Apple's new policy & impact on digital performance marketing

There had been lot of speculations within the industry on the new Apple iOS 14 privacy policy since Q4 2020. It was supposed to be rolled-out in January 2021, but then later delayed to the end of Q1 2021, to be finally launched on 26th April 2021.

I would like to share some information on the possible impact on iOS digital performance campaigns due to Apple's latest privacy policy. At Careem, we discussed with growth teams around the world to find some conclusions, hypothesis and the way forward in the new world dominated by limited ad tracking (LAT) for iOS devices.

This article is an effort to explain the basics of what has changed (for non-technical readers to understand) and later covers detailed technical information along with estimations and hypothesis on possible impact and the way forward.

Table of contents

1. What is Apple iOS 14 privacy policy? 
2. What exactly Apple has done?
3. What does it mean for iOS app advertisers?
4. What does it mean for digital acquisition performance marketing?
5. What does it mean for digital engagement performance marketing
6. Which brands are impacted more?
7. Which markets are impacted more?
8. Timelines of impact
9. What does it mean for performance creative strategy?
10. Segmented vs general performance marketing
11. Action plan
12. Platform specific updates 
13. Thoughts? 

1- What is Apple iOS 14 privacy policy? 

Apple wants advertisers not to track individual users against their behavior, interests and actions in context of privacy.

2- What exactly Apple has done?

Apple has made it mandatory for each iOS app to ask from their users if they allow the app advertiser to track user-behavior (through mobile device identifier) for marketing purposes. This is done through a framework and communicating with user though a native pop-ups from the app - called AppTrackingTransparency (ATT)

2-1 > Device IDs > Apple has stopped allowing advertisers to track iPhone user's device IDs without their consent. (before this change, the default opt-in rate for MENA region was ~97%)

2-1 > Opt-out users > Advertisers (promoting iOS apps) will no longer be able to target iPhone users for personalized paid digital marketing unless they opt-in specifically (expected global opt-in rate is ~15% to ~30%) [ref_1]. The opt-in rates should vary for brands with high or low market equity, however, only time will tell what percentage of the users are opting in per brand / app.

[Update as of 16th May 2021] > During the first 20 days of the iOS 14.5 release, the overall global opt-in rates are around 14% - this is close to the range of expected opt-in rate of ~15% to ~30%.

3- What does it mean for iOS app advertisers?

Digital key performance indicators (KPIs) to take hit due to lack of precise user level targeting, tracking, measurement and optimization.

3-1 > Targeting > Apple iOS users won't be precisely targeted for either acquisition or engagement campaigns (unless they opt-in). This lack of precise targeting will result in compromised performance for acquisition or engagement campaigns.

PS: details for targeting will be discussed later under acquisition and engagement section

3-2 > Tracking > Advertisers have to rely on Apple's new tracking framework (called SKAdnetwork) which is limiting the conversion / event tracking to specific time window (mostly first 48 hours of the install), this implies that not all user conversion data will be tracked.

3-3 > Measurement > Granular user-level data wont be measured without user consent. There is no solution as per now for individual user performance measurement for digital marketing platforms

3-4 > Optimization > Marketing platform's (such as Google, Facebook etc) machine learning based optimization capabilities are supposed to be impacted, this means the way these platforms used to operate to deliver performance needs to be re-looked. Also, with lack of tracking and targeting, the campaign level manual optimizations done by digital experts are not the same which results in compromised overall campaign performance.

The following snapshot summarized what's changed, impact, available solutions and their effectiveness.

Additionally, on the impact - there is a data science component to performance marketing that will struggle to build models such as look-alike, channel specific LTV which are used to target potential high value customers, cross-sell, and exclude already acquired or low-value users. 

Here is a very interesting short video on the updated Apple's user privacy policy - Its great how they guys explained this topic in a funny way :)

4- What does it mean for digital acquisition performance marketing?

Lack of targeting the right audiences comes up negative implication on acquisition campaigns. Targeting the right audiences simply means showing ads to users who have the app installed.

4.1 > Audiences > Apple iOS users won't be targeted specifically for acquisition campaign (can't exclude all existing app users from an acquisition / install campaign) - This means the acquisition ads will be shown to users who have, or doesn't have the app installed. Which means that part of the budget which is spends on users - who already have the app installed - is not contributing towards acquisition, in-fact, its helping with engagement.

4.2 > Platforms > Advertisers cannot precisely exclude complete iOS install base on platforms such as Facebook, Instagram, Google, Twitter, Snapchat etc. (Apple Search Ads is the only exception as it is compliant with iOS 14.5 and rely on iCloud accounts instead of device IDs or Identifier for Advertisement (IDFA)

4.3 > KPIs > CPI > For app based businesses, cost per install (CPI) is expected to go up - Increase in KPIs is directly proportional to the acquisition budget spent on users who already have the app installed. If platforms started to rely on other signals (besides IDFA) to identify if a user already have the app installed, then the increase in CAC can be controlled. So far (as of 5th May 2021), there is no such information on how the marketing platforms (such as Google and Facebook etc) will deal with this situation except to reply on Apple's provided reporting methodology called SKAdnetwork .

4.4 > KPIs > CAC > Similar to CPI, cost per acquired customer (CAC) will also go up but this KPI will have more impact than CPI because the acquisition ads wil be shown to users who already have the app + the conversion / event tracking is limited to a specific time window(s) . In short, user events will be allowed to be tracked for a limited time frame (mostly within 48 hours), after the time expires, the conversion event will not be associated to the acquisition campaign. Hence the real performance of the campaign will not be know.

PS: Conversion tracking window will be explained in detail later in the article

4.5 > User level tracking > For all opted-out users, there will be no user-level tracking. User-level tracking means an individual user / device ID will no longer be connected to a digital channel / campaign. Previously, each individual device ID was tracked against all user actions / conversions on the app. However, now for opted-out users, there is no such thing as user-level details, the lowest level of details provided are campaign level through SKAdnetwork.

4.6 > What's best advertisers can do > There is very limited advertisers can do to exclude existing iOS base / installers (using phone numbers or emails) to run acquisition campaigns. As per estimates, roughly ~20% to ~30% of the existing iOS install base can be removed from digital acquisition campaigns - this is considering 1) marketing platform (Facebook, Google etc) audience match rates and 2) install base without emails / phone numbers.

4.7 > Why only ~20% to ~30% base exclusion > Lets assume following scenario in which an advertisers receive 100 installs, 50 signups from installs and 25 new users from installs. In order to exclude the installers post 14.5 era, the advertisers has to rely on signed-up user data (email and phone numbers). If the signup-rate is 50% then only 50% of the install base has emails and phone numbers. Uploading these signed-up users data to platform in order to exclude from acquisition campaign will only help with the industry match rate of around ~35% to ~55% (lets has 50%). In short, out of the total installs, only (~20% to ~30%) of the install base can be excluded from acquisition campaigns. The remaining 75% of the install base will still be seeing the acquisition ad.

PS: no change for CRM systems including push notifications, email and small message services (sms)

5- What does it mean for digital engagement performance marketing?

Advertisers tend to target different stages of the funnel and customer journey to keep them engagement. Some of the major use-cases for digital engagement is re-activation. cross-selling and activation. All engagement use-cases are impacted due to targeting capabilities for opted-out users.

Lets consider different stages of the funnel:

5.1 > Installed not signed-up users > These are the users who have installed the app but have not signed-up so far. This means the advertiser doesn't have the user email or phone number which can be used to target users on digital platform. If an advertiser wants to target this specific cohort then its not possible for opted-out user now because there is no common denominator between advertiser and the user device (previously it was device ID)

5.2 > Signed-up not converted users > These are the users who have signed-up with the business but have not converted so far. This means the advertiser have the user email or phone number which can be used to target users on digital platform. If an advertiser wants to target this specific cohort then its possible for opted-out user but this also has limitations. The major limitation is audience match rate with the platform (Facebook, Google etc). Not all users who have signup also hared the same email or phone number on their social profile. As per the industry standard, match rates hover around ~35% to ~55%. This means, on average ~55% to ~65% of the audience will still not be targeted.

5.3 > Any stage of the funnel after sign-up > Since digital engagement has several use cases including re-activation, churn control, cross- selling etc, any other stage of the user funnel targeting will be depended on the user email / phone which automatically brings down the total addressable audience by ~50%

6- Which brands are impacted more?

Apple's new privacy policy stands same for all but it may probably have different impact for different brands depending upon brand trust and penetration. In short, brands with high trust and market penetration will relatively have less impact in comparison to new brands or brands with low market share.

6.1 > Brand trust > If a brand has built positive impression then its highly likely that the opt-in rate for that particular brand will be high. Appsflyer has already shared study in which they have seen some brands having initial opt-in rates of up-to 46%.

6.2 > Brand penetration > Apps with high penetration is the market has pretty much already acquired a considerable share of the market. This means the acquisition budget split between users who already have the app installed and who doesn't will be distributed relatively better as compared to the brand which is new or with low brand equity in the market. Percentage increase in CPI depends on the brand penetration and existing market share. The higher the market share within in a geo, the lower will be the impact on CPI.

6.3 > Impact estimation > Its not easy to estimate the impact iOS 14 privacy policy can bring, however based on certain assumptions and hypothesis, following seems to be the impact brands may have.

• Brands with high penetration and trust are expected to have ~46% to ~53% impact on the iOS campaigns CAC - this is subject to iOS 14.5 fully rolled-out. Opt-in rates for such brands is expected to be ~30% to ~40%.
• Brands with low market penetration and trust are expected to have ~84% to ~91% impact on the CAC - this is subject to iOS 14.5 fully rolled out. Opt-in rates for such brands is expected to be ~2% to ~5%.

Opt-in rates are also impacted by how seamlessly a user has been asked to opt-in for advertising. Higher the opt-in rate for an app, the better performance marketing can deliver results.

7- Which markets are impacted more?

Markets with high iOS share [ref_2] will be impacted such as Saudi Arabia, United Arab Emirates, United Kingdom etc. No significant impact on Android heavy markets such as Pakistan, Algeria, Palestine etc

8- Timelines of impact?

This increase in CPI and CAC is expected to happen gradually over a period of four to five months.

8-1 > Adoption rate > Adoption rate for iOS 14.5 is expected to be somehow similar to the following iOS 14.3 - apparent reason believed to be Apple wants to roll-out this gradually just in the best interest of all and give enough time for advertisers to gradually learn.

8-2 > Complete roll-out > End of Q2 2021 should have minimum 55% iOS 14.5 adoption with Q3 2021 should end with ~85% adoption. Following graph shows the actual iOS 14.3 adoption rate over a period of six weeks. [ref_3]

9- What does it mean for performance creative strategy?

Since the target audiences can not be precisely defined for opt-out users, the asset communication should make sense for audiences at different stages of the funnel. Performance marketing creatives needs to be generalized considering unique selling proposition and call to actions.

9-1 > Call to action (CTA) > From the above example, instead of saying "Install Now", lets have a generalized version such as "Order Now", "Buy Now" or "Book Now" etc because the creative will be seem by users who already have the app installed (we have already established that roughly ~25% of the install base can be excluded from acquisition campaign))

9-2 > Unique Selling Proposition (USP) > Unique selling proposition should not be limited to specific users of the funnel. The above example offers 20% discount on first three rides of the users, whereas the due to iOS 14.5 audience reach limitations, exiting users will also be seeing this ad and the communicating becomes conflicting which may lead of bad user experience.

10- Segmented VS blanket digital marketing approach?

iOS app marketing seems to be currently shaping close to traditional offline marketing with an online ad-placement, where advertisers will have less control on the target audiences. Its easy to understand from an offline billboard advertisement where a potential, non-potential or existing customer see the advertisement - the efficiency and tracking is pretty much questionable for a specific cohort.

10-1 > Digital marketing approach > Reporting > Until the advertising industry have a solid solution to this (limited tracking and reporting) situation, what KPIs should a digital performance campaign be reported against? As we already established that there is no such thing as pure acquisition or engagement campaign, then what's best way to move forward with this? Some of the experts believe that digital performance campaigns should start reporting against both acquisition and engagement combined - this too will be reported for only tracked users. Following are some the metrics advertisers may consider to report acquisition and engagement combined under SKAdnetwork campaigns.

L0 is the most important KPI, following by L1 through L3. L0 is overall performance output of tracked users (either though acquisition or engagement). Total revenue generated divided by overall spend on the campaign.

11- Product team action plan

Product and technology teams have to take certain action on the app to ensure compliance with Apple's privacy policy - this simply means to update app SDKs and working with performance marketing team on the AppTrackingTransparency (ATT) pop-up.

11-1 > Software development kits > Existing SDKs which are trying to track users should be updated with new ones while considering following two options:

1. App users may not be asked to allow advertisers to opt-in for tracking, and this by default will not save users device Ids for tracking / attribution purposes. Some companies are specifically taking this route to learn the implications of adding another step (consent pop-up) at the top funnel before having this for wider audiences.
2. App users should be asked to allow advertisers to track user behaviors by sharing consent using Apple’s AppTrackingTransparency (ATT) framework (the pop-up which a user will respond to allow or not to allow advertisers for tracking and targeting purposes)

The new SDKs (specifically for mobile marketing platforms such as Adjust, AppsFlyer etc) should ensure that other tracking methodologies are also not in place to track the users such as finger printing [det_1]. Early May 2021, Apple rejected several apps which moved away from tracking device IDs, but trying to track / attribute using finger-printing methodology [ref_4]. Most of these apps had Adjust MMP SDK integrated which was eventually updated on immediate basis. REF:

Marketing platforms also have their SDKs which are supposed to help with performance marketing and lifting some limitations with the standard SKAdnetwork methodology. Snaphat has recently released a new version while considering iOS 14.5 limitations. Similarly Facebook has upgraded its SDK and encouraging advertisers to move to the latest version.

11-2 > AppTrackingTransparency (ATT) - iOS app advertisers must use AppTrackingTransparency framework if the app collects data about end users for purposes of tracking and targeting. The AppTrackingTransparency framework presents an app-tracking authorization request to the user and provides the tracking authorization status. In simple words, its the pop-up shown to the app users to allow or disallow advertisers to track user for marketing purposes [ref_5].

The important part here is how the pop-up is designed to maximize the opt-in rates. Following is Facebook opt-in screen request (left side) - once the user press "Allow", the native pop-up has to be shown (right side) for the final user consent input. Its critical to conduct experiments on the opt-in screens to learn which gives the best opt-in rates - the higher the opt-in rates are the more control advertisers will have to deliver better. [det_2]

11-3 > Measurement & Success for ATT > Some of the metrics which should be tracked closely while moving towards ATT framework are:

• Percentage existing users who opt-in
• Percentage new users who opt-in
• Percentage change in funnel (install to signup to new user)
• Conversion metrics - other relevant conversion events relevant to the app business model.

12- What is SKAdnetwork?

Its important to have good understanding of how the new limited tracking ecosystem works - marketers should have knowledge and understanding about Apple's new attribution / tracking framework called SKAdnetwork.

SKAdnetwork framework will operate differently for tracking and attribution app users in comparison to how the mobile marketing platforms (MMP) operate. One of the fundamental difference is MMP is a middle-man between advertiser and publisher, where as in SKAdnetwork, advertiser is the last one to receive the tracking and attribution information from publisher.

12-1 > Data flow of Mobile marketing platform (MMP)

MMP Data (which is owned by advertiser) has complete transparency of user at every stage of the funnel. Impressions and clicks data is sent directly to MMP from the publisher side, whereas the conversion data is sent directly from advertisers app to MMP (though SDK). These two primary data sources are then finally processed to have complete user level reporting of behavior on apps from different publisher platforms.

12-2 > Data flow of SKAdnetwork

Following is a simplified snapshot of how SKAdnetwork operates, which clearly states that advertiser is the last to know about the conversion information.

The above is a holistic view of SKAdnetwork, However, to really understand, one have to understand the following flow diagram on what exactly is happening at each stage of the user from clicking an advertisement to installing and opening it for the first time.

12-2-1 > Data flow of SKAdnetwork > Conversion values

Conversion value is an action done by user on the app - mobile app users perform certain actions on the app which every advertiser wants to track to better understand user behavior and optimize digital campaigns. In the new world of SKAdnetwork, advertisers are limited with the numbers of user events to track - which are only six events. The final reported event will only be one per user (from the six) which has the highest conversion value [det_3].

Conversion value received (explained in detail under section 12-2-2 > Data flow of SKAdnetwork > Timer concept) can be organized into receiving much more information rather than just one user event. This can be done by understanding how the conversion values are organized, what is the structure of conversion value and how growth and product team can align on tracking the user behavior and action information.

Conversion values are 6-bit binary data which can be interpreted into 64 different way (lowest value being 0 and highest 63). These conversion values can be set in both basic or advanced manner.

Following is an example in which these conversion values can be configured to give basic information of the user event. Each of the three examples are only communicating just one thing - example 1 communicate user revenue, example 2 communicate if a user has signed up and example 3 is communicating that a user has made a purchase and become a new user after making a conversion.

However, conversion values can be configured in an advanced manner to communicate a lot more information which varies per business. For example a gaming app would like to track revenue generated along with levels user has achieved in the game.

Here is an example of how a ride hail business could prefer to track the conversion values while considering cohort, signup data, predictive life time value (pLTV) & revenue information within one conversion value.

From the 6-bit conversion value - the above example is divided into getting four attributes per user. Three days cohort and revenue are 2-bit, where as signup and predictive life time value is 1-bit.

1. Conversion days: (3 days cohort) > Communicate time (in days) a user took to convert. This works well for businesses where most of the users converts within 3 days of install.
2. Signup information > Communicated if a user has signed-up or not. This is important if users doesn’t convert within allocated time. (T1 & T2) 
3. Predictive LTV information > The deeper the action of user in conversion funnel the higher the LTV would be. However, this will holistically classify user as LTV+ or LTV- due to the limitation of 1-bit.
4. Revenue information > Revenue indicator is simply if the user is high value or low value. 1, 2 or 3 can be interpreted as three different average order values (AOVs). 

12-2-2 > Data flow of SKAdnetwork > Timer concept

The infamous timer concept is specifically designed to send conversion value if its happening within two particular time windows (let's call them T1 and T2).

• T1 is the 24 hours window where the attribution and conversion value (if any) is sent to the advertiser. This time is basically to capture any possible conversion to happen on the app to be finally sent when T2 starts. T1 is subject to reset (for a maximum of six times) only if higher conversion value is happening within the next allocated 24 hour window. (Conversion value have been explained in detail under section 12-2-1 > Data flow of SKAdnetwork > Conversion values)
• After the T1 expires (first 24 hours or multiple 24 hours window with a maximum rest limit of six times), T2 is initiated and will send the install attribution and conversion data anytime within its window of 24 hours. Once T2 expires, no further conversion data will be sent, irrespective of any number of conversion happened.

Lets consider following two scenarios where we have 1) no conversion data within first 24 hours and 2) having conversion data within first 24 hours window.

In the following scenario, a use installs the app and doesn't do any conversion event on the app within 24 hours window. This means, there is no conversion event during T1, so once T2 starts, only the install attribution source is sent to advertiser.

In following scenario, a user installs the app and did a conversion event within first 24 hours (green), so this means the T1 will be reset to give another 24 hours to track any upcoming conversion. Next T1 window also have a conversion (blue), which means another reset of T1. For the next T1 reset, the user has not done any conversion (grey). This means the T1 will expire after 24 hours completion and the T2 will trigger. Anytime within T2, SKAN will send attribution and conversion value to the advertiser (to be viewed under SKAN reporting).

Its important to note that:

• Not all three conversion events will be sent, only the one with highest value per user will be sent, which will always be the last conversion event.
• T1 will only rest if the conversion value is higher than the previous conversion value.
• Lastly, Apple doesn't provide reporting interface to visualize the conversion data (Apple just server the functionality by passing only one SKAdnetwork post-back per user to be attributed network / publisher - and the network forward conversion information to the advertiser or app). Most of the Apps are integrated with mobile marketing platform (such as Adjust, AppsFlyer, Singular etc) which have build the reporting interfaces for SKAdnetwork tracked conversions. If you are an advertiser and working with any MMP, then the reporting interface is most likely to be just configured.

13 - Marketing team action plan

Digital marketing teams has to perform certain actions to ensure best possible tracking, campaigns optimization, and to comply with all the changes required on marketing platforms.

13-1 SKAdnetwork Setup > Digital marketing team should ensure to have at-least the basic conversion value setup in-place to ensure tracking of user behavior within the limitations of SKAdnetwork. The basic setup can be easily done if the mobile marketing platform is integrated to the app - all whats needed is to configure the six most important user events to be passed on to the SKAdnetwork.

Following is the snapshot of Adjust MMP interface which allows to select six custom events for reporting purposes.

13-2 > Digital marketing platforms > Introduction of SKAdnetworks has several implications on marketing platforms - which means social platforms such as Google, Facebook etc have to comply with them in order to facilitate the advertisers. Advertisers has to consider following changes, which are:

• Campaigns limit > There are now limitations on the number of SKAdnetwork enabled iOS campaigns / ad-groups. Total allowed SKAdnetwork campaigns per app per platform is 100, however, most of the platforms are allowing different limits for maximum number campaigns / ad-groups.
• View attribution > View-though attribution is now available with SKAdNetwork 2.2 - which enables advertisers to see the performance of different ad-formats such as video, audio at view level user interaction [ref_6].
• Deep linking > Deferred deep-linking is not available for SKAdnetwork enabled iOS campaigns.
• Engagement performance > No app engagement campaign performance visibility because re-attribution will no more work for SKAdnetwork campaigns.

13-2-1 > Digital marketing platforms > Publisher's Limitations > Updates on the publishers due to SKAdnetwork are slightly different for each major social platform.

• Facebook / Instagram - Allows a maximum of 9 campaigns with up-to 5 ad-groups per campaign [det_4].
• Snapchat - Allows a maximum of 10 ad-groups (can fall under any SKAN enabled campaign) and also suggested to have a new ad-account for all iOS +14.5 campaigns [det_5].
• Google - Recommending to have 8 campaigns per app - though its not a hard limit as Google claims to use other 92 campaigns for machine learning purposes but recommends not to exceed more than 8 campaigns per app, else it may impact the reporting & performance. No particular limit on the number of ad-groups, they are still the same (100 per campaign). Also, bidding on tCPA may result in better scale and more stable performance vs tROAS bidding [det_6].
• Twitter - Allows maximum of 70 ad-groups per app (within any number of campaigns). For now, Twitter is not supporting view-through attribution but it will come soon (as per 29th May 2021) [det_7].
• TikTok - Limit of 11 active campaigns and 1 ad-group per campaign for each iOS app [det_8].

Platform registration - each marketing platform has to be registered with Apple's SKAdnetwork to function - most of them has shared comprehensive guides on changes required at the account level [det_9].

13-2-2 > Digital marketing platforms > MMP updates > If you’re using a Mobile marketing platforms (MMP), you should follow their specific instructions on how to ensure the best possible tracking.

• Adjust [det_10]
• AppsFlyer [det_11]
• Kochava [det_12] 
• Branch [det_13]
• Singular [det_14] 

13-3 > Experiments > Marketing team (along with product team) should own the learning and testing of consent pop-up to maximize the user opt-in rates. Experiments will eventually ensure setting up the best message to maximize consent for advertising. Ideally, following should be considered while working on the pop-up.

• Competitors > What are other apps doing, and could I gain an advantage from higher performing user-acquisition campaigns?
• Networks / Publishers > Are there measurement or performance gains from being able to match identifiers with networks?
• User Experience > How does the ATT prompt affect in-app user conversion behavior and user on-boarding? Consider this from a funnel perspective - its simply adding one more step to the user-funnel, which may lead to some percentage drop in install to signup or signup to first order funnel metrics. Also, are there any other prompts or permissions that users contend with? such as push notification pop-up?

13-4 > Giving knowledge within the organization > Digital marketing team should ensure to develop basic understanding on the implications of SKAdnetwork among other organizational stake holders before the impact really took over. Recommendation is to take a pro-active approach and give basic understanding to stake holders (such as finance and product) on how much it may impact the CAC. Ideally a redo of the digital budget plan with increased CAC for iOS over the year (expected increase in CAC for iOS may vary from 25% to 50% depending upon the brand equity and user consent rate)

13-5 > Budget shift > Android vs iOS budgets are also a question since the iOS tracking is compromised. Markets would still prefer to have more control and transparency against the amount spent. Organization may consider skewing some budget towards Android vs iOS.

Some organizations with double digit millions of dollars per annul budgets are also redoing the yearly budget plans while considering the iOS 14.5 adoption rate over the year 2021 and gradually moving some percentage of budget towards Android campaigns in an effort to control the CAC.

Following snapshot share by Singular (MMP), which shows that Android spends have increased by 8% since Apple's new privacy policy is enforced in April 2021.

13-6 > Fraud from ad-networks - Markets who work with ad-networks understand the fact that several types of frauds happening due to click-spamming, click injection, SDK spoofing (to add fake events) etc. As per the estimate ~50% of the installs delivered from ad-networks in MENA region are fake! The fraud is not only sending fake installs and conversions but also real-user conversions stolen from other channels (called attribution or organic hijacking). Industry has developed over-time to provide some specialized tools to catch such frauds to save advertises money. Such tools include, and not limited to mFilterIT, Machine, TrafficGuard, Scaler, Intercepted etc.

Following snapshot shows fraud percentage by type in MENA region [ref_7]

Introduction of SKAdnetwork for opt-out users also limit the efficiency of fraud tools due to data concerns and the way SKAN works. These specialized fraud tools have to make changes to provide fraud reports within the best available data to ensure the ad-network payout has minimum fraud percentage.

13-6-1 > Data concerns > No user level data > These fraud tools run analysis on user level data, which is no more available after SKAdnetwork. This means the fraud analysis has to run only on aggregated data (not user level) which gives a lot of opportunities to fraudsters to manage fraud within the aggregated approach, hence catching the fraud has become more challenging.

13-6-2 > SKAN functionality concerns > The other set of concerns are more towards not having encrypted data. The way SKAdnetwork operates is by sending data first to ad-networks and then finally to publishers. Within this, there is a probability of encrypted data being modified.

• Conversion value is not encrypted and can be modified > Cannot rely on event based optimization (event spoofing)
• IP address (which is only way to look at geo) is not encrypted > Cannot rely on GeoLocation information (invalid geo)
• SKAN supports view-through, but doesn’t validate the impressions > Spamming is ever-prevalent and will continues (click spamming)
• SKAN TransactionID can be “replayed” to escape detection > cannot rely on “encryption” by Apple (fake device)

These fraud tools have started to focus more on the click / impression fraud detection because they are limited to only aggregated data (not user level data) without complete conversions.

13-6-3 > Adnetwork Payout > If an advertiser is working on impressions and clicks (which almost no-one works) then there is no change. However, if an advertiser is paying for installs or conversion events, then the best possible reporting is based on aggregated fraud percentage at a campaign / source level (not user level). This means that fraudsters can make their overall numbers look fine (as per industry standard CTR, clicks to installs, install to conversion etc) by hiding under the umbrella of aggregated data.

13-7 > Alternates to SKAdnetwork > Some mobile marketing platforms (including Adjust) has worked on secondary attribution solution which are not depended on device IDs and fingerprinting technology. Its simply gathering following device information to attribute installs against specific clicks or views.

1. Device type (e.g. phone)
2. Device name (e.g. Samsung Galaxy S7)
3. Operating system (e.g. Android)
4. Operating system version (e.g. 7.1.2)
5. IP address (e.g. 77.185.208.234)
6. User agent (e.g. browser and operating system)

From the above six device attributes, some can change over short period of time (such as IP address), therefore, Adjust (MMP) claims to attributes a users when all six are matched with install (means the six user-attributes captured upon a click / impression are matched with the six attributes from Adjust SDK upon the install) [det_15]

14 - Thoughts?

• Some consumers may believe that personalization is great because it enables advertisers to show relevant advertisement based on their interests & behaviors. However, the obvious counter part is users don't want to be tracked / chased in the context of privacy.
• From advertiser's point of view, iOS marketing seems to be going back in dark ages where there was no such sophistication / options in the in digital acquisition, re-activation, cross-selling and retention.
• Lastly, this situation created a vacuum in the industry for martech companies to take this as a challenge and come-up with some solution without impacting Apple's privacy condition.
• Do you think this is going to happen with Android?

Please do share your feedback, suggestion, ideas on this major privacy change which could add value for others!

15 - References & Details

• Reference_1 > https://www.flurry.com/blog/ios-14-5-opt-in-rate-att-restricted-app-tracking-transparency-worldwide-us-daily-latest-update/
• Reference_2 > https://gs.statcounter.com/vendor-market-share/mobile/united-arab-emirates/#monthly-202003-202103
• Reference_3 > https://www.flurry.com/blog/ios-14-5-app-tracking-transparency-idfa-release-adoption-upgrade-apple-users/
• Reference_4 > https://developer.apple.com/app-store/user-privacy-and-data-use/
• Reference_5 > https://developer.apple.com/documentation/apptrackingtransparency
• Reference_6 > https://developer.apple.com/news/?id=wajvzt18
• Reference_7 > https://app-benchmarks.adjust.com/?vertical=All&region=All&user_type=All&platform=All&start=2019-10-01%2000%3A00%3A00%20%2B0000&end=2019-12-31%2023%3A59%3A59%20%2B0000
• 
  
• Details_1 > https://pixelprivacy.com/resources/browser-fingerprinting/
• Details_2 > https://help.adjust.com/en/article/ios-14-user-privacy-frameworks#AppTrackingTransparency
• Details_3 > https://www.appsflyer.com/blog/skadnetworks-conversion-values-power/
• Details_4 > https://www.facebook.com/business/news/how-to-prepare-for-changes-to-facebook-ads-from-ios-14-update
• Details_5 > https://businesshelp.snapchat.com/s/article/snap-ios14?language=en_US
• Details_6 > https://blog.google/products/ads-commerce/preparing-developers-and-advertisers-for-policy-updates/
• Details_7 > https://business.twitter.com/en/blog/twitter-ios14-updates.html
• Details_8 > https://www.tiktok.com/business/en/blog/supporting-our-partners-through-ios-14-and-beyond
• Details_9 > https://help.adjust.com/en/article/skadnetwork-partner-integrations
• Details_10 > https://www.adjust.com/blog/topics/ios-14/
• Details_11 > https://www.appsflyer.com/ios-14/
• Details_12 > https://www.kochava.com/ios14-idfa/
• Details_13 > https://help.branch.io/faq/docs/ios-14-faqs 
• Details_14 > https://support.singular.net/hc/en-us/articles/360047706852-Working-with-iOS-14-FAQ-for-Partners
• Details_15 > https://help.adjust.com/en/article/attribution-methods#probabilistic-matching