UPDATE - September 4, 2020
Yesterday, Apple confirmed that they will postpone the required adoption of the IDFA opt-in till early next year.
However, while the timeline has changed, the course of action has not. We feel the following recommendations remain relevant for all app marketers running performance marketing.
Since 2012 (9 iOS versions ago!) our focus at Jampp has been to help your apps grow by developing cutting-edge technology according to the technical capabilities and privacy safeguards of the different operating systems, and the ad-tech ecosystem.
During this journey, we faced many industry-wide changes that helped shape the foundations of our core technology. As disruptive as the IDFA update is, we believe that all efforts towards enhancing user privacy should be embraced and welcomed.
Our priority is to develop technologies that fully respect users' decisions while allowing advertisers to reach their potential customers (as well as reconnecting with their current ones) in the best possible way. This means ads that are less intrusive and more relevant for users, more cost-effective for advertisers, and more profitable for publishers.
Yet the upcoming changes will affect our ecosystem the most. Let’s go over how those will affect Jampp’s services.
What’s changing in iOS 14?
Unless you’ve been living on another planet for the last few months, you are probably aware of what’s coming (if you have been under a rock though... I’d encourage you to stay there for the time being—I don’t even know where to start). Just in case, a quick recap... all apps in iOS 14 will be forced to ask for permission from the user to store and use Apple’s Identifier for Advertisers (IDFA) for tracking purposes (the same way they ask permission for notifications, location, camera, etc.).
Certain aspects of the copy will be customizable by publishers, the above example was provided by Apple.
Technical Note: If the user doesn’t allow the app to collect the data, the IDFA value will be zero’ed (000-000-00...) in the same way as the current “do not track” option that is available at the operating system level, making it impossible to track.
When users opt out of IDFA tracking, this diminishes the ability to serve personalized and retargeting ads, as well as the ability to accurately attribute their marketing campaigns and understand campaign ROI.
The IDFA isn’t technically gone, but because very low opt-in rates are expected, the industry needs to embrace alternatives.
To be clear, users that choose not to be tracked will still be served ads, but it will be harder to show them relevant ads and it will be harder to understand the performance of those ads.
How SKAdNetwork will replace Identifier for Advertisers (IDFA)
Apple will offer an attribution solution called SKAdNetwork if and when users opt out of IDFA.
SKAdNetwork allows marketers to track their campaigns to determine which ones led to installs or purchases, but without disclosing granular, user-level data.
Unfortunately, SKAdNetwork has significant limitations such as including only aggregated campaign data, no impression or view-through data, very limited down-funnel reporting, and no control over attribution windows. For example, we will know that a certain campaign gets an install (or an in-app conversion) but we will have no information about which user made that install (or even which creatives were served).
So, depending on the user consent there will be different scenarios:
- If a user accepts to share their IDFA: Attribution can be done deterministically with IDFA (like now).
- If the user doesn’t accept to share their IDFA:
- SKAdNetwork (Apple’s Solution) will handle attribution and the limited tracking described above: we believe this is going to be the main method.
- Probabilistic Attribution by the MMP: by using different signals (time, device, user agent, etc), MMPs will be able to match-attribute ads with conversions. This isn’t new, it’s something that existed for a long time, but the industry gradually switched to IDFA as it brought a better way to attribute.
How are MMPs responding to the changes?
By now, all MMPs have announced their plans for how to deal with the changes. In all cases they will combine the available attribution methods, including IDFA (with users who opt-in), probabilistic attribution, and SKAdNetwork with tools for Conversion Value mapping. We have been working with all of them to make sure our integrations are up to date. This is the information as of now from the main players:
- Adjust: For more information, we recommend reading their overview of the changes, and the announcement of their latest update to their SDK.
- AppsFlyer: Their approach for Probabilistic attribution is slightly different as they plan to do Aggregated Attribution when no IDFA is available. This means DSPs and ad networks won’t know exactly which impression or click generated the conversion; therefore, optimization will be similar to SKAdNetwork. For more information, see their SKAdNetwork Solution and FAQ: Impact of Apple (iOS) Limit Ad Tracking on attribution.
- Branch: To learn more, visit Branch’s resource hub for all iOS 14 related questions.
- Singular: If you’re just starting to look at this topic, we recommend their webinar for a clear overview and their FAQ’s post for specifics.
- Kochava: For more information visit their resource center for all questions regarding iOS 14.
How to prepare your app for the changes in IDFA
Get your app ready to support the AppTransparency Framework
Your Product team is probably already working on this, but if not, your app will need to implement the App Tracking Transparency framework so users can opt in to share their IDFA for tracking purposes. This way, when users opt in on the publisher side and on your app, IDFA matching attribution can continue to work.
And it will also need to implement the SKAdNetwork framework for this new attribution method to work to track installs and also post-install conversion values.
Fortunately, most MMPs (Appsflyer, Branch, Kochava, Singular, Adjust…) are providing support for both frameworks automatically with their latest SDKs.
- Talk with your MMP and make sure you update to their latest SDK version that will include AppTrackingTransparency framework and SKAdNetwork framework management to simplify your workflow.
- Work with your Product team to understand how to optimize the User Experience flow for the opt-in, for example, by adding informative messaging before the opt-in pop-up appears. We recommend experimenting and A/B testing different copy and moments to display the pop-up. It’s important for people to understand the value they get in exchange for sharing their data (in most cases, hours of free entertainment).
Design your Conversion Values
One limitation that SKAdNetwork has is the fact that the app will not be able to send multiple in-app events and revenue data points.
There’s a limitation on the number of different conversion data points to be sent (precisely 6 bits of information) that needs to be leveraged to inform Jampp (and all your paid acquisition channels) about the value of each conversion so we can optimize towards your business outcomes (either CPI, CPA or Target ROAS/LTV).
Although 6 bits of data is really limiting, there are ways to use this value to send relevant data for optimization and lifetime value calculations.
Using independent frameworks like Elixir (video), or the solutions provided by the MMPs, you can use these bits to map conversion events or revenue information. There are many models that can be adopted, and each one will have different characteristics and benefits according to your business model and KPIs.
- Revenue model: 2 bits are used for encoding the count of days after the install from 0 to 3.
- Retention model: uses the 6 bits to encode the count of days after the install when the user opens the app every day.
- Simple event-based model: 3 bits used to map events (ie: 010 means "sign-up" and 011 means "purchase") while the other 3 bits are used to encode the days after the install.
- Predicted LTV-based optimization: uses 2 bits to encode the days after the install and the rest of the bits to send a predicted LTV value in incremental buckets.
It’s important that you let your DSP know which one you are using in order to ensure proper campaign optimization.
- Talk with your product team to align the development process on how to properly match conversion values with SKAdNetwork in a way that is helpful for your business goals and paid acquisition.
- Optionally, talk with your MMP to get access to their SKAdNetwork configuration tools for conversion value mapping.
- Talk with your DSP to share your conversion value mapping so they (we) can properly implement it in our platform and optimize towards your goals.
How does iOS 14 affect your programmatic campaigns?
Impact in app User Acquisition Campaigns (iOS)
Today, the majority of User Acquisition activity relies on IDFA (and deterministic user-level attribution) to correctly target, optimize, and attribute. As seen above, with iOS 14, that will change and UA campaigns will be attributed by different methods depending on the user consent: IDFA, SKAdNetwork, and Probabilistic models.
Jampp was one of the first DSPs to be approved on Apple's SKAdnetwork and we will provide you with full transparency on the spend, performance, and relevant metrics for each attribution method used in your campaigns so you can continue to be in control of your investment.
Impact in App Retargeting Campaigns (iOS)
Retargeting currently relies entirely on IDFA. So, it will be limited to those users who opt-in and give consent. No alternative attribution methods are available for Retargeting at the moment so all campaigns will be IDFA-based, for now. Jampp will continue to support iOS Retargeting campaigns using IDFA user-level data when users opt in, with the same level of precision and segmentation capabilities than today.
At the same time, we are exploring alternatives for user segmentation using probabilistic user targeting and matching that we will be announcing in the future.
Impact on Lift Measurement (iOS)
Always-On Lift Measurement relies on IDFA user-level information to automatically create a hold-out group that is not exposed to ads and track their behavior to compare it with the behavior of the users actually exposed.
Lift Measurement will continue to work perfectly within the universe of users who opt-in to share their IDFA. We are working with our Data Science teams to adapt our statistical significance thresholds to ensure proper lift analysis in an acceptable period of time considering that the universe of users to measure will be lower than today.
At the same time, we are updating our algorithms to enable us to augment the Lift Measurement we get from IDFA users regardless of attribution methods (probabilistic models or SKAdNetwork). This way, we will provide a holistic picture of the real incremental value of your campaigns at the same time that we protect user-consent.
What about Android?
None of the changes in iOS 14 has any direct impact on Android. While there is some speculation that Google will be following a similar path in the future, there have been no announcements to this effect, so Android will continue to operate “business as usual” for the time being. Naturally, everything we learn from the iOS experience will be potentially applicable to a world in which Android’s tracking might also be limited.
What to expect in terms of Performance
SKAdNetwork is a new attribution method that is still unexplored for advertisers, publishers, and networks alike. It’s impossible today to predict the impact that it will have on conversion rates and CPI/CPA/ROAS/LTV.
At the same time, apart from the change in attribution methods, the fact that a big part of the traffic will lack IDFA means that prices can fluctuate as well. We could potentially see an increase in prices in “IDFA traffic” as it becomes more scarce; and we could also see a decrease in prices overall, given that most of the traffic will not have IDFA and, without this data, bidding will become less aggressive. It’s hard to forecast which effect will be stronger.
We have already been testing different scenarios, including Limit ad tracking (LAT) traffic for iOS campaigns and probabilistic models to simulate SKAdNetwork attribution; and our machine learning models are developed to quickly adapt to market fluctuations and automatically adjust bidding tactics to ensure reach while achieving your CPI/CPA/ROAS goals. So, while some fluctuations in results can be expected, our platform will adjust bidding and optimizations to avoid major hiccups. Early results (at a small scale) are encouraging so far.
In general, we believe the adtech industry gets (an undeserved) bad reputation when it comes to its handling of user data. Most companies in the ecosystem (including DPS, SSPs, and MMPs) only obtain anonymous identifiers that have not been the source of any serious data breaches. We do not hold personally identifiable information as some of the walled garden platforms do.
We believe that app users should be empowered to choose if they want to share data or what level of data they are comfortable sharing and, of course, they should be aware of what they get in exchange.
Aggregated attribution, optimization, and ethical machine learning methods with no user-level data (either with SKAdNetwork or future alternatives) are here to stay and will help us all to build a more sustainable and user-centric industry.