of creative assets and we've we've seen poor performance in those instances and and then the frequency of edits so again with automated ipads there's no need to make changes all the time let's try and make sure that we're exiting learning phase um and then making all the changes at once so uh just finish up here very quickly i want to kind of go through some testimonials and case studies associated with each of the objectives that we call that start around performance scale and efficiency so firstly in terms of better performance i just think this example of merca scatter slots is is great right where they've seen 90 lower cost per action 40 higher impressions and a 30 higher return on ad spend i'm gonna run quickly through these just i think we're we're kind of at time uh from a scale point of view uh ilian cube rush adventure using automated ipads so 150 more installs and 29 higher return on ad spends and then finally for um efficiency i think this quote

from applovin really sums it up and and this is exactly what we're trying to do here this is probably you know really critical if if your company doesn't have the resources to invest in in monitoring and tweaking these all the time but even if you do it's what we're hoping to achieve with automated ipads is to allow your activity to um to run and to invest more time into creative ideation the market research so uh that's it here is there's just some final tips that i'll leave up there and i think with that we're done yeah we've got loads of questions and not even at the time so let's try and pump through them pretty fast um so obviously we have the elephant in the room that's coming up with the uh idfa being removed um and i suppose you could argue there are also legacies around the gdpr and various other kind of um issues that still kind of hang over us how will um triple a for one of a better phrase how will that work

post on idfa no a great question i think i think there's still a lot of uh just a lot of stuff still up in the air and and our understanding is that the impact on automated app ads from idfa changes won't be mature like there's no reason for us to believe anything different than the impact on the normal business as usual activity so um i mean it's not a great answer to the impact my personal feeling is that what we what we end up with is and i think the general trend has been this from kind of the reason why i mentioned gdbi the general trend is that we have less and less confidence over the identity of the individual for good and bad uh bad because we can't give them things that they were necessarily interested in but good because obviously we're respecting privacy so i think it means that we we still have to do the work but the outcome is going to give us different results and therefore we have to interpret them differently is that a fair assessment um well what i think is kind of interesting is i think it will require a change absolutely i think exactly to your point

where people are really used to that that you know okay i know exactly who this is i know what's going on and i think people will will likely need to get more comfortable with looking at some of this and aggregate i think that really plays into a lot of the strengths of automated ipads where we're no longer where we're kind of almost relinquishing really between that and effortlessly relinquishing a little bit of control um on that side and i think if we can start looking at the net performance and saying hey is that actually working for us and then actually i think that becomes less um sinister it becomes less kind of like crazy and also it means it focuses on what really matters because we're looking at uh behavioral uh and cultural choices and rather than trying to sort of target individuals which i think could lead it has in many cases led us down some rabbit holes um so um yeah another one again i've only got a couple of minutes so which one will go how do you maintain low cpi when you're dealing with daily user acquisition budgets um yeah well so so this is and this ties into the last

uh point you made a little bit as well which is one of the i guess real drivers between this automated application we found is that we've had a set of recommendations for the longest time normal manual what we found is a lot of people aren't adhering to the best practices i think the temptation to to go in and fiddle with you know what's changing and stuff like that i think it's almost too trunks too strong to resist so in terms of achieving lower cpi's number one and if you're able to really hone in on your targeting uh i think the the risk is no i don't care about mid value users i don't care i wanna i want we want all of us to target who we think are the most valuable users and that our definition i think is going to lead to higher cpi's especially as you acquire those users and that pool starts to to look to look different and mess with these new people with automated app ads the system has the flexibility to find you know higher cpi or higher quality right there and the last thing i'd say is when most people say cpi they're also talking about performance so it's it's it's it's very rare cpi for

the sake of it right it's very easy to get low cpu installs so the um the app install and event uh optimization goal i think could be a great option here yeah no i the the bit i was going to say there is absolutely that which is that you know actually saying we want low cpi i can totally understand that it's the thing we want but actually we want effective users for our game and it's not about like finding the the cheap users it's about finding the audience for our game i think the moment we kind of forget about the the um the purpose of this process is to find our audience and we just try to get numbers we lose track of what success looks like i think 100 and if we found in in multiple tests in the past that oftentimes developers will have a very strong idea of who their audience is and who their core audience is and that's that's fine that's absolutely correct but what a lot of people miss is by focusing on a core audience there might be another audience who maybe isn't at the same level but they're also valuable they're also roi positive and so hopefully automated

ipads by using the right creative mix you can continue attracting your core audience but maybe broaden the appeal as well yeah and i think that broadening appeal is really important because if you're going to scale then often you'll find that different aspects of the game will scale to different audiences anyway uh so last question tanveer ahmed is saying how quickly and how much budget is required for a triple a you know for your triple a um that's the difficulty is triple a means something in games we have we have that internally all the time so with um you know auto ads um how much is budget and how quickly can you optimize so i i think the key piece here is the learning phase so what we're looking for is is we need a minimum of those 50 events within the seven day period so how what the budget is going to look like depends on what your usual costs are for that those 50 events and the more the better so not a great answer but again i think that's like what we've seen performance start to improve after that stage and then the more volume we were able to drive into

it the better it gets uh that makes a lot of sense i mean i think in my head well it seems like there was still going to have lots of value in the manual ads particularly as we're testing concepts you know i think you know that whilst there's some fantastic stuff in terms of optimize automation if we're just trying to see is this idea any good i think we're going to still be using manual personally but as we start scaling our game we want to start broadening our breadth of audience and i think that's where we can discover who actually is interested in our game rather than what we who we think is interesting absolutely the last thing i'd add on that is that like one of the key functions of automated ipads is to really kind of muscle up on the exploration part of the activity so really finding not just finding the users and and continuing to convert to users but finding more users so so making sure that it's always on the lookout for additional pockets people are not i guess not optimizing too quickly um at the expense of sustained performance vimbar thank you so much for that that was great

you