A blueprint for iOS campaign success

It’s no secret that your mobile revenue is heavily dependent on iOS users. That’s why measuring your iOS campaign performance is critical to maintaining a high return on ad spend (ROAS).

But, in the absence of user-level data, how do you actually do that?

In this playbook, AppsFlyer and Google team up to share winning strategies and tools to effectively measure and optimize your iOS campaigns with SKAdNetwork—an essential component for growth. Follow this blueprint, which includes invaluable tips and recommendations to capitalize on the opportunities in the iOS world.

What’s inside?

• SKAdNetwork basics
• Innovative solutions to overcome SKAdNetwork limitations 
• Industry-specific SKAdNetwork use cases
• Best practices for setting up your Google iOS App campaigns for success

The post Boost your app growth with SKAdNetwork appeared first on AppsFlyer.

The exciting and much-anticipated release of SKAN 4.0 not only brings a lot of new features with it, it also brings a lot of options for both ad networks and app developers. With each version update, SKAN brings more and more value, but also becomes increasingly complex; user acquisition and SKAN strategy are the name of the game, and will set the savvy iOS developers apart from the rest.

First thing’s first

Before we start talking about strategy, let’s first review what the SKAN 4.0 changes are all about. There are 4 categories for which SKAN 4.0 changes fall into:

1. LTV Measurement

In SKAN 4.0, Apple introduces up to 3 (!) postbacks per user that work in the following order:

1st postback

When looking at the first postback with regards to LTV measurement, there are two main features:

1. You can measure up to 48 hours of user activity, if you want to (previously, assuming a user was not active for 24 hours, you would not be able to measure anything beyond that time frame)
   
2. The introduction of coarse-grain conversion value. coarse-grain is a new type of conversion value that has only 3 possible values: low, medium, or high. Similarly to how you currently map events or revenue to a conversion value 0-63, you can now map an event or revenue to low, medium, or high. The returned postback will contain either a fine grain conversion value, coarse-grain conversion value or “null” (depending on the privacy thresholds defined by Apple). Why is this important? We expect the null rates to drop dramatically thanks to the addition of the coarse-grain conversion value. If previously, Apple needed to mask the conversion value when privacy thresholds weren’t met, there’s now a “middle ground” with the coarse-grain conversion value. 

2nd postback

The introduction of a second postback is great news: You can now measure user activity for day 3 to day 7. The three disclaimers to the second postbacks are:

1.  Only the coarse-grain or nulled conversion value is returned.
   
2. The delay from the time the postback was ready to be sent and until the postback is sent varies from 24 hours to 144 hours.
   
3. The 2nd postback will only be sent if the user was active in this time frame.

3rd postback

The third postback is very much like the second postback, but measures user activity from day 8 to day 35. Similarly to the second postback, the third postback has three disclaimers

1. It can only return a coarse-grain or nulled conversion value. 
   
2. The delay from the time the postback was ready to be sent and until the postback is sent varies from 24 hours to 144 hours.
   
3. The 3rd postback will only be sent if the user was active in this time frame.

lockWindow

Apple is also introducing a new concept – ‘lockWindow’. This gives the app developer the option to stop measuring user activity in order to get the postback as soon as possible. Locking the measurement window can be done for each postback within its activity window. For example, you can lock the window of the second postback anywhere between day 3 and day 7 of the user’s activity. 

2. Ad network optimization / reporting drilldown

One of the critical updates for SKAN 4.0 will be very useful for ad networks: better granularity in campaign dimension reporting. In previous SKAN versions, campaign IDs were limited to 2 digits, meaning ad networks could only assign an ad a number between 0-99. In SKAN 4.0, the campaign ID is reborn as “source identifier”, and can contain up to four digits (up to 10,000 values). 

It’s extremely important to note that the ad network will only receive all four digits in a postback if the privacy thresholds are met. When privacy thresholds are not met, the ad network may receive three digits or two digits (two digits will be the minimum). 

Over the past few months, we’ve held multiple discussions with ad networks on how they plan to utilize this new feature. As it stands, ad networks have two main approaches:

1. The extra two digits will be used for internal optimization
   
2. The extra two digits will be used for additional reporting breakdown, based on the advertiser’s needs (e.g. country dimension or creative ID)

3. Additional attribution channels

One of the main pain points previous SKAN versions presented was the lack of support for web-to-app attribution. As a result, some of marketers’ main acquisition channels (e.g. Google Ads) were not covered by SKAN. 

Apple is now adding support for web-to-app attribution for Safari.

While this new solution supports Safari only, our data shows that Safari is the mobile browser of choice for 93.65% of iPhone users.

4. Privacy thresholds

In previous versions of SKAdNetwork, if a cohort of users did not pass the privacy threshold, Apple would drop the conversion value and/or source app ID from the postback and both would be nulled. 

In SKAN 4.0 this concept is broadened; there are four tiers of crowd anonymity: 0, 1, 2 and 3. Apple determines which crowd anonymity tier each install belongs to, and may drop one or more of the following fields based on the tier: fine grain conversion value, coarse-grain conversion value, source ID, or source app ID. 

Building your strategy

Now that we understand SKAN 4.0, let’s build a plan on how to tackle it. 

While it will likely take a while for the industry to align with SKAN 4.0, you can already start thinking about how you would like to set up your LTV measurement to make the most of the additional capabilities. 

What should you measure and when? Here are the factors to take into account:

1. Data freshness – how soon do you need your data for optimization purposes?
   
2. Measurement optimization – how can each postback (first vs second vs third) measurement be optimized? 
   
3. Reporting – what could your reporting look like?

Data Freshness

A very important aspect of the ability to run campaigns effectively is taking decisions as fast as possible. With SKAN 4.0, unless the lockWindow function is used, the first postback will arrive 3-4 days after the install occurs.

When considering what you would like to measure, you will need to take into account that there’s a tradeoff between capturing user activity and data freshness. 

Measurement optimization

A common misconception we’ve encountered is that the coarse-grain conversion value is a derivative of the fine-grain conversion value. This is not true at all. The coarse-grain conversion values are an entirely separate method of mapping events or revenue. 

In AppsFlyer’s Conversion Studio, advertisers will be able to map a specific in-app event or revenue range as ‘low’, ‘medium’ or ‘high’ for each postback. For example, a developer could potentially map a purchase event as ‘medium’ on postback #2, and map a level_complete event as ‘medium’ on postback #3. 

Depending on your app category, the coarse-grain conversion value can carry significant value and insight; a subscription-based app could map the subscription event to ‘high’ on the 2nd postback, for example. You can check out some category-specific examples for SKAN 4.0 measurement in the interactive SKAdventure simulator.

Reporting

Reporting with SKAN 4.0 could get a bit tricky. Since the source ID can be either 2,3 or 4 digits, different dimensions may be returned on every postback. 

Let’s look at an example: If an ad network is dedicating the 3rd digit to indicate the country, some of the postbacks received will include the country dimension, while others  may not due to privacy thresholds. Moreover, the 2nd and 3rd postbacks only ever contain 2-digit source IDs by default, meaning the country data will only ever be available up to day 2 post-install. 

App developers will need to prepare accordingly, understanding what the new capabilities—combined with the new restrictions—mean for their LTV data reports.

Get the latest marketing news and expert insights delivered to your inbox

Looking ahead

SKAN 4.0 brings a LOT of great new capabilities and many options for app developers to define how and what they want to measure. While this news is still fresh, we highly recommend getting yourself familiarized with it and getting a head start on planning. 

It may take some time for the industry to fully adopt SKAN 4.0, as adoption depends on every stakeholder in this ecosystem: app developers, MMPs, ad networks, publishers and end users. 

The post SKAdNetwork 4.0 is out, let’s build your strategy appeared first on AppsFlyer.

SKAN makes the world go round, at least for app marketers!

The significant impact of Apple’s SKAdNetwork (SKAN) cannot be overstated. Despite its attempts to help advertisers measure and optimize ad campaigns while maintaining user privacy, the ability to do so remains a significant challenge for many marketers. 

Although the upcoming SKAN 4.0, which is expected to go live “later in 2022” according to Apple, offers significant improvements to measurement and optimization, it is becoming increasingly more complex. There are still uncertainties about the solution, while ad networks will choose their own way to adopt the new features, creating a lack of industry ‘standard’. So you can expect an adjustment period for marketers, media sources, and MMPs alike.

Effectively implementing SKAN in gaming is a unique challenge of its own. In this article, we’re going to take a deep dive into user acquisition channels for gaming and how you can get the most out of them in a SKAN environment.

There are lots of options for marketers in the mobile ad ecosystem, each with their own strengths and weaknesses. In this post, we’ll focus on the following media sources for SKAN: Meta (Facebook), TikTok For Business, Snap, and Google. 

AppsFlyer’s SKAN Index, which ranks the best media sources in mobile advertising, shows that Meta Ads captured the #1 ranking for in-app purchases in gaming apps, followed by TikTok. In general, both are showing better results than Google and Snap in the SKAN environment.

Apple’s privacy threshold: what do we do?

There’s one big challenge with aggregated reporting like SKAN. The fewer the postbacks, the easier it is to identify a specific user. To combat this, Apple introduced the privacy threshold.

Apple states that postbacks “may include a conversion value and the source app’s ID if Apple determines that providing values meets Apple’s privacy threshold.” In other words, Apple’s privacy threshold is designed to prevent unique users from being identified. As a result, if publishers drive installs that don’t meet the number that Apple deems to be the threshold, they will receive postbacks with a null conversion value.

So what does this mean for advertisers?

Advertisers need to find a way to structure their campaigns to high volume installs across all their channels in order to receive the most data possible. This may mean shifting from hyper-segmented campaigns to aggregation and volume-focused campaigns. 

SKAN user acquisition strategies for gaming apps

Now that you’re well versed with the challenges of UA in the world of ATT, let’s dive into the nitty gritty tactics.

TikTok 

TikTok limits the number of UA campaigns you can run at once. An effective strategy is to group countries and consolidate lookalikes and interest group audiences with similar bids or similar ARPU (average revenue per user).  

For example, if you have two separate campaigns targeting English speaking audiences in the UK and US with similar bid and optimization goals, you can consolidate those to a single campaign. 

According to Matej Lancaric, who consults many gaming companies, “I usually group all Tier 1 countries together not looking at the language, but only ARPU or LTV (lifetime value), then create Tier 2 segments, and a third, ‘Rest of the World’ segment.”

Here are the geo tiers according to ATM:

• Tier 1:  geo set that every CPA marketer desires to work with. The wealthiest countries and the most competitive GEOs.
• Tier 2: Less competitive locations and lower average income per person.
• Tier 3: Developing countries and consumers with low purchasing power.

In cases where you want greater flexibility in testing audiences or creatives, you can utilize the multi-ad group feature which allows 2 ad groups per dedicated campaign.

Leverage TikTok’s Automated Creative Optimization (ACO)

Lancaric recommends using Automated Creative Optimization (ACO) in order to manage your ads more efficiently and effectively, since ACO automatically identifies high-performing combinations of your creative assets. 

ACO enables you to generate 150 ads based on 30 videos, 5 ad texts, and dynamic CTAs. Quick napkin math: generate 30 x 5 x Dynamic CTA = 150 ads. TikTok makes it extremely easy, but don’t get too carried away with an unmanageable number of ads.

Here are a few additional tips with ACO you should keep in mind:

1. ​Always upload a variety of creative assets. For ACO, you’ll need at least 3 assets for it to be effective. 
2. Leverage ad-level SKAN reporting to understand creative-level performance, and optimize performance based on these results.
3. There are 3 SKAN metrics in Ads Reporting – Conversion (SKAN), CPA (SKAN) and CVR (SKAN) for your creative performance evaluation of iOS 14 dedicated campaigns.  
4. Don’t make campaign adjustments in the first 72 hours after launching the iOS 14 campaign. 
5. Wait until you see the initial data first. When evaluating overall campaign performance, wait an additional 24 hours (ideally 72 hours) so that the campaign can fully receive all  SKAN conversion post-backs. 
6. Super important: Expect a longer learning phase for the iOS 14 campaigns due to SKAN delays and IDFA loss. 

Meta ads (Facebook)

While only Apple truly knows the magic privacy threshold number, Meta data shows that you will need to drive at least 88 app installs per day. Again, if the threshold is not met, you’ll likely see null conversions in your reporting. 

Simplifying your campaign structure by combining ad sets and campaigns can help you meet that 88 app install threshold. Consider increasing budgets and moving up the funnel to a wider audience or capture more frequent events.

And with any platform, don’t forget to rigorously run and test creatives to drive down CPI. To save on costs, you can run creative tests on Android and push the winners to iOS campaigns.

Leverage Meta’s automated optimizations

If you’re short for time, consider switching campaigns to Meta Advantage+ app campaigns (formerly known as AAA or Automated App Ads) structure to help overcome the privacy threshold. Advantage+ helps test creatives, simplifies your audiences, and optimizes towards installs.

You can split your mobile app installs’ event optimization to optimize for purchases and installs: 50% of the time the campaign will optimize for installs and the other 50% will optimize for purchases. This may yield lower CPIs and target high quality players. 

Be patient and go broad

To meet threshold demands, go as broad as possible when you can. Start by consolidating lookalikes and audiences with similar interests and behaviors. Run User acquisition campaigns with considerably higher budgets. 

As a rule of thumb, look at your CPIs on Android, multiply them 5-8x to calculate your budget. In other words, if you run $150-200/day on Android, you will likely need to start at least $500-600/day on iOS.

Don’t forget to wait at least 72 hours to evaluate performance for iOS 14+ campaigns. This is again, very important. Don’t panic if you don’t see data in the first 24-48 hours. They will show up eventually. 

It is also recommended to consolidate countries and bucketing geo tiers based on your own proprietary data. 

Here’s an example of bucketing geos based on the LTV for a specific game. But don’t copy, do your own homework!

T1: US, UK, CA, DE, FR, AU, KR, JP, CH

T2: DK, FI, NO, SE, NL, ES, IT, HK, SG, NZ

T3: PL, CZ, SK, BE, AT, IE, TR, UAE, SA, TW, TH, BG

Google

Google’s peak for iOS campaigns was when they offered value-optimized campaigns via Target ROAS (tROAS). As soon as ATT hit, tROAS was deprecated and modeled conversions were introduced.

As stated previously, modeled conversions take up to 5 days to appear, which is expected behavior. It’s important to note that reporting for recent campaign performance will continue to be affected by modelling lags, and you should continue to be prepared for performance fluctuations. 

But in the modern world of user acquisition, 72 hours is already too long of a window, so five days is a huge challenge for marketers.

Get the latest marketing news and expert insights delivered to your inbox

tROAS to return?

On the bright side, Google has plans to bring back tROAS from the dead. With it being in the beta for so long, they’ve confirmed plans to finally make it available to more developers.

According to Lancaric, “From first hand experience, so far so good. Advertisers can expect user acquisition campaign performance to fluctuate a lot, but there’s definitely lots of potential.”

While you can’t apply the tried and true Google Ads best practices, you can rely on what you use on other channels: fewer campaigns combined with the geo strategies mentioned previously to drive at least 100 installs per day to meet the privacy threshold.

Snapchat

Lancaric noted that Snap’s threshold was at around 75 installs per campaign per day. You should expect delays in your reporting from 24-48 hours from the actual install or subsequent in-app conversion event as a result of Apple’s delay framework.

To meet this threshold, a similar approach to Google, Meta, and TikTok is recommended. Take a patient approach to testing and optimization decisions.

This means waiting 48+ hours before adjusting your bids and budgets, and starting with high bids and budgets to meet the threshold. Then you optimize bids downwards. Or else you run the risk of prematurely pausing ad creatives that may lead to strong UA performance. 

Key takeaways

1) Don’t panic. Take your time while evaluating iOS14 campaigns. Wait 72 hours until you see the first data set come in. You should not make any adjustments at this point! Allow the first cohorts to mature and then make decisions!

2) Diversify UA. While we talked about top UA channels on SKAN, you have other sources of high quality traffic that work well. Check out the recent Performance Index to figure out what to add into your UA portfolio.

3) Consolidate and simplify campaign structures. Run fewer campaigns, allocate higher budgets to meet the privacy threshold. 

4) Focus on country tiers. Do your homework, check the LTV on the country level and group them into tiers to help meet the privacy threshold. 

The post How to optimize SKAdNetwork campaigns on Meta, Google, TikTok, and Snap appeared first on AppsFlyer.

There’s no doubt that Apple’s iOS 14+ has introduced important measures to safeguard user privacy. But it has also created significant challenges for marketers who have had to adapt to a completely new way of measuring their campaigns. 

One of the most complicated areas involves SKAdNetwork’s conversion value and timer extension mechanisms. This is a problem, since post-install data is what matters most in the freemium-driven app ecosystem. 

Before iOS 14, marketers could measure in-depth and at length to determine user/campaign value. But in Apple’s SKAdNetwork, data is far less in-depth and at length…You get 64 combinations from 6 bits to map post-install activity and a few days at best of activity data. In other words, conversion values are SKAN’s alternative to [limited] LTV.

With limited options, it’s imperative that advertisers make the most of what is possible. To be clear, what is possible can be plenty if done right. But what does ‘right’ look like? How should marketers map their conversion value schemes?

To answer this question, we looked at data from our Conversion Studio and analyzed the configurations of 900+ apps to help advertisers learn the bits and bytes about the bits and bytes.  

The data shows that for many marketers, conversion values are tough to master. While some have learned to squeeze the data lemon (unsurprisingly, mainly from the gaming crowd), others are not even close to utilizing the full potential of this new mechanism.  

In this article, we’ll show you how different verticals are: 

1. Mapping conversion value schemas
2. Defining the activity window timer 
3. Using the 64 combination capacity, or not 

For a quick recap on conversion values, watch this video below:

For more SKAN Insights, visit the Inside SKAN series.

What are the types of conversion mappings?

Before we show you the data, here’s a high-level explanation of the different types of models we’ll present in the charts (for more detailed info click here). 

Revenue: Revenue generated by the user during the activity window before the one-time postback is sent. It can be recorded using a single event, or associated with revenue by stream (IAP, IAA, subscription revenue) using select in-app events. 

In-App Events: Post-install in-app behavior such as tutorial completions, logins, purchases, shares, and level completions. Marketers can measure the number of unique users performing an event, the number of times an event was performed, or both.

Funnel: Measures the occurrence of in-app events based on a list of sequential events in a funnel. A user performing any event in the funnel is regarded as having performed all the events preceding that event. This is the most efficient way to measure sequential events, using fewer values for the same amount of in-app events measured.

Revenue and in-app events are the most configured schemes

A sub-division looks like this:

Key findings

1. Gaming apps are driven by revenue compared to events-driven non-gaming apps.

The majority of gaming apps are driven by revenue, with 84% of gaming apps utilizing revenue configurations. In contrast, 85% of non-gaming apps measure in-app events. 

Why? Gaming apps have a lower time to first purchase – particularly among the highly popular casual games – compared to non-gaming apps. The latter rely on in-app events to be key indicators for purchases that occur beyond the activity window.

2. Almost 40% of Shopping apps use a combination of in-app events and revenue.

As mentioned above, in-app events like registrations and email signups can be key LTV indicators for shopping apps. 

No less than 77% of shopping app configurations involve counting events, while 39% use a combination of events and revenue. The primary goal for shopping apps is to drive the first purchase as quickly as possible while measuring key events in the process.

3. 92% of Health & Fitness apps measure in-app events.

Most Health & Fitness apps rely on subscription models, requiring apps to focus on building loyalty and provide consistent value to their users. 

With free trials also being common, users take much longer to make a purchase, so it’s imperative that Health & Fitness advertisers focus on measuring early signals of value through in-app events – such as registrations, workouts completed, and shares.

4. 23% of Social Casino apps configured funnel events and revenue.

As one of the highest-grossing app verticals with an excellent ability to identify early patterns of high-value behavioral signals, more than a fifth of social casino apps measure the combination of funnel events and revenue.

5. Hardcore games are laser-focused on revenue.

Hardcore games heavily rely on in-app purchases and not in-app advertising, which is why 93% of these games utilize revenue configurations.

Get the latest marketing news and expert insights delivered to your inbox

Activity window correlates with duration of the funnel

Key findings

1. As the default setting, 24-hour activity windows are the most common.

The majority of apps across all verticals are utilizing the default 24-hour activity windows to collect data, but we expect this to change as SKAdNetwork usage matures. 

Beyond the fact that 24h-hour window is the default, it’s important to remember that data can only be collected once, so there is an intentional tradeoff. 

Shorter activity windows enable faster data collection, but leave you with limited information. Alternatively, longer windows provide richer data, but access to it will be delayed.

Also, the timer cannot be extended for all users – so even if it’s extended, it will only apply to a portion of users. The only way to extend the timer is by updating the conversion value, which can only happen if the app is opened.

So, even if you set the timer to 72H, if the user didn’t open the app for 24H, the conversion value is locked. For example, if a user opens the app on day 0, day 1, and day 3, the activity of day 3 won’t be recorded – because the app wasn’t opened on day 2.

Lastly, don’t forget that some major ad networks still do not support optimization beyond 24 hours, affecting marketers’ timer decisions.

2. Non-gaming verticals with longer purchase funnels use the 72-hour window.

Finance apps typically have longer purchase cycles, which is why 40% of them are using the maximum 72-hour window. 

Similarly, apps in other non-gaming categories use much longer purchase cycles than Gaming apps whose purchase journeys are significantly shorter.

3. 86% of Social Casino games have 24-hour windows.

On the opposite side of Health & Fitness apps, we have Social Casino apps whose ultimate goal is to get players engaged, hooked, and spending as quickly as possible. 

Gaming apps utilize more of the conversion value capacity

Key findings

1. Unlike non-gaming apps, more Gaming apps utilize conversion value capacity.

Gaming app marketers are well known for their level of data-driven expertise and ability to rapidly react to change. 

The sheer competitiveness of this vertical dictates that, with 63% already utilizing practically all options, compared to only about 40% of non-gaming apps (Food & Drink apps are an exception with no less than 80% using all options).

2. Hardcore and Midcore games have significantly higher conversion value capacity.

74% of Hardcore and 76% of Midcore games utilize 60-64 conversion values due to their more complex funnels. 

Users have a lot more actions and experiences to take at their disposal. They also both utilize longer activity windows, which provide more time to record more events.

How to squeeze the conversion value lemon 

While it’s key to learn from industry best practices, your success with conversion goals depends on how you tailor your strategy to your unique app. 

Here are a few strategies to keep in mind:

Having 64 options is a limiting factor, but still offers plenty of value if bits are properly allocated and utilized. Make the most of ranges and combinations and focus on the post-install actions that matter most. Split, swap, and combine your conversion mapping configurations until you find the right mix. 

Industry best practices are a good place to start. If you don’t have a lot of data to work with, the statistics above are an excellent starting point.

There’s no “right” answer, only what’s best for you. How you implement, measure, and optimize your conversion values is an ongoing process. Test, test, and test again until you find the right mapping (having a UI certainly makes it easier).

Measure smarter with funnels. We’re still in the early days of SKAdNetwork. Therefore, most advertisers are under-utilizing funnel configurations, which can be a more efficient way to allocate your bits. 

Instead of dedicating 3 bits to measure three separate events, a funnel configuration can measure sequential behavior using only one bit.

Predictive modeling is not a luxury anymore, it’s a must. The postback timer limits the amount of data you can collect to predict LTV for your users. Leverage predictive modeling to identify the most effective metrics and behaviors that can best identify your most profitable users. 

You can employ a predictive benefit score that uses all 6 bits to communicate any score, a specific combination (which will very likely differ from one advertiser to another) will indicate a valuable user cluster. 

Remember that while the benefit score is associated with a specific user, once the information comes back from SKAN it’s presented in an aggregate way for a user cluster.

Key takeaways

• Conversion value mapping is one of the biggest challenges app marketers are facing post-iOS 14.5. The right schema can create a significant competitive advantage. 

• Despite limited options, conversion values can deliver sufficient insights for measurement if done right.

• There are major differences between Gaming and non-gaming apps due to marketers’ know-how, funnel complexity and length. 

• Apps from every vertical are underutilizing funnel usage as adoption of the SKAdNetwork is still in its early days. We expect this to change as marketers get access to more data and adoption matures.

• There’s no right answer on the best conversion models, only what is best for you and your business. However, modeling your conversion mapping after industry best practices is a great place to start.

For more SKAN Insights, visit the Inside SKAN series.

The post SKAdNetwork conversion value mapping data shows gaps between gaming and non-gaming apps appeared first on AppsFlyer.

Welcome to our new SKAdNetwork feature requests series. In this new series we aim to create a platform where marketers and industry leaders can share ideas and SKAdNetwork feature requests with Apple. 

I’ll start: I’d like to talk about organic (unattributed) data.

The thing with organic data is that in most cases organic data is not really organic. 

Every organic install can be attributed to some marketing effort (word-of-mouth can be considered a marketing effort as well). So when I say “organic”, I mostly mean unattributed install data, which is, to put it simply, everything that is not directly attributed. 

iOS 14 and the loss of organic data

Measuring campaign performance has always been a challenge for marketers. Even in a fantasy world, where you could have access to all the data on the planet, you would still need to figure out how to set your benchmarks and which metrics to focus on. What’s more, the metrics you measure will likely change over time according to the needs and requirements of your business.

Whether you’re measuring installs, in-app CVR, eCPI, ROAS, ROI day 7, or anything else, comparing organic (unattributed) data with non-organic data is a critical aspect of performance measurement.

Image: Organic data compared with non organic data

In the wake of iOS 14+, marketers are truly struggling with the loss of organic data.
I speak with customers and partners often, gathering feedback and conducting research. This notion resurfaces time and again, making it harder than ever for businesses to assess the performance of their campaigns. 

Without organic data, marketers can’t really measure the quality of their paid users. Making organic SKAdNetwork data available as part of the SKAN postbacks is undoubtedly the next reasonable step.

Suggested solution

As a Product Manager, I always try to pick the “low hanging fruit”, or in other words: deliver the products that require little effort to make a huge impact. Without getting into implementation details, providing unattributed install data seems like low hanging fruit for Apple (pun not intended). 

So how can this be done?

SKAdNetwork is a distributed attribution solution where the clicks and impressions are stored on the end user’s device. For every install, where the advertiser uses registerForAttribution() or updateConversionValue() to request attribution data, if:

• the click and impression database on the device is empty, or 
• The lookback window expired

Apple should send a postback of organic (unattributed) data. The postbacks won’t include any attribution data, but should include conversion values based on user behavior. 

Here’s how I imagine such a postback would look like:

Since advertisers are already set up to receive a copy of SKAN postbacks to their endpoint of choice, organic postbacks can be sent exactly the same way, to the same destination. This way, advertisers get all of their data to the same location, where it can be grouped, analyzed, and enriched.

Get the latest marketing news and expert insights delivered to your inbox

What’s next?

Since the reintroduction of SKAdNetwork last year, AppsFlyer has been in constant  contact with Apple to provide feedback on important capabilities that we believe are still missing. 

Among other things, we argued that attribution data belongs to the advertiser, and therefore the advertiser should receive a copy of the winning postbacks. Apple listened. Starting in iOS 15, app developers can receive a copy of winning postbacks directly to their endpoint of choice.

We hope that this new series will continue to promote SKAdNetwork innovation and help marketers get the most out of SKAdNetwork.

Have more ideas on the evolution of SKAdNetwork? Shoot us an email to skadnetwork@appsflyer.com and we’ll host your idea (with full credit) in one of our blog posts. 

Or, In the meantime, you can contribute by completing the survey:

The post SKAdNetwork feature requests: Organic (unattributed) data appeared first on AppsFlyer.

• Publisher and advertiser app connections revealed
• May 20th and the curious case of the conv value null’s share
• The story behind the low NOI figures (and some reassurances) 

Introduction

Apple’s privacy-centric aggregated attribution framework SKAdNetwork (SKAN) is the new sheriff in town. With most iOS users denying access to their user-level data in versions 14.5 and above, you rely on SKAN to measure the success of their campaigns.

SKAN has turned mobile app marketing in iOS on its head, introducing completely new mechanisms to balance between data privacy and marketing measurement. At the heart of this new world order are sophisticated timer mechanisms, elusive privacy thresholds, and a unique conversion value system to grade the success of a campaign.

Since the introduction of Apple’s AppTrackingTransparency (ATT) framework in iOS 14, privacy-driven policies have taken center stage to eliminate bad practices surrounding user data. This is, without a doubt, an important step forward. 

However, many challenges have also been introduced with the new frameworks, particularly the limited data provided by SKAN to optimize campaign performance.

AppsFlyer’s Inside SKAN guide uncovers insights and learnings from SKAdNetwork during this time of transition. We have a clear goal in mind — to help you navigate this largely uncharted territory so that you can drive sustainable growth and profitability for your business. 

Chapter 1

Publisher and advertiser app connections revealed

Which app categories advertise most and with what kind of publishers? Neither Mobile Measurement Partners (MMPs) nor ad networks have been able to independently answer this question. 

Until now. 

As part of SKAdNetwork —  Apple’s solution for aggregated attribution post iOS 14 — new data has become available that enables the connection of data between advertisers and publishers for the first time. 

Previously, the MMP and advertiser weren’t exposed to the Apple publisher app ID (this data isn’t available in Android either). But the SKAdnetwork postback includes the source-app-id. Now, with this new data, we can see how publishers connect to advertisers, and vice versa. 

This data is extremely valuable to app marketers to inform targeting (especially in the age of privacy when data is limited and contextual signals become increasingly important), drive ad revenue (allowing publishers to prioritize categories and genres with high affinity to their own), and inform cross-promotion optimization, to name just a few key benefits.

In this guide we’ll uncover a world of insights from SKAdNetwork to help advertisers navigate this largely uncharted territory.

Sankey diagrams explained

A Sankey diagram is a specific type of flow diagram visualization, where the width of the arrow is proportional to the flow or value. 

The easiest way to explain a Sankey diagram is to show you one, and to dig straight into the advertiser-publisher data from SKAdnetwork:

On the left side are the publisher apps, and on the right side are the advertiser apps, showing us which app categories connect, and to what extent. 

There are three types of insights in these Sankey diagrams:

1) From left to right, or publisher app category to advertiser app category, we can see that 51% of installs driven by ads in Social apps were of Gaming apps. 

2) From right to left, we can see that 35% of Gaming app downloads were the result of ads in Social apps.

3) Left or right side only shows distribution of each type; for example, 48% of paid app installs were driven by ads on Social apps (publisher side).

Key findings

The data is based on tens of millions of SKAdNetwork postbacks aggregated between May 10 and June 15, 2021.

1) Overall: Social and Gaming publishers reign supreme

On the publisher side:

• Ads on Social & Gaming apps drive 92% of paid installs: 48% from Social, 44% from Gaming.

This is no surprise considering a) the popularity of social network ad platforms among advertisers (Facebook, Snap, Twitter, TikTok etc. are all integrated with SKAdNetwork), and b) the sheer size of Gaming in the app space and their heavy use of advertising to drive both demand and revenue as publishers.

• More than half the installs driven by Social publishers are of Gaming apps. An additional 10% led to Shopping app installs, and 7% to Education as well as to Health & Fitness apps.
• A whopping 97% of installs driven by Gaming publishers are of Gaming apps. The appeal is clear, gamers play many games and are always on the lookout for the next game to play.

On the advertiser side:

• Gaming apps account for 70% of paid installs, with Shopping a distant second at 5%. 

Gaming app marketers rely heavily on paid acquisition to drive demand. Because their space is hyper competitive like no other industry, they must spend money to make money and stand out. 

As the savviest marketers, they also have confidence in their ability to leverage data to ensure their ad investments are profitable. 

Get the latest marketing news and expert insights delivered to your inbox

2) Non-gaming publishers: Social apps command 86% share

• When we exclude the massive source of Gaming apps as publishers, we can see Social take over non-gaming publishing with an 86% share. 
• Although Shopping is the second biggest vertical, it is not used at all as a publishing platform. Given that the sole reason users browse Shopping apps is to buy goods, which of course is aligned with any brand’s goal, there is no reason to disrupt the flow with ads.
• Entertainment apps are the second largest publisher with a combination of streaming apps that often use ads as a revenue stream to complement/supplement subscription revenue, and the resemblance to Gaming in quite a few apps categorized as Entertainment in the App Store. 

3) Gaming sub-category*: Revenue model determines publishing space distribution

• Ads in Hyper Casual (HC) apps drive 48% of paid Gaming app installs. Since these games are completely reliant on in-app advertising (IAA) as a source of revenue, and because of their popularity, they dominate the publishing field among Gaming apps. 
• The rest of the Gaming publishing space is also driven by the revenue stream: Casual games often have a mixed income from IAA and IAP (in-app purchase), while Hardcore and Social Casino are far more reliant on IAP revenue. It is also a question of size as these subcategories are far more niche and appeal to more advanced gamers. 

Genre groupings were comprised of the following app store categories:

• Hyper Casual: Apps with at least 90% of revenue coming from ads
• Casual: Casual, Puzzle, Card, Board, Word, Educational, Trivia, Family, Sports
• Midcore: Adventure, Simulation, Action, Arcade, Racing
• Hardcore: Strategy, Role Playing
• Social Casino: Casino (not real money)

4) Gaming genre: Casual and Puzzle games hold a 45% share of publishing space

• A category level breakdown shows Casual apps driving 30% of paid Gaming app installs, followed by Puzzle at 15%. In this App Store-based categorization, Hyper Casual apps are included in Casual. 

As explained above, these apps rely on IAA and therefore dominate the publishing space. Puzzle games are a form of Casual games as they are relatively simple to play and rely on a mix of IAA and IAP to drive revenue.

• Although Racing games are not major publishers, they do command 15% of the paid install space. 

To summarize

• Thanks to SKAdNetwork, new data connecting publisher and advertiser apps has become available for the first time. 
• Given the popularity and size of social network ad platforms and Gaming in the app space, ads on their apps drive the vast majority of paid installs.
• Because of their hyper competitive space, Gaming app marketers rely heavily on paid acquisition, accounting Gaming apps for the bulk of paid installs.
• Social takes over non-gaming publishing, with Entertainment grabbing second place. 
• Shopping is not used as a publishing platform at all, despite being the second largest vertical, because ads tend to interrupt the shopping flow. 
• Ads in HC apps dominate the publishing field among Gaming apps, primarily due to popularity and for being completely reliant on IAA-driven revenue.

Chapter 2

May 20th and the curious case of the conv value null’s share

On May 21st, the share of conversion values (CV) with null entries in Apple’s SKAdNetwork postbacks changed dramatically. 

While some media sources experienced a drastic drop in CV null values, dropping from a share of 80% to 20% over the course of one day, others demonstrated the opposite trend, displaying a spike from 0% to over 30% CV nulls:

The data, based on 47 million postbacks from 10 major media sources during the measured timeframe, suggests that Apple changed the logic that determines its privacy threshold on May 20th. As a result, an immediate impact on the CV null rates was seen the following day. 

It’s important to mention that since Apple has yet to communicate anything regarding this matter, we can only make a confident assumption that this is indeed the reason. 

In this chapter, we’ll provide our take on this dramatic shift, its potential cause, implications, and recommended action items for advertisers.

The likely culprit: Apple’s privacy threshold change

As mentioned above, we believe Apple updated its privacy threshold on May 20th.

But what is this elusive privacy threshold, you ask? Great question. In its documentation, Apple only states: “The postback may include a conversion value and the source app’s ID if Apple determines that providing the values meets Apple’s privacy threshold.” 

What does meeting this threshold actually mean? We can only assume by connecting a few dots at this point. More on that a little later on.

There’s a strong indication that Apple’s privacy threshold is essentially an attempt to prevent any way of identifying or even coming close to identifying a unique user. 

In an aggregated solution such as SKAdNetwork, the lower the number of postbacks, the higher the chances to identify a specific user. Therefore, when granularity is as low as publisher level on a per-day basis, it’s possible for only a small number of postbacks to be included. 

In such a case, when Apple believes there’s a risk of identifying a user, it will mark all daily postbacks from that publisher as null.

Needless to say, this has massive implications for advertisers. Because it’s the only piece of data that is available on post-install activity from SKAdNetwork, it informs predictions of future value, and therefore campaign optimization. Without CV, advertisers are left with install data alone, which carries very little weight or value without the needed post-install data. 

Why did some networks experience spikes while others sharp drops?

We believe this is because each network has its own internal logic when it comes to figuring out the privacy threshold. They offer their advertisers guidance accordingly; each with its own set of campaigns hierarchy, ad-sets, publishers etc., so any change, such as the May 20th event, can bring about significant differences between networks’ performance. 

For example, an ad network with hundreds or even thousands of publishers will look at the data on a publisher level, which can include very few postbacks. Other media sources like social networks, analyze their data at ad-set level because they have just one or a few publishers. 

Facebook, for example, has recently advised its advertisers that the privacy threshold, as it applies to Facebook, is 128 installs per day per campaign.  

Overall, and beyond the fact that Apple limits SKAdNetwork campaigns to 100, undoubtedly a change in its own right, the privacy threshold creates added complexity for networks. Supporting this, several ad networks we’ve approached on the matter acknowledged the problem, and confirmed they were working to reduce their share of null values.

Fresh insights on Apple’s privacy threshold

To help shed more light on the recent privacy threshold modification, we compared the share of CV nulls with the number of postbacks per campaign, and then again with the number of campaigns.

The data presented is based on the DAILY number of postbacks per app, campaign, and network. It does not, however, include the source app ID dimension, which we believe is included in Apple’s logic. Since we don’t know which other factors are involved, these numbers should serve to highlight general trends rather than specific benchmarks.

The data above shows a clear correlation between CV nulls rates to the number of daily postbacks per app, per campaign, where the percentage of nulls begins a dramatic drop from 80% to 0% after exceeding 10 postbacks.

On the flip side, when we look at the CV null ratio compared to the number of campaigns, we can see that the percentage of nulls rises from around 10% to more than 30% roughly after 57 campaigns. Although this is not a consistent trend, evident by the sudden dips around 61 and 71 campaigns, it is an upward shift nonetheless.

Recommended next steps

• We recommend that advertisers revamp their campaign structure to favor a high volume of installs per campaign across all networks they advertise on, and follow network guidance where applicable (e.g. Facebook).
• It might be beneficial for networks to seek clarity from Apple, so that they can update their campaign taxonomy recommendation accordingly, but also receive a proper heads-up when logic or values are updated in the future. 
• Advertisers should be agile in executing on recommendations such as campaign aggregation, moving away from hyper-segmented campaigns.

The final word

Privacy thresholds are an important component in today’s privacy-centric reality. Enhanced transparency about how, what, and when they are put in place can help marketers optimize their campaigns while persevering the new privacy standard. 

To ensure maximal relevance, we’ll be sure to update the post as more information becomes available, and continue to closely monitor, analyze, and shed more light on this and future trends.

Chapter 3

The story behind the low NOI figures (and some reassurances) 

Mobile app marketers have had to rethink how to measure, attribute, and optimize their campaigns in the age of privacy. They’re struggling with the reality of having to determine their LTV and ROAS based on limited post-install data provided by Apple’s SKAdNetwork (SKAN).

But iOS 14 is also proving to have a negative impact on another key part of performance — the volume of non-organic installs (NOI) — leaving many advertisers concerned. 

Indeed, according to our data:

The average app is experiencing a 37% drop in the number of NOIs attributed by SKAN compared to pre-iOS 14 attribution figures.

So if you’ve seen it in your app, know that you are not alone. 

This analysis was based on a traffic comparison between July and February 2021, sampling the same apps and the same networks in the same timeframe. In other words, the combination of app and network had to appear in both timeframes for a valid comparison to be made. 

6 reasons behind the NOI decline

1) Low SKAN version 2.2+ implementation rate leading to missed VTAs

Because view-through attribution (VTA) is enabled only on SKAN versions 2.2 and above, we’ve found that 22% of postbacks originating from these versions are view-throughs. 

On the flip side, no less than 76% of postbacks came from SKAN versions that were below 2.2. In theory, up to 22% of these 76% could be VTAs that have not been recorded because of networks and publishers that do not support version 2.2+, and because of end users who did not update the publisher app version.

According to our data, 85% of publishers and 73% of networks have implemented versions 2.2+. However, this list doesn’t include several major players that are responsible for a significant volume of traffic and have yet to implement version 2.2+. 

It’s also important to remember that it takes not two but three to tango. 

That means that for a VTA to be recorded three conditions have to be met: both the network and publisher need to have SKAN version 2.2+, and in addition the end user has to have the updated publisher app version. 

2) Some publisher apps have yet to implement SKAN

In order for a SKAN postback to be sent, three things need to happen:

1. The ad network needs to sign the ad
2. The advertiser app needs to update the conversion value
3. The publisher app needs to trigger the SKAN flow

As we analyze only advertiser apps and networks, we still need publisher apps to play their part. We know that some of them have yet to implement SKAN, which is another reason that contributes to the NOI drop.

3) No web attribution in SKAN

The web is becoming an increasingly important touchpoint for app marketers because it is an attractive source for both acquisition and engagement with existing customers. According to our data, about 10% of installs have a web touchpoint. 

However, SKAN does not support web attribution…

Furthermore, many journeys begin on Google Search which is also a web touchpoint. Although not the default search engine in iOS’ Safari browser, it still commands about 10% share of iOS installs from Google’s channels. 

4)  Self-attribution for non-consenting users no more 

Prior to iOS 14, Self-Reporting Networks (SRNs) used to claim credit for any engagement as long as it was captured within their lookback window, regardless of the stage of the credited touchpoint in the funnel. 

In the SKAN reality, however, SRNs get credited for last touch attribution just like any other media source. If they did serve the last touch, these networks would receive a postback indicating a win. If they didn’t, no indication of a preceding engagement is credited. 

Clearly, there’s a delta here that advertisers see when comparing their SRN dashboards with SKAN traffic vs. pre-SKAN days. 

5)  Decrease in iOS budgets

Our data shows a drop of 15%-20% in iOS budgets, which also has an impact on NOI in specific and attribution in general. Having said that, the share of iOS budgets within the overall budget pie has only decreased by 10%. 

This trend, however, is far from being consistent. While some companies have shifted their budget towards Android campaigns, others are continuing to maintain stable iOS-focused ad spend.

6) Industry wide performance and tech adjustments still WIP

The paradigm shift in the mobile app space cannot be overstated. 

It requires heavy lifting on the tech side to build new models and new ways to optimize. But bear in mind this takes time and we are only a couple of months in. 

In this transition phase, limited data means targeting and optimization is just not there yet, and as a result there’s a drop in traffic. 

What you can do now  

Despite the undeniable level of uncertainty, there’s still plenty that can be done during this transition phase to improve measurability and performance, whether driving traffic or securing high LTV and positive ROAS.

Here are a few next steps and recommendations you can put to use that can boost your confidence to invest in iOS campaigns: 

• Make sure your publishers and networks upgrade their SKAN version to 2.2+. And if they’re not at all integrated with SKAN, make sure they do so as well. You’ll gain far more VTAs if more publishers upgrade. 

• Know that 64 conversion value options are enough to measure ROAS and predict LTV. Although you do lose a degree of data richness, it doesn’t mean that you can’t build campaigns that target high quality users. There are ways to make these 64 options work for you, as Eran Heres, CTO at CrazyLabs said in our recent iOS 14 guide for Gaming app marketers:

“We’ve built our own model that is based on the prediction of a user’s in-app purchase revenue and ad revenue. By leveraging all 64 values we’re able to measure the ROAS and predict user LTV with good accuracy.”

In the context of sufficient data for predictions, a recent Deconstructor of Fun post written by Tiffany Keller, Director of Product at Sybo, commented that “saying VO bidding killed all the whales for Mid and Hardcore Gaming is like saying that within 24-72hr of installing, only knowing someone made a $15 IAP, but not that they also bought a $2 starter pack can’t give us a good prediction of the install’s general LTV bucket.” 

• Leverage higher than anticipated ATT prompt opt-in rates for benchmarking. The share of consenting users is significant and higher than expected, which is good news not only for this cohort but mainly for the purpose of applying learnings and benchmarking to your non-consenting cohort. 

• Bring users to your app from the web where the cost of media is cheaper and the transition from web to app can work like a charm, ensuring measurability is fully retained.

• Utilize accurate attribution while preserving user privacy. Use all available data points, but ensure compliance with the way the data is being shared with your partners.

• Find creative ways to increase your opt-in rates.
  • Timing – Try to pinpoint the precise moment in your users’ journey to prompt the ATT dialogue, while creating different paths for existing and new users. Also, know that timing the ATT prompt to appear before the attribution SDK is triggered can have a major impact (180% higher IDFA rate), and in most cases, delaying by only a few seconds is all it takes. 

• Customization – The unbolded text in the ATT prompt can be edited so you can better explain why you want to capture the user’s IDFA.

• Pre-ATT prompt – Displaying a pre-prompt means you’re gating Apple’s ATT dialogue behind your own native prompt, enabling you to customize the design, timing, and messaging to best fit your app.

• Use IDFV across your apps to create a 1st party data powerhouse, and use existing audience segmentation integrations to communicate via in-app messaging and push notifications (mainly relevant for Gaming).

• Longer term solutions: Put incrementality and predictive analytics to use. These fields are currently being heavily-invested in by the entire industry, including attribution companies, SRNs, and ad networks. 

Advertisers should follow suit and begin developing these capabilities by employing a data science team. 

Although this extra step is not likely to address immediate pains, it will enable you to get more applicable data for internal predictions – a must-have in the age of privacy.

Summary and a few words of encouragement 

1) The current direction of the privacy wave is pushing all industry players to rely more on innovation, data science, and integrations, all processes that take time to bear fruit. 

2) Considering all the recently applied privacy regulations (LAT, GDPR, CCPA, etc.) we need to remember that the share of users who opted-out of tracking was not small even prior to iOS 14, especially in iOS markets, where LAT users were about 30%.

3) Aggregate data enables you to remain compliant with privacy rules, while conversion value management tools and predictive analytics give you the flexibility to make better decisions based on limited data.
The bottom line: Retaining volume and performance in the age of privacy is more than possible. It is only a matter of time.

The post Inside SKAN: Data & learnings from Apple’s SKAdNetwork appeared first on AppsFlyer.

SKAdNetwork is a privacy-centric interface operated by Apple. It helps ad networks and advertisers measure their ad activity (such as impressions, clicks, and app installs) on an aggregated level, without accessing individual user data.

What is SKAdNetwork?

First introduced by Apple back in 2018, SKAdNetwork (often shortened to SKAN) provides attribution data to advertisers on iOS so they can measure and optimize campaigns. Because it does this on an aggregated level, without revealing any user-level or device-specific data, it offers an accurate yet privacy-preserving source of campaign insights.

How does it work?

Before we launch into the mechanics, let’s take a look at the four main players involved, each having specific roles and responsibilities: 

1. Publishing app – where the ad is displayed.
2. Ad network – connecting app advertisers to publishers.
3. Target app – the app that is being advertised.
4. Mobile measurement partner (MMP) – responsible for connecting all the dots by attributing and optimizing app data and campaign performance metrics. Holistic SKAdNetwork management allows marketers to measure, visualize, and optimize with ease.

There are two types of engagements that SKAdNetwork registers:

1. Views – whether an ad was viewed
2. StoreKit renders – whether or not a StoreKit rendering was generated.

The SKAdNetwork flow

1. An ad is displayed in the publishing app. As soon as the ad is displayed, the publishing app starts the three-second timer and notifies SKAdNetwork that it has started.
2. If the ad is displayed for at least three seconds, the publishing app will then notify SKAdNetwork that the three-second timer is up and this activity will be recorded as a successful view. If the user engages with the ad, the publisher renders the advertised app StoreKit. 
3. Once it’s displayed, SKAdNetwork registers that the StoreKit was rendered successfully. The next step would hopefully be for the user to download the advertised app. 
4. If the StoreKit was rendered, the user can download the app right then and there. If the user installs the app and launches it within the SKAdNetwork attribution window, the install is attributed to the ad network and the device sends the install postback to the ad network and a copy to the advertiser.

A note on postbacks

A postback is where an attribution provider (in this case, SKAd Network) sends data back to the ad network and advertiser on user behavior — for example, whether they clicked on an ad or downloaded an app. This provides vital feedback that enables advertisers to optimize their campaigns. 

Unlike standard postbacks, SKAdNetwork postbacks are not immediately sent to the ad network and advertiser when the app is first launched. In fact, the attribution window (the period in which an ad impression or engagement is considered responsible for a conversion) can be up to 30 days between click and install, depending on the ad type. 

SKAdNetwork postbacks are built on a timer mechanism that only sends the postback when the timer runs out. This timer delays the postback by a minimum of 24 hours. Once the timer runs out, the ad network and advertiser receive the postback. 

It’s important to note that postbacks don’t contain any device or user data. Together with the timer delay, this ensures that user data remains private by removing the ability to single out a specific user.

If the advertiser is working with an MMP, like AppsFlyer, postbacks will be reported to the MMP via dedicated dashboards and APIs, or as configured by the advertiser.

The challenges – and how to address them

Compared with previous methods of attribution, SKAdNetwork does present a number of complexities and challenges for advertisers. To name a few:

• No real ROI/LTV – SKAdNetwork measures mostly installs, conversion values, and post-install data, but in a very limited, time-restrictive way. Prior to SKAN 4.0, the lifetime value (LTV) was constrained within 24 hours or the advertiser-defined activity window. However, with the introduction of SKAN 4.0, advertisers can now receive up to three postbacks, each corresponding to distinct timeframes (0-2 days, 3-7 days, and 8-35 days).
• Granularity – In SKAN versions 2 and 3, data granularity was restricted in terms of dimensions, and limited to only 100 SKAN campaigns. However, SKAN 4.0 provides increased granularity, allowing for up to 10,000 SKAN campaigns.
• Postback delay – Postbacks are delayed by at least 24 hours, limiting how fast you can optimize campaigns.
• Ad fraud risk – Data can easily be manipulated in transit, resulting in wasted budget.
• No attribution support for re-engagement activity, which is vital for reducing churn and building user loyalty. . 

Fortunately, there are ways to navigate these challenges, so you can continue to measure and predict the value of your acquired users. 

The key to unlocking value from SKAdNetwork is to understand Apple’s unique conversion value mechanism.

Conversion values – and how to make them work for you

Conversion values are configured by app developers to measure post-install activity, and tie it back to the install. A single conversion value is included in the one-time postback that Apple sends to the ad network and advertiser. 

As a result, the information in this conversion value is all you get on a user’s post-install activity (assuming they didn’t consent to “tracking”), making it extremely important. After all, in a freemium world, optimization is driven by post-install data.

Starting with SKAN 4.0, there are two types of conversion values: 

• Fine-grained conversion values are defined by six bits, which are binary measures, meaning they can be turned on or off (0 or 1). This opens up the potential for 64 measurement combinations within those six bits – from 0 to 63. Although this might sound limited, there are still plenty of options to work with to measure revenue, engagement, funnel progress, and more. 
• Coarse-grained conversion values are divided into three types: low, medium, and high. These values are assigned by the advertiser to indicate varying levels of user engagement, allowing advertisers to receive at least some attribution data in cases where the privacy threshold is not met. 

As long as you properly map your conversion values based on your internal logic, these values can be used in any way you want. They’re yours to control and assign to the KPIs that are most valuable to you. 

The 64 fine-grained values and 3 coarse-grained values, each with their own unique decoding configured by the app developer or advertiser, are then attributed to the source of the install, enabling campaign measurement and optimization.

Different approaches work for different app types. For example, data from AppsFlyer’s Conversion Studio shows that gaming apps are laser-focused on revenue – as such, it’s a model that’s being involved in most conversion value schemes. On the non-gaming side, in-app activity is the most configured option.

Learn more here about how to make the most of conversion values, including benchmarks on activity window timer, and optimal usage of the 64-combination capacity. 

SKAdNetwork 4.0 (SKAN 4.0)

On October 24, 2022, Apple released the next version of SKAdNetwork (4.0). This introduced significant changes that allow advertisers and ad networks to measure more while maintaining user privacy.

Three postbacks instead of one

SKAdNetwork 4.0 enables advertisers to receive up to three postbacks, each based on a specific activity window (0-2 days, 3-7 days, and 8-35 days). This allows advertisers to understand how users engage with their app over time. 

The first postback is sent within 24-48 hours as before. However, the length of the timer increases for the second and third postbacks, which are sent after 24-144 hours.

Although it’s important to note that these three postbacks cannot be connected and tied to a specific user, advertisers are still able to count unique event occurrences.

LockWindow

Although each postback is based on a limited activity window, SKAN 4.0 also introduces a new ability called LockWindow, which allows app developers to finalize the conversion value and lock the measurement window in order to receive postbacks sooner. For example, app developers can lock the window of the second postback anywhere between day 3 and day 7.

Crowd anonymity

Crowd anonymity is a new term that Apple uses to describe the privacy-preserving way in which SKAN delivers attribution data. In short: the more installs you get, the more data you get.

To maintain user privacy, Apple limits the data that SKAN shares in postbacks. Postback data tiers are based on the conversion-volume of the campaign, like this:

Apple decides which crowd anonymity tier each install belongs to, and shares data accordingly. 

For example, for ads in tier 0, the postback will not include a conversion value (null). Whereas for ads in tier 1, the first postback will include only a coarse conversion value and a two-digit source identifier. However, for ads in tiers 2 or 3, the first postback will include a fine-grained conversion value and a two to four-digit source identifier. The second and third postbacks will only be shared if the crowd anonymity tier is above 0, and will include only a coarse conversion value and a two-digit source identifier.

Here’s a summary:

Hierarchical conversion values and source identifiers

In earlier versions of SKAdNetwork, postbacks included a conversion value only in cases where Apple’s privacy threshold was met. When crowd anonymity is low, Apple takes extra precautions to protect user privacy by masking the conversion value and source app ID. 

SKAdNetwork 4.0 introduces a new set of “coarse-grained” conversion values (in addition to the 64 “fine-grained” values that exist today).  

As previously mentioned, coarse-grained conversion values are categorized into three types: low, medium, or high. Advertisers assign these values to signify varying degrees of user engagement, enabling them to obtain partial attribution data when the privacy threshold isn’t reached, particularly at lower levels of crowd anonymity.

Another change introduced in SKAN 4.0, which significantly impacts campaign granularity, involves the source identifier. Apple has taken the step of renaming its campaign identifier field as the source identifier, and along with this change, has expanded its range from two digits (equivalent to 100 options) to four digits (equivalent to 10,000 options).

Although the source identifier is presented as a single number, Apple urges advertisers to interpret it as a sequence of three hierarchical numbers. By adopting this approach, advertisers gain the ability to comprehensively measure various parameters, including factors like ad placement, geographic targeting, creative components, and more.

Coarse-grained values are provided in the 2nd and 3rd postbacks, or in the 1st postback if the privacy threshold is not met; whereas the fine conversion value is included in the 1st postback only.

Hierarchical source identifier

From SKAN 4.0 onward, Apple is renaming its campaign identifier field to source identifier and increasing its range from 2 digits (representative of 100 options) to 4 digits (representative of 10,000 options). 

Although the source identifier is a single number, Apple encourages advertisers to use it as 3 hierarchical numbers — allowing them to measure more parameters, such as ad placement, GEO, creative, and more.

As with hierarchical conversion values, hierarchical source identifiers also adhere to Apple’s privacy threshold — meaning the higher the level of crowd anonymity, the higher the level of provided granularity.

Like hierarchical conversion values, these hierarchical source identifiers also abide by Apple’s privacy threshold. This means that the level of detail provided is directly tied to the level of crowd anonymity, making the system adaptable to diverse privacy requirements.

Web-to-app support

Previously, advertisers could only measure app-to-app flows, but web-to-app was not supported. In SKAN 4.0, Apple supports web-to-app attribution for Safari, enabling advertisers to measure campaign performance across multiple channels.

Find out more about SKAN 4.0 changes and building your strategy

How can advertisers make the most of SKAdNetwork?

Here are a few steps to ensure your business is SKAdNetwork-ready:

• Data aggregation – Be sure to collect all SKAdNetwork information from each ad network.
• Data validation – Ensure all postbacks are signed by Apple and aren’t manipulated in transit. Working with a trusted MMP can help you address this with ease.
• Data enrichment – Match SKAdNetwork information with other data points, such as impressions, clicks, cost, organic traffic, and more, for complete ROI analysis.
• Data enablement – Use dedicated dashboards and interfaces to view your data in a clear, convenient way. An MMP can help with this. 
• Seamless integration – Make sure your mobile attribution solution offers full encapsulation. This will save you time and effort, especially when it comes to future changes in the SKAdNetwork protocol.
• Conversion events – Be sure to measure server-side, dynamic, and flexible in-app events.

AdAttributionKit

At WWDC 2024, Apple introduced AdAttributionKit (AAK), a new framework that enables wider ad attribution capabilities while seeking to meet regulatory demands. According to Apple, the AAK framework builds upon the core principles of the SKAN methodology.

Although similar, AdAttributionKit differs from SKAN in three key ways: 

1. Alternative app store support: AdAttributionKit is built on SKAN’s foundation to offer better integration and expanded capabilities across more marketplaces, making it easier for marketers familiar with Apple’s tools to adapt. However, since no marketplaces are live at the moment, this feature’s immediate impact is minimal.
2. Re-engagement capabilities: Re-engagement capabilities previewed during WWDC23 have been incorporated into AdAttributionKit to enable measuring conversions from ads clicked by users who have already installed the app. Although re-engagement isn’t supported by SKAN, marketers are successfully using deep linking and other methods to measure user interactions. 
3. New developer mode: AdAttributionKit introduces a new developer mode that facilitates app development and testing by simplifying the measurement process. This makes it easier to test attribution setups without complex configurations or waiting for live data.

While AdAttributionKit offers promising new features, advertisers should continue leveraging SKAN measurement capabilities while staying informed on AdAttributionKit’s future developments. 

Frequently asked questions

Key takeaways

As the mobile advertising industry continues to adapt to a privacy-centric reality, the ability to measure, attribute, and optimize will continue to improve. Better models, increased usage of predictive analytics, acquired expertise in SKAN, and innovation across the ecosystem have all driven improvements in measurability. 

As an advertiser, you should:

• Leverage easy-to-use tools to map and constantly update your conversion value schema to make the most of the 64 options, instead of wasting time on development.
• Use predictive analytics to overcome time limitations and leverage early signs of engagement to predict long-term campaign performance. Put mobile attribution in SKAdNetwork on “autopilot”, removing measurement and timing barriers, so you can maintain and strengthen your competitive edge in this new reality.
• Keep your data safe from all types of fraud in the new iOS 14 ecosystem. An MMP can offer protection from SKAdNetwork fraud, by ensuring that you’re getting accurate data on your campaign performance. This will protect your ad spend before, during, and after installs, with end-to-end coverage against infrastructure weaknesses, data limitations, and reporting loopholes.
• Connect with an ecosystem – partner with an MMP that has joined forces with ad networks such as Facebook, Twitter, Snap, or ironSource. Solid cooperation ensures that postbacks, conversion value schemas, and data are delivered to your MMP and chosen partners smoothly and easily.

Mobile marketing dictionary from A-Z

Thanks for your download!

The post SKAdNetwork (SKAN) appeared first on AppsFlyer.

Background

Founded in 2014, Clover drives the essential connections of the future by helping singles looking for love and friendship to find their match.

Although Gen Z and millenials are searching for the right match on their own timeline and waiting longer to get married than previous generations, they still enjoy the fast pace of online dating. With over 9 million users worldwide Clover fills this market need. As a subscription-based service, having users convert into paying customers and measuring ROI are vital to Clover’s success.

To zero-in on high quality users in their core demographic, the team runs campaigns on major social networks, including Snapchat, Facebook, and TikTok.

They analyze their campaign performance across these key channels on a weekly and sometimes daily basis to make agile UA optimizations, which is a big reason they are able to keep all user acquisition in house.

Challenge

Like so many other advertisers, Clover knew that the release of Apple’s iOS 14 and the introduction of the App Tracking Transparency (ATT) framework – a prompt asking users for permission to “track” their device data – would pose new challenges when acquiring users on Apple devices.

Being unable to leverage the Identifier for Advertisers (IDFA) – the identifier used to measure user engagement and remarketing – meant the team needed to understand how their current acquisition strategy that relied on IDFA needed to evolve in order to maintain their high rate of growth in a quickly evolving mobile marketing ecosystem.

Clover’s overall strategic focus was converting free users into paid subscribers as quickly as possible. But it could take weeks or months before a free user became a subscriber.

Although Clover optimized for installs and not post-install events, understanding how these downstream events indicated future subscription revenue was key.

If a user downloaded the app and then registered or signed up for a free trial within a certain time frame, Clover could be relatively confident that the user would go on to become a subscriber. Being able to quickly identify trends in these post-install events made them a valuable barometer for subscription performance, and indicated install quality that informed strategic budget allocation.

The introduction of the new iOS 14 changes also brought SKAdNetwork into the story.

The new limitations around IDFA forced Clover to shift some of their iOS ads to SKAdNetwork, Apple’s proprietary campaign performance API. SKAdNetwork is designed to provide attribution data while maximizing user-privacy.

As a result, one of SKAdNetwork’s features is delayed postbacks: Postbacks are delayed at least 24 hours after install to ensure that the user/ device cannot be identified. Once a user installs the app, SKAdNetwork’s 24-hr timer starts running.

The advertiser can set a number of events that effectively extend this timer (such as resetting each time the app is opened) but if these events don’t occur within the 24-hr window, the conversion value is locked in and the postback is sent. This means if a user registered for a free trial in this time frame, but at some point much later went on to subscribe, the user would forever be recorded as only registering for a free trial.

This timing change introduced by SKAdNetwork made measuring events over 24 hours post-install difficult, and severely limited Clover’s ability to engage in meaningful analysis of the quality of their installs.

Solution

Clover turned to AppsFlyer’s solution for iOS, which brings the AppsFlyer SDK, infrastructure, and platform together to preserve as much data insight as possible while remaining compliant with Apple’s new privacy policies.

The SKAdNetwork infrastructure operates without IDFA, so Clover could ensure measurability of their campaigns independent of the ATT prompt response. In cases where permission to use the IDFA was given, Clover could additionally rely on AppsFlyer’s deterministic attribution for their mobile measurement. Combined with the SKAdNetwork dashboard that provided a comprehensive visual drill down of critical performance KPls, such as ROI, CPI, and ROAS, Clover could make data driven decisions with confidence.

Clover first integrated AppsFlyer’s plug-and-play SKAdNetwork solution with Snapchat in a simple set up that required minimal time and resources. This gave them the ability to test and optimize on Snapchat before integrating with Facebook and TikTok.

Results

Implementing SKAdNetwork through AppsFlyer gave Clover valuable insight into what privacy-centric measurement looked like in practice, getting them ahead of the competition once the release day came.

Adapting to the delayed postbacks and working with aggregated data led to key workflow and strategy changes. But Clover was able to make these adjustments with plenty of time to spare.

Ultimately, moving to aggregated user level data didn’t have as big of an impact on Clover’s ability to achieve their goals as they first thought it might.

The majority of their key goals could still be consistently reached using AppsFlyer’s suite of solutions. The team now has to exercise more patience, and allow a few days to pass to see what impact adjustments have on performance, but they also have the peace of mind knowing that they were fully prepared for success in a post-IDFA world when Apple flipped the switch to enforce mandatory ATT.

“Although losing user level data is one of the biggest changes mobile marketers have confronted, AppsFlyer’s SKAdNetwork solution made our transition into a privacy-centric reality painless.” – Natasha Upal, VP of Growth

The post Embracing SKAdNetwork to adapt to the privacy-centric future appeared first on AppsFlyer.

In yesterday’s WWDC 2021, Apple announced several more privacy updates for iOS, including a key update to SKAdNetwork that allows advertisers to receive a copy of postbacks for “winning” app installs.

Until now, postbacks were sent only to the ad networks, and then forwarded to the advertiser or MMP on the advertiser’s behalf.

While this update could have easily gone under the radar, it’s a critical step in the right direction for SKAdNetwork: it addresses and partially solves privacy as well as business issues in regards to data ownership around app installs and post install events. 

The ecosystem has been advocating for this change for some time, and it was one of our key concerns back when SKAdNetwork was reintroduced last year.

The advertisers want to own their data

Back in late December of last year, we ran a poll among our customers asking who they feel should receive the postback from SKAdNetwork. The results were conclusive – advertisers strongly felt they (or the MMP on their behalf) should be receiving SKAdNetwork postbacks.

We’ve spent the last year innovating on privacy to give our customers even more control over their data, how it is shared and with whom. This new announcement from Apple is a welcomed one indeed, and helps give advertisers the ownership they deserve of their critical data.

Apple has released a steady cadence of improvements since relaunching SKAdNetwork last June, and we look forward to the additional upgrades that are certain to come.

Prepare for iOS 15 with AppsFlyer

The postback copy will be supported in iOS 15, which is already in beta. We recommend advertisers get started with testing this capability, so that they’re prepared once iOS 15 is publicly available.

AppsFlyer’s SK360 solution gives advertisers the ability to manage their SKAdNetwork attribution alongside AppsFlyer’s traditional attribution.

Advertisers can now add the new endpoint https://www.appsflyer-skadnetwork.com to their app’s property list editor for AppsFlyer to manage SKAdNetwork. 

What AppsFlyer does for iOS advertisers:

1. Provides full flexibility for designing, testing, managing and optimizing their conversion value logic
2. Validates the postbacks against those reported by the ad network
3. Connects ad network campaign name, ad set name, and ad name to the SKAdNetwork-reported campaign
4. Connects pre-install campaign data such as cost, clicks and impressions
5. Provides the data in SKAdNetwork dashboard as well as SKAdNetwork APIs
6. Prevents and blocks SKAdNetwork fraud 

Get the latest marketing news and expert insights delivered to your inbox

Coming soon: SKAdNetwork postback real-time testing dashboard

AppsFlyer will soon be releasing a new real-time testing environment and dashboard to be used with a test profile (or even production data, if you so choose).

This environment will allow advertisers to see the SKAdNetwork postbacks that arrived from the end user’s device to the AppsFlyer endpoint (log-level data). 

Stay tuned!

The post SKAdNetwork postbacks: A welcome update for advertisers appeared first on AppsFlyer.

SKAdNetwork – Apple’s privacy-centric solution for deterministic attribution in iOS 14 — brings limitations, complexities, and restrictions compared to previous methods of attribution. 

Fortunately, there are ways to navigate these challenges to largely retain the ability to measure and predict the value of your acquired users.

The key to unlocking value from SKAdNetwork is to understand Apple’s unique conversion value mechanism. In this post, we’ll dive deep into conversion values – what they are, the bits that they’re composed of, and how to make them work for you.

What are bits and how do they work?

Conversion values are configured by the app developers to measure post-install activity and tie it back to the install.

*For DE subtitle scroll down

A single conversion value is included in the one-time postback that iOS sends to the ad network. As a result, the information in this conversion value is all the information you can get on a user’s post-install activity (only if they did not consent to ‘tracking’), making it extremely important. 

But what does that configuration actually look like, and how can it be customized to provide meaningful insights? 

This is where bits come in. A conversion value is defined by 6 bits, which are binary measures – meaning they can be turned on or off (0 or 1). Think of it as a light switch.

This opens up the potential for a variety of measurement combinations within those 6 bits – 64, in fact, from 0 to 63.

Although 64 options can be considered limited, there are still plenty of options to work with to measure revenue, engagement, funnel progress, gender, device, and more. 

These values can be used in any way you want – they’re yours to control and assign to the KPIs that are most valuable to you. As long as you properly mapped your conversion values based on your internal logic. 

The 6 values, each with its own unique decoding configured by the app developer/advertiser, are then attributed to the source of the install, enabling campaign measurement and optimization.

Optimizing bits to measure value

Mastering this method of measurement is the key to overcoming the limitations of SKAdNetwork and unlocking its potential. 

With your 6 bits, you still have the scope to measure early signals of engagement, retention and monetization, and construct a picture of a user’s predicted lifetime value (pLTV) from those signals. 

To do so, you must define your bit assignment strategy: what are the events that matter most to your own LTV calculations? From there, you can map those events onto the corresponding conversion values. 

You have plenty of flexibility as to how to make best use of your 6 bits to achieve this goal. We can broadly define these in three categories – flat, split, and combo-split.

1. Flat

With this method, all 6 bits are used to measure a single KPI. In the example shown below, the app developer has devoted all 6 bits towards measuring revenue. 

The bits here show a specific binary value of 110001, which then returns a corresponding conversion value that informs the app that the user has generated a specific amount of revenue – in this case $49 (see table above).

2. Split

Instead of devoting all 6 bits to a single KPI, split up the bits to measure multiple aspects of user behavior within the same conversion value. For instance, you could devote three bits to measuring revenue, and three to measuring game progress – enabling you to tie these measures together for individual users.

In this example of splitting, the app configured their conversion values based on the following logic:

The conversion value 46 means that the user has spent more than $20 while playing 25 levels.

3. Combo-split

The final category turns up the complexity another notch: combo-split. It takes the split method (in this example, revenue and game progress) and adds a final deterministic on/off signal via one of the remaining bits (e.g. whether the user was logged-in or not).

Using conversion values for predictive analytics

We’ve looked at some of the structural details around conversion values – now let’s talk about why this matters so much for predictive analytics. 

We mentioned above how conversion values are mostly based on signals early in the funnel of the user journey, as that is the nature of SKAdNetwork. On top of that, SKAdNetwork’s limitations (such as sending only 1 postback, and that without a timestamp) greatly impact your ability to cohort users and predict their value. 

Consequently, building effective predictive models to unlock the power of those early signals is more important than ever. 

Advanced marketers have had such models in place long before the arrival of SKAdNetwork, allowing them to predict user value early in the process and optimize quickly.

Now, with only the early signals provided by SKAdNetwork, predictive modeling is all but a must-have for marketers.

Predictions before and after SKAdNetwork

As each app has its own unique way of calculating user LTV – composed of its own range of in-app events, benchmarks and weighting – predictive modeling even before SKAdNetwork required machine training on historical data measured via the MMP’s SDK, followed by a scoring mechanism based on the completed in-app events. In the example below, the deeper the action in the funnel, the greater the weight given in the LTV prediction.

With SKAdNetwork, an additional step is needed.

Once the machine training period and event value configuration have taken place, you’ll need to determine your bit tactics to make sure they capture the event values and combinations that constitute a valuable user.

SKAdNetwork has made it a necessity for developers and advertisers to have a sophisticated prediction model built on these conversion values. Predictive analytics enables you to take user activity during their initial couple of days with the app, and correlate it with their long-term LTV. 

Of course, this has a huge range of benefits: you can cut your losses early in the event of high-risk activity, or alternatively identify potential success out of the gate and double down on it. It drastically reduces any wasteful learning periods and lets you take action.

The flexibility of conversion values – and the fact that you have control over them – give you solid tools for predicting future value and optimizing accordingly. 

Get the latest marketing news and expert insights delivered to your inbox

Taking control of conversion values

Understanding conversion values and making them work for you are crucial to unlocking the value of SKAdNetwork. That said, we won’t pretend this is an easy feat – especially if you want to deliver high-performing campaigns. 

As an advertiser or developer, control over conversion values is in your hands. However, there are a whole range of challenges to overcome – and that’s why you should make sure only a single entity is controlling this field on your behalf. MMPs are the best choice. 

MMPs are not only a trusted, unbiased source: they provide extensive ad network integrations. Ultimately, SKAdNetwork doesn’t exist in a vacuum, and enrichment with other attribution models can greatly enhance campaign optimization.

In terms of technical resources, MMPs also provide seamless client/server SK sync, scale-driven machine learning, and extensive engineering support to help you stay future proofed for later versions of SKAdNetwork.

Key takeaways 

Although iOS 14 and SKAdNetwork mark a major shift in the mobile marketing landscape, it is still possible to navigate those changes without compromising much on measurability or your ability to predict user LTV. 

Remember:

1. Although the 6 bits that make up conversion values can seem restrictive, an effective bit strategy gives scope for advanced measuring.
2. With the limitations of SKAdNetwork, predictive modeling is now a must-have for marketers.
3. MMPs can greatly aid your ability to deliver high-performing campaigns, and there is immense value in allowing them to control your conversion values.

The post Unlocking the power of SKAdNetwork’s conversion values to measure and predict campaign value appeared first on AppsFlyer.