Table of Contents
The TL;DR version
- With iOS 14.5, all ad networks (including Facebook) will lose the ability to automatically optimize paid UA campaigns based on which ad creatives ‘perform better’ (known as driving more users and in-app events as reported to them by the advertising apps).
- This will mean losing the ability to provide a/b testing for ad creatives.
- It will create a world where a large portion of UA spend (billions of dollars globally) is flowing to inefficient ad creatives that perform poorly, simply because the ad networks won’t be able to know better.
- A new creative optimization methodology needs to be implemented for apps and games’ mobile marketing and UA teams, that, before being deployed on live campaigns understands ad and app store page creative performances.
- The strength of your paid UA campaign will be as strong as the ad creatives you provide to the networks, and they won’t be able to “save” you from poorly performing ad creatives.
- To be proactive with your growth and KPIs, prepare for that leak to happen and start implementing a methodology that “qualifies” your ad creatives to avoid significant UA spend waste.
The ‘I need the long read’ version
Any day now the rollout of iOS 14.5 will start, and like every Apple iOS update, the vast majority of Apple iPhone users will be using the new OS in a matter of weeks.
When this happens, we as mobile marketers will lose access to in-app behavioral data used for app advertising – the fuel that has driven much of the industry. Yes, one of the core reasons for the huge growth the mobile app industry has seen in the past several years (both apps and games) has been due to unprecedented access to high-quality users by ad networks selling hyper-targeted inventory.
In the past, from the ad network’s standpoint, it went like this:
“Are you looking for a new player that’s likely to convert into an in-app purchaser?”
“No problem, we can show your ad to these 100,000 users that just yesterday made an in-app purchase in a similar game. They’re installing a new game every week, plus they’re going on vacation so they’ll have some spare time.”
This “skill” that ad networks will cease to exist. As you’ll see below, there will be a few workarounds, all of which are temporary; Apple made it very clear that they don’t approve of the usage of such data for ad targeting (read Apple’s A Day in the Life of Your Data if you want to get a sense of how they view the world).
Winning by Creative Optimization
Creative optimization is the practice of improving the creatives and messaging within ad and app store creatives, to get more high-quality users to stop, be interested enough to respond to the ad, and eventually install the app. The reason I emphasized “high quality” is because it’s not just about getting more users to respond to the ads, it’s about getting more high-quality users (however you define them – retention, usage, registration, in-app purchases, subscriptions, etc.)
If your ads and app store page are getting more installs without driving the underlying value KPIs you’re also measured on, why should you care?
For years now, the practice of creative optimization has been transferring from the minds and hands of mobile marketers and UA people to those of Facebook, Google, and other ad networks.
They have built such sophisticated machines for crafting audiences that are likely to not only respond to your ad, but also to become high-quality users, that we didn’t have to do much but feed that machine with a stream of new creatives.
The machine would quickly de-prioritize a subpar creative, move budgets away from it, and into ads that do perform. As marketers, we could simply observe the output of that machine and understand which creatives work the best for our target audiences.
If we’d hit a home run, the machine would automatically divert budgets to those creatives and we would be able to reap the benefits of seeing our KPIs go up. Click through rates, conversion rates, and eventually revenues through ROAS figures.
That being the situation, we had no idea why these creatives work best. We didn’t even know who the Lookalike audiences were that the machine had crafted. How would those people look if you would put them in the same room? Would you be able to associate all of them or most of them with a certain demographic data point? Age? Gender? Location? Household income? Interests?
According to many experts, you wouldn’t be able to.
So we were living in a world where things simply worked as long as we fed the machine with more and more creatives and observed what worked and what didn’t, without understanding the nuts and bolts keeping the machine running.
Fast forward to now – the machine is about to stop working or break beyond repair. This is of course Facebook and Google’s ability to get our ads in front of an extremely high-quality audience and then surfacing the ads that lead to the most “value” we signaled to them, be it dollars in revenues (VO campaigns) or certain events like registration (AEO campaigns).
And we don’t know how it really worked. Yes, we knew Facebook was crafting audiences based on their behavior in other apps and websites, but what were these events? Why did they lead to more quality users?
Which means we’ll be left in a weird place. We’ll still need to drive new users to our apps and games, and we’ll still be measured by our ability to bring users that drive revenues. But we won’t have the ability to measure revenues accurately, and we won’t be able to rely on Facebook or any other ad network to tell us which creatives are “working”.
What will the impact be on paid User Acquisition Campaign Performance?
First, it’s important to understand that the impact of iOS 14.5 won’t be distributed evenly between all apps and games. Those that will be impacted the most are apps and games that were dependent on super-users, the fraction of the population that would be your perfect users.
The most impacted end is apps and games that were dependent on in-app purchases or subscriptions. The lower end of the spectrum is apps and games that were dependent on in-app advertising revenues, especially apps and games that have a target audience that is very broad, like hypercasual games or utility apps (everyone needs a flashlight or a scanner).
The impact will not be very clear to see, but it’ll be there. Many ad networks, as well as MMPs or Attribution providers, will continue to show “modeled” data about campaign performance. What will be modeled? Ad set and ads creative level metrics such as installs, conversion rates, and conversionValue counts.
That modeled data will have varying levels of accuracy but it won’t be great.
Why won’t it be great? Let’s unpack that:
- The only source of verified install attribution will come from Apple’s SKAdNetwork for users that didn’t opt-in to be tracked both in the advertiser app and the publisher app, that’s a pretty steep ask. Installs will only be reported at the campaign level.
- The attribution logic of Apple’s SKAdNetwork recent version, 2.2. is extremely skewed and complex, you can read this piece here to learn more. The bottom line is that the install credit will go to the last ad the user viewed that was signed with SKAdNetwork, with a preference to a certain type of ad Apple endorses (Storekit-rendered ads, basically a rendered app store page as an ad impression).
The first part means that no one, and I mean no one, will have accurate data about how many installs each ad creative has generated. We can make assumptions, sure, but nothing close to any real reality. Making matters even worse, install data at the campaign level will be sent 24-48 hours in delay.
The second part means that even the verified installs that the ad networks and MMPs will be getting as data will be extremely skewed. One anecdote here is that Storekit-rendered ads will have an attribution window of 30-days and all other ad formats will have an attribution window of 24 hours.
So if a user viewed a Storekit ad in the past month somewhere, it can get the install credit even for organic installs such as search installs, top chart or category chart installs, or even featuring installs. It can get the install credit even if the user viewed a regular ad after they viewed that Storekit ad.
The strength of the assumptions the ad networks and the MMPs will make in trying to understand the modeled data for each ad creative will be even weaker, as the underlying data coming in from SKAdNetwork won’t truly reflect the actual campaign that drove that install.
So the impact on paid UA campaigns will be hidden behind layers of tough to interpret modeled data. Some campaigns will still show a positive ROAS on the surface of it, but the truth is that you won’t really have good enough data to reach those conclusions.
The combination of the broader targeting that your ads are going to get, and the lack of ability by Facebook and the other networks to divert budgets and impressions to ads that perform better, will result in less quality installs.
It could also result in fewer installs, but this depends on the economics of the advertising markets and whether CPMs will go down (driven by lower spending by brands that aren’t able to prove a positive ROAS), allowing brands to increase budgets. Opinions are split on this one, but we’ll see that effect very soon.
You will see that effect if you were to look holistically at your app’s acquisition, retention, and monetization performance. What you’ll see is that the cohorts of iOS 14.5 users are less likely to retain and monetize but are costing you roughly the same to bring in.
Bottom line – the KPIs you’re measured on are going to be impacted, and that impact can’t be concealed by vanity modeled metrics.
Why will Facebook stop acting as your automated creative optimization “friend”?
In this era of significantly lower user acquisition campaign performance, you will also face Facebook and the network’s inability to de-focus ads that don’t perform. All of a sudden, Facebook won’t really know which ad creative they should (or shouldn’t) show to drive installs.
They simply won’t have the data (even installs per ad creative) to do so.
So instead of trusting the ad networks to tell you which ad creative is better, you’ll be operating in a world where your KPIs are tied to the strength of the creatives you feed the machine.
You’ll feed the machine bad creatives, and the machine won’t have any alternative but to treat it almost like your top-performing creative. Even the ability to a/b test ads on Facebook will be disabled because Facebook will have no data point to treat as “success”. No Installs, no nothing.
You’ll have to introduce a strong methodology for qualifying ad creatives and finding the best ones, so instead of trusting the machine to do the work, you’ll “force” it to operate better because you feed it mostly with top-performing creatives.
We’ll get into that in a bit.
Why will your App Store creatives become much more important?
A very overlooked area of this creative optimization problem lies with the intent level of users landing on your app store page have. When paid traffic is highly targeted, users come into your app store page with generally high-intent (you could easily see that when comparing your app store conversion rates between Facebook traffic to low-quality from ad networks running rewarded interstitial ads).
If user intent is generally high, your app store page, the second step of their install journey, has less “work” to do to convince users to install or re-enforce their install decision.
By definition, as the ad networks lose their ability to drive a similar quality audience as they did pre iOS 14.5, the average level of user intent will be lower.
If it’s lower, your app store page will need to do more convincing in order to get users to install. If your app sells shoes, instead of getting people that have to buy shoes this week, you’ll start getting people that are somewhat interested in getting cool new shoes, but who’re just in the explorative state of their shoe buying journey.
The second type of users need much more convincing, and pretty different messaging to get them to install, than the “I need it NOW” audience.
So in order to maintain your app store conversion rates for paid traffic, you will have to strengthen your app store page and adjust messaging from what worked well for hyper-targeted audiences to what will work well for a broad audience.
How will you know if your creatives are working?
You don’t know what you don’t know. Being dependent on SKAdNetwork dashboards visualizing data that is either skewed, incomplete or inaccurate, would lead you to:
- Think your UA performance is still intact when in fact, it was significantly impacted.
- Make the wrong decisions when allocating budgets to the wrong campaigns/ad groups/ads.
The only way to actually understand the impact of the new reality of very little user-level data would be to measure your aggregated KPIs over time, identify the points in time when you made creative changes, and observe the impact on the top-level KPI.
For example – updating your ad creatives to fit a new messaging strategy and observing overall App Units, Revenues and Retention as they get reported by App Store to Connect before and after the change, eliminating any other change that might have taken place around that time.
This will be the only way to really identify and quantify the impact more broad targeting and a lack of automated creative optimization has on your growth KPIs.
To begin with, you could mark the date of iOS 14.5 rollouts, avoid very significant changes in that time period, and observe what happens after that change, compared to a period prior to the change. Doing this, you can then answer:
- What’s the impact on my app store conversion rate for paid traffic in the various channels?
- What happened to cohort retention and revenues?
- What happened to the absolute volume of paid App Units?
- What happened to organic (Browse and Search) impressions and App Units as a result of less efficient paid campaigns?
As for this last point – note that the volume of organic search and browse impressions and App Units is usually correlated to the overall level of paid UA spend in certain channels. This is because users, instead of responding directly to the ad, either search for it in the App Store or find it on the top charts as the volume of paid App Units drove it to a higher ranking.
A 2-fold proposed methodology
In order to combat this expected effect, and to account for the loss of ad creative optimization, we propose implementing the following methodology:
Measure the strength of ad creatives in a replicated environment
Instead of relying on ad networks to tell you which creatives perform better and allocate spend efficiently, you’ll need to supply them with already “qualified” ad creatives that you know work well for your target audience.
As a/b testing ad creatives in Facebook and other networks will be disabled (because they don’t get any ad-level install data back from SKAdNetwork), we propose testing these ad creatives in a sandbox environment such as ASOWorld.
The gist of it is: set up a replicated App Store page using ASOWorld and run a test ad campaign on the channel you want to assess the ad strength for.
You can test multiple ads simultaneously, driving traffic from all ads to your test App Store page that’s displaying your current App Store creatives (the control variation).
As users won’t be sent to the actual App Store, but to the ASOWorld replicated App Store environment, you’ll be able to gauge:
- Top funnel ad performance metrics such as CTR and CPM.
- App Store conversion rate per ad
- Cost per install
When setting up the test campaign to drive traffic from your target audience, you’ll be able to gain extremely valuable insights that will allow you to “feed” Facebook and other networks with ad creatives you know are going to perform well.
Then you just rinse -> wash -> repeat, always iterating your creative ads and improving performance further and further by understanding how your target audiences respond to different messages and creatives within your ads before introducing them to Facebook.
If you find that a certain ad is performing poorly, you save yourself from spending a significant budget on it (as Facebook/ ad networks won’t know it’s not performing well).
Improve your app store page creatives to fit the best ad creatives
Once you find ad creatives that seem to be working very well, you should switch to testing your app store page creatives to find the right messaging for this audience to maximize conversion rates even further, bringing your CPIs down.
Remember, what worked before might not work in the future, your app store page has a lot more convincing to do now that the average intent of your paid app store page visitors will be lower.
Let’s bring this all to a close by thinking about the entire industry trying to increase paid UA budget efficiency as much as possible in order to combat the negative impact of iOS 14.5.
It’s clear that some UA and mobile marketing teams will be able to feed the networks with better ad creatives and have better app store pages that support them and convert more of these ad impressions. They’ll take advantage of the data they have from their new creative optimization methodology.
Comparing their performance against a competitor that doesn’t employ such methodology, these folks will most probably experience very significant “leaks” where a lot of their UA spend will flow to lower-performing creatives, and they’ll have a much harder time improving performance without accurate data on which ads and which app store pages lead to better metrics (such as App Store Conversion Rates and Cost per Install).
It’s as if this entire era is taking us back to a time when the most creative marketer understood their users the best; not marketers who won by tapping into a black box that drove them high-quality users AND did all the creative optimization work for them.