Eric Seufert discusses in details the background of Apple's App Tracking Transparency privacy policy, Apple's potential motivation and how the policy will impact mobile marketers.
Advertisers thought you could just not use the IDFA (i.e not show the prompt) and instead use other identifiers. But Apple is not allowing for any loopholes (e.g. fingerprinting) that would undermine its new policy.
Back in December Facebook also revealed that the ATT prompt applies broadly to all mobile advertising, including mobile web campaigns.
No mechanism may be used for aggregating data at the user level for the purposes of tracking in the case of an ATT opt-out.
- For web campaigns, ATT governs the use of tracking pixels for user level data collection
- For app campaigns, ATT governs the use of the IDFA for user level data collection
For all campaigns (app and web), other identity options like device fingerprints, email addresses, et cetera, as well as back end conversion APIs like Facebook's, can't be used as they are now if the user has not given consent.
SKAdNetwork is how attribution will be done for mobile app campaigns for users that have opted out of ATT. Private click measurement (PCM) is how web events will be attributed for opted out users.
Both SKAdNetwork and PCM aggregate advertising data at the campaign level and allow for one conversion event (or “instrumented prioritized onsite event”) to be transmitted back to the advertising network in a postback, for every acquired user.
ATT is the future, and the most radical interpretation of how severely ATT will be enforced is probably the correct one. Advertisers need to adapt to that (rather than grieving or bargaining or conceiving of workarounds).
What might be the primary motivation of ATT for Apple is to regain control of content distribution on iOS, and move away from app discovery being driven by advertising.
Apple is most likely using ATT strategically to strengthen its market power, with respect to operating the app store. Because of this, Eric expects that ATT policy will be enforced to the maximum extent.
If products can't survive or thrive through advertising, then those products become more dependent on the platform operator for distribution.
Apple gets to define tracking, and then it gets to define the privacy spectrum. With its own approach at the extreme protective end, and tracking at the extreme exploitative end. But in fact, Apple’s privacy rules are at the middle of the privacy spectrum.
Users now expect all of their content to be available across all their devices. The idea of a direct relationship between a platform and a proprietary hardware (e.g. the App Store and the iPhone) is becoming anachronistic and ATT is Apple’s attempt to make this last longer.
If ad platforms can no longer efficiently connect supply and demand via serving ads, they'll try to keep content interactions on their own properties so that they can leverage first party data and sell advertising. They’ll become content fortresses. Example: OTT apps in China and Facebook with instant games, instant articles, FB shops, etc.
If all the publishing happens directly in Facebook’s services instead of Facebook serving as a routing service between publishers and users, it increases the dependency to Facebook and other platforms. It could result in less real choice for consumers and a much less competitive content ecosystem on mobile.
The SKAdNetwork framework that Apple is providing to advertisers is inadequate, unnecessarily punitive and injures advertising beyond what’s required from a privacy standpoint.
SKAdNetwork doesn't include parameters that are necessary for optimizing ad campaigns and that wouldn't reduce privacy protections. One example is the absence of a creative ID, which wouldn’t help attributing a specific user to a campaign yet is critical to advertisers (as Apple knows since there is a creative ID in the new Apple Ads attribution - AdServices).
As of March 2021, SKAdNetwork is completely dysfunctional: only a few networks are allowing advertisers to run SKAdBNetwork verified campaigns and even for these there are problems with reporting (receiving only a portion of postbacks, incomplete parameter information like conversion values and publisher site IDs).
Connect to existing consumers with first party data in as many ways as possible, so you can retain them.
As targeting gets less precise, acquisition will become less efficient and to stay viable you need a higher LTV over a longer lifetime.
Ad platforms use to do the filtering and provide qualified users. With less qualified users, the product will need to segment users into buckets and optimize the experience of these buckets to better monetize. It will need to go beyond early stage offers and content recommendation.
It will be hard to build niche products and advertise to small audiences, and it will impact B2C products in many app verticals that saw success with low DAU, high ARPDAU.
Just because impressions become cheaper doesn't mean all advertisers are able to maintain or increase spend. The entire funnel degrades (CTR, retention rates, monetization rates) and this impacts ROI.
Products will need to be appealing to a broader audience to reduce the negative impact of losing granular targeting.
Most ROI models will need to be completely recalibrated because the composition of cohorts will change as targeting becomes much less granular and LTV projections will change.
Only D0 can be used to project out to some target LTV, as you can only transmit data back to the network within a limited timeframe
Creative testing will be slower, because there are a limited number of campaign IDs you can use and most of them will be used by ad platforms to model ad group and ad creative performance.
Advertisers thought you could just not use the IDFA (i.e not show the prompt) and instead use other identifiers. But Apple is not allowing for any loopholes (e.g. fingerprinting) that would undermine its new policy.
Back in December Facebook also revealed that the ATT prompt applies broadly to all mobile advertising, including mobile web campaigns.
No mechanism may be used for aggregating data at the user level for the purposes of tracking in the case of an ATT opt-out.
- For web campaigns, ATT governs the use of tracking pixels for user level data collection
- For app campaigns, ATT governs the use of the IDFA for user level data collection
For all campaigns (app and web), other identity options like device fingerprints, email addresses, et cetera, as well as back end conversion APIs like Facebook's, can't be used as they are now if the user has not given consent.
SKAdNetwork is how attribution will be done for mobile app campaigns for users that have opted out of ATT. Private click measurement (PCM) is how web events will be attributed for opted out users.
Both SKAdNetwork and PCM aggregate advertising data at the campaign level and allow for one conversion event (or “instrumented prioritized onsite event”) to be transmitted back to the advertising network in a postback, for every acquired user.
ATT is the future, and the most radical interpretation of how severely ATT will be enforced is probably the correct one. Advertisers need to adapt to that (rather than grieving or bargaining or conceiving of workarounds).
What might be the primary motivation of ATT for Apple is to regain control of content distribution on iOS, and move away from app discovery being driven by advertising.
Apple is most likely using ATT strategically to strengthen its market power, with respect to operating the app store. Because of this, Eric expects that ATT policy will be enforced to the maximum extent.
If products can't survive or thrive through advertising, then those products become more dependent on the platform operator for distribution.
Apple gets to define tracking, and then it gets to define the privacy spectrum. With its own approach at the extreme protective end, and tracking at the extreme exploitative end. But in fact, Apple’s privacy rules are at the middle of the privacy spectrum.
Users now expect all of their content to be available across all their devices. The idea of a direct relationship between a platform and a proprietary hardware (e.g. the App Store and the iPhone) is becoming anachronistic and ATT is Apple’s attempt to make this last longer.
If ad platforms can no longer efficiently connect supply and demand via serving ads, they'll try to keep content interactions on their own properties so that they can leverage first party data and sell advertising. They’ll become content fortresses. Example: OTT apps in China and Facebook with instant games, instant articles, FB shops, etc.
If all the publishing happens directly in Facebook’s services instead of Facebook serving as a routing service between publishers and users, it increases the dependency to Facebook and other platforms. It could result in less real choice for consumers and a much less competitive content ecosystem on mobile.
The SKAdNetwork framework that Apple is providing to advertisers is inadequate, unnecessarily punitive and injures advertising beyond what’s required from a privacy standpoint.
SKAdNetwork doesn't include parameters that are necessary for optimizing ad campaigns and that wouldn't reduce privacy protections. One example is the absence of a creative ID, which wouldn’t help attributing a specific user to a campaign yet is critical to advertisers (as Apple knows since there is a creative ID in the new Apple Ads attribution - AdServices).
As of March 2021, SKAdNetwork is completely dysfunctional: only a few networks are allowing advertisers to run SKAdBNetwork verified campaigns and even for these there are problems with reporting (receiving only a portion of postbacks, incomplete parameter information like conversion values and publisher site IDs).
Connect to existing consumers with first party data in as many ways as possible, so you can retain them.
As targeting gets less precise, acquisition will become less efficient and to stay viable you need a higher LTV over a longer lifetime.
Ad platforms use to do the filtering and provide qualified users. With less qualified users, the product will need to segment users into buckets and optimize the experience of these buckets to better monetize. It will need to go beyond early stage offers and content recommendation.
It will be hard to build niche products and advertise to small audiences, and it will impact B2C products in many app verticals that saw success with low DAU, high ARPDAU.
Just because impressions become cheaper doesn't mean all advertisers are able to maintain or increase spend. The entire funnel degrades (CTR, retention rates, monetization rates) and this impacts ROI.
Products will need to be appealing to a broader audience to reduce the negative impact of losing granular targeting.
Most ROI models will need to be completely recalibrated because the composition of cohorts will change as targeting becomes much less granular and LTV projections will change.
Only D0 can be used to project out to some target LTV, as you can only transmit data back to the network within a limited timeframe
Creative testing will be slower, because there are a limited number of campaign IDs you can use and most of them will be used by ad platforms to model ad group and ad creative performance.
Advertisers thought you could just not use the IDFA (i.e not show the prompt) and instead use other identifiers. But Apple is not allowing for any loopholes (e.g. fingerprinting) that would undermine its new policy.
Back in December Facebook also revealed that the ATT prompt applies broadly to all mobile advertising, including mobile web campaigns.
No mechanism may be used for aggregating data at the user level for the purposes of tracking in the case of an ATT opt-out.
- For web campaigns, ATT governs the use of tracking pixels for user level data collection
- For app campaigns, ATT governs the use of the IDFA for user level data collection
For all campaigns (app and web), other identity options like device fingerprints, email addresses, et cetera, as well as back end conversion APIs like Facebook's, can't be used as they are now if the user has not given consent.
SKAdNetwork is how attribution will be done for mobile app campaigns for users that have opted out of ATT. Private click measurement (PCM) is how web events will be attributed for opted out users.
Both SKAdNetwork and PCM aggregate advertising data at the campaign level and allow for one conversion event (or “instrumented prioritized onsite event”) to be transmitted back to the advertising network in a postback, for every acquired user.
ATT is the future, and the most radical interpretation of how severely ATT will be enforced is probably the correct one. Advertisers need to adapt to that (rather than grieving or bargaining or conceiving of workarounds).
What might be the primary motivation of ATT for Apple is to regain control of content distribution on iOS, and move away from app discovery being driven by advertising.
Apple is most likely using ATT strategically to strengthen its market power, with respect to operating the app store. Because of this, Eric expects that ATT policy will be enforced to the maximum extent.
If products can't survive or thrive through advertising, then those products become more dependent on the platform operator for distribution.
Apple gets to define tracking, and then it gets to define the privacy spectrum. With its own approach at the extreme protective end, and tracking at the extreme exploitative end. But in fact, Apple’s privacy rules are at the middle of the privacy spectrum.
Users now expect all of their content to be available across all their devices. The idea of a direct relationship between a platform and a proprietary hardware (e.g. the App Store and the iPhone) is becoming anachronistic and ATT is Apple’s attempt to make this last longer.
If ad platforms can no longer efficiently connect supply and demand via serving ads, they'll try to keep content interactions on their own properties so that they can leverage first party data and sell advertising. They’ll become content fortresses. Example: OTT apps in China and Facebook with instant games, instant articles, FB shops, etc.
If all the publishing happens directly in Facebook’s services instead of Facebook serving as a routing service between publishers and users, it increases the dependency to Facebook and other platforms. It could result in less real choice for consumers and a much less competitive content ecosystem on mobile.
The SKAdNetwork framework that Apple is providing to advertisers is inadequate, unnecessarily punitive and injures advertising beyond what’s required from a privacy standpoint.
SKAdNetwork doesn't include parameters that are necessary for optimizing ad campaigns and that wouldn't reduce privacy protections. One example is the absence of a creative ID, which wouldn’t help attributing a specific user to a campaign yet is critical to advertisers (as Apple knows since there is a creative ID in the new Apple Ads attribution - AdServices).
As of March 2021, SKAdNetwork is completely dysfunctional: only a few networks are allowing advertisers to run SKAdBNetwork verified campaigns and even for these there are problems with reporting (receiving only a portion of postbacks, incomplete parameter information like conversion values and publisher site IDs).
Connect to existing consumers with first party data in as many ways as possible, so you can retain them.
As targeting gets less precise, acquisition will become less efficient and to stay viable you need a higher LTV over a longer lifetime.
Ad platforms use to do the filtering and provide qualified users. With less qualified users, the product will need to segment users into buckets and optimize the experience of these buckets to better monetize. It will need to go beyond early stage offers and content recommendation.
It will be hard to build niche products and advertise to small audiences, and it will impact B2C products in many app verticals that saw success with low DAU, high ARPDAU.
Just because impressions become cheaper doesn't mean all advertisers are able to maintain or increase spend. The entire funnel degrades (CTR, retention rates, monetization rates) and this impacts ROI.
Products will need to be appealing to a broader audience to reduce the negative impact of losing granular targeting.
Most ROI models will need to be completely recalibrated because the composition of cohorts will change as targeting becomes much less granular and LTV projections will change.
Only D0 can be used to project out to some target LTV, as you can only transmit data back to the network within a limited timeframe
Creative testing will be slower, because there are a limited number of campaign IDs you can use and most of them will be used by ad platforms to model ad group and ad creative performance.
Notes for this resource are currently being transferred and will be available soon.
ATT replicates LAT on an app-by-app basis.
Initially, advertisers believed that the only practical consequence of an ATT opt-out was that an app could not access the IDFA and that you could use other data for app tracking.
[💎@02:05] Advertisers thought you could just not use the IDFA (i.e not show the prompt) and instead use other identifiers. But Apple is not allowing for any loopholes (e.g. fingerprinting) that would undermine its new policy.
The scope of iOS 14 became much clearer in December when Facebook revealed that it was told by Apple that it must expose the ATT prompt in its apps or risk having them removed from the app store.
[💎@03:15] Back in December Facebook also revealed that the ATT prompt applies broadly to all mobile advertising, including mobile web campaigns.
Apple sees ATT as a policy, not an IDFA gating mechanic and adherence to ATT guidelines is expected to be total.
[💎@04:00] No mechanism may be used for aggregating data at the user level for the purposes of tracking in the case of an ATT opt-out.
[💎@04:18] For all campaigns (app and web), other identity options like device fingerprints, email addresses, et cetera, as well as back end conversion APIs like Facebook's, can't be used as they are now if the user has not given consent.
Ad networks leverage user-level tracking to get full transparency into what users do/buy on websites and in apps in order to build personal profiles and know which ads to expose users to.
[💎@05:50] SKAdNetwork is how attribution will be done for mobile app campaigns for users that have opted out of ATT. Private click measurement (PCM) is how web events will be attributed for opted out users.
[💎@06:08] Both SKAdNetwork and PCM aggregate advertising data at the campaign level and allow for one conversion event (or “instrumented prioritized onsite event”) to be transmitted back to the advertising network in a postback, for every acquired user.
What’s being lost with ATT is both measurement and targeting.
If a user has opted out of the ATT opt-in window, their IDFA will be inaccessible and any data used to attribute them to a source campaign will be transmitted at the campaign level, not the user level.
This will be done using SKAdNetwork for mobile app campaigns or some variant of PCM for mobile web campaigns (such as Facebook's AEM, which was modeled on PCM).
Facebook is only accepting data via the conversions API in accordance with its AEM structure, and it isn't matching users that have opted out for custom audiences.
Only up to one event will be attributed to a campaign per user.
There has been no good news since June on ATT.
[💎@10:03] ATT is the future, and the most radical interpretation of how severely ATT will be enforced is probably the correct one. Advertisers need to adapt to that (rather than grieving or bargaining or conceiving of workarounds).
ATT fits into the broader privacy transformation being applied to digital advertising on the browser with Apple's ITP launched in 2017 and the upcoming deprecation of third-party cookies on Chrome that catalyzed Google's privacy sandbox. They are the signs of a future of interest-based targeting and aggregated campaign level measurement.
Apple’s motivations for ATT:
[💎@12:22] What might be the primary motivation of ATT for Apple is to regain control of content distribution on iOS, and move away from app discovery being driven by advertising.
It gives Apple more leverage in battles like the one with EPIC. If Apple is seen as the primary point from which content is discovered, then it has a much better case for demanding that developers must use app store payment processing.
It's impossible to really assess how Apple will regulate and enforce ATT without first considering what it is trying to achieve with ATT.
[💎@13:45] Apple is most likely using ATT strategically to strengthen its market power, with respect to operating the app store. Because of this, Eric expects that ATT policy will be enforced to the maximum extent.
Ads personalization is a public good. But the IDFA has been abused and does need to be retired. Cf. more in “The IDFA is the hydrocarbon of the mobile advertising ecosystem”.
But this is not what ATT is designed to do, and ATT is problematic.
Apple doesn’t provide real consumer choice and implies that the difference between being tracked and not tracked is the difference between the data being collected by developers or not.
Freemium economy is based on in-product personalization and ad personalization. If that’s not possible the monetization model needs to change or risk not being viable anymore.
[💎@16:20] If products can't survive or thrive through advertising, then those products become more dependent on the platform operator for distribution.
Apple doesn’t track users with its own ad network, but that’s based on Apple’s definition of tracking.
What Apple does is group users into interest-based segments based on usage of its first party products. What recourse do users have?
[💎@17:17] Apple gets to define tracking, and then it gets to define the privacy spectrum. With its own approach at the extreme protective end, and tracking at the extreme exploitative end. But in fact, Apple’s privacy rules are at the middle of the privacy spectrum.
Privacy, as it relates to digital products, is a tradeoff between data collection and usage and product functionality. Apple’s position is that the user data collected belongs to the developer of the product but may not be exchanged with a 3rd party. That’s ok, but it is the middle point of the spectrum and does not represent total privacy.
Very efficient and very targeted advertising should, in a sense, remove any need for a user to ever search for content. And that would make the App Store a middleman.
The context here is that the device on which you consume digital content is becoming irrelevant: users expect all their content to be available across all their devices.
[💎@20:01] Users now expect all of their content to be available across all their devices. The idea of a direct relationship between a platform and a proprietary hardware (e.g. the App Store and the iPhone) is becoming anachronistic and ATT is Apple’s attempt to make this last longer.
Apple wants to tie content distribution and hardware. If Apple loses this, then they need to compete with developers for content. And so Apple has been trying to preserve this by measures like blocking “app stores within the app stores”, preventing game streaming services from publishing to the app store and forcing the use of its own payment gateways.
Cf. article “The profound, unintended consequence of ATT: content fortresses”.
[💎@21:15] If ad platforms can no longer efficiently connect supply and demand via serving ads, they'll try to keep content interactions on their own properties so that they can leverage first party data and sell advertising. They’ll become content fortresses. Example: OTT apps in China and Facebook with instant games, instant articles, FB shops, etc.
Advertising is just leasing eyeballs out. If that is not cost-effective then platforms will ask publishers to move interactions onto the platforms’ own properties, making it possible for them to sell advertising there.
[💎@21:56] If all the publishing happens directly in Facebook’s services instead of Facebook serving as a routing service between publishers and users, it increases the dependency to Facebook and other platforms. It could result in less real choice for consumers and a much less competitive content ecosystem on mobile.
[💎@22:27] The SKAdNetwork framework that Apple is providing to advertisers is inadequate, unnecessarily punitive and injures advertising beyond what’s required from a privacy standpoint.
Google’s privacy sandbox is thoughtful, and being rolled out slowly. But SKAdNetwork interested no one when it came out in 2018.
[💎@23:04] SKAdNetwork doesn't include parameters that are necessary for optimizing ad campaigns and that wouldn't reduce privacy protections. One example is the absence of a creative ID, which wouldn’t help attributing a specific user to a campaign yet is critical to advertisers (as Apple knows since there is a creative ID in the new Apple Ads attribution - AdServices).
[💎@23:44] As of March 2021, SKAdNetwork is completely dysfunctional: only a few networks are allowing advertisers to run SKAdBNetwork verified campaigns and even for these there are problems with reporting (receiving only a portion of postbacks, incomplete parameter information like conversion values and publisher site IDs)
SKAdNetwork is broken and advertisers can not rebuild their ad serving and ad targeting infrastructure.
[💎@25:00] Connect to existing consumers with first party data in as many ways as possible, so you can retain them.
Although you can’t use email addresses for targeting, there’s nothing stopping an advertiser to use them to communicate with their users.
Acquisition won’t be something an advertiser can rely on to yield a 90 day or a 100 day LTV.
[💎@25:50] As targeting gets less precise, acquisition will become less efficient and to stay viable you need a higher LTV over a longer lifetime.
Personalization increases the degree to which users monetize.
[💎@26:40] Ad platforms use to do the filtering and provide qualified users. With less qualified users, the product will need to segment users into buckets and optimize the experience of these buckets to better monetize. It will need to go beyond early stage offers and content recommendation.
As traffic becomes less targeted, many products should prepare for a dramatic collapse of product metrics like early stage retention, ARPU, ARPDAU, RRPU, etc.
You want to present to users the content they are the most likely to engage with. This will become a significant competitive advantage.
Targeting broadly gives an ad platform as much room as possible to find the right fit between users, products and the ad creative. This relies on the feedback loop between ad impressions, clicks, engagement, and monetization.
If the feedback loop between ad impressions, clicks, engagement, and monetization is broken, ad platforms will no longer be able to build proclivity profiles of users that can be used for granular user level targeting.
[💎@29:01] It will be hard to build niche products and advertise to small audiences, and it will impact B2C products in many app verticals that saw success with low DAU, high ARPDAU.
[💎@29:40] Just because impressions become cheaper doesn't mean all advertisers are able to maintain or increase spend. The entire funnel degrades (CTR, retention rates, monetization rates) and this impacts ROI.
[💎@30:00] Products will need to be appealing to a broader audience to reduce the negative impact of losing granular targeting.
Little is known about how the platforms are going to accommodate ATT, and it’s impossible to give guidance now.
Big changes to think through:
Change #1
[💎@30:29] Most ROI models will need to be completely recalibrated because the composition of cohorts will change as targeting becomes much less granular and LTV projections will change.
Change #2
[💎@31:00] Only D0 can be used to project out to some target LTV, as you can only transmit data back to the network within a limited timeframe
Change #3
[💎@31:30] Creative testing will be slower, because there are a limited number of campaign IDs you can use and most of them will be used by ad platforms to model ad group and ad creative performance.
You will most likely have to lower the number of creatives and variants you produce, if you want to get a sense of performance.
Change #4
You will have to be very good at assessing and reporting monetization based on early stage product behavior. It can be defining a “value path” that creates a combined event so you can provide context.
ATT replicates LAT on an app-by-app basis.
Initially, advertisers believed that the only practical consequence of an ATT opt-out was that an app could not access the IDFA and that you could use other data for app tracking.
[💎@02:05] Advertisers thought you could just not use the IDFA (i.e not show the prompt) and instead use other identifiers. But Apple is not allowing for any loopholes (e.g. fingerprinting) that would undermine its new policy.
The scope of iOS 14 became much clearer in December when Facebook revealed that it was told by Apple that it must expose the ATT prompt in its apps or risk having them removed from the app store.
[💎@03:15] Back in December Facebook also revealed that the ATT prompt applies broadly to all mobile advertising, including mobile web campaigns.
Apple sees ATT as a policy, not an IDFA gating mechanic and adherence to ATT guidelines is expected to be total.
[💎@04:00] No mechanism may be used for aggregating data at the user level for the purposes of tracking in the case of an ATT opt-out.
[💎@04:18] For all campaigns (app and web), other identity options like device fingerprints, email addresses, et cetera, as well as back end conversion APIs like Facebook's, can't be used as they are now if the user has not given consent.
Ad networks leverage user-level tracking to get full transparency into what users do/buy on websites and in apps in order to build personal profiles and know which ads to expose users to.
[💎@05:50] SKAdNetwork is how attribution will be done for mobile app campaigns for users that have opted out of ATT. Private click measurement (PCM) is how web events will be attributed for opted out users.
[💎@06:08] Both SKAdNetwork and PCM aggregate advertising data at the campaign level and allow for one conversion event (or “instrumented prioritized onsite event”) to be transmitted back to the advertising network in a postback, for every acquired user.
What’s being lost with ATT is both measurement and targeting.
If a user has opted out of the ATT opt-in window, their IDFA will be inaccessible and any data used to attribute them to a source campaign will be transmitted at the campaign level, not the user level.
This will be done using SKAdNetwork for mobile app campaigns or some variant of PCM for mobile web campaigns (such as Facebook's AEM, which was modeled on PCM).
Facebook is only accepting data via the conversions API in accordance with its AEM structure, and it isn't matching users that have opted out for custom audiences.
Only up to one event will be attributed to a campaign per user.
There has been no good news since June on ATT.
[💎@10:03] ATT is the future, and the most radical interpretation of how severely ATT will be enforced is probably the correct one. Advertisers need to adapt to that (rather than grieving or bargaining or conceiving of workarounds).
ATT fits into the broader privacy transformation being applied to digital advertising on the browser with Apple's ITP launched in 2017 and the upcoming deprecation of third-party cookies on Chrome that catalyzed Google's privacy sandbox. They are the signs of a future of interest-based targeting and aggregated campaign level measurement.
Apple’s motivations for ATT:
[💎@12:22] What might be the primary motivation of ATT for Apple is to regain control of content distribution on iOS, and move away from app discovery being driven by advertising.
It gives Apple more leverage in battles like the one with EPIC. If Apple is seen as the primary point from which content is discovered, then it has a much better case for demanding that developers must use app store payment processing.
It's impossible to really assess how Apple will regulate and enforce ATT without first considering what it is trying to achieve with ATT.
[💎@13:45] Apple is most likely using ATT strategically to strengthen its market power, with respect to operating the app store. Because of this, Eric expects that ATT policy will be enforced to the maximum extent.
Ads personalization is a public good. But the IDFA has been abused and does need to be retired. Cf. more in “The IDFA is the hydrocarbon of the mobile advertising ecosystem”.
But this is not what ATT is designed to do, and ATT is problematic.
Apple doesn’t provide real consumer choice and implies that the difference between being tracked and not tracked is the difference between the data being collected by developers or not.
Freemium economy is based on in-product personalization and ad personalization. If that’s not possible the monetization model needs to change or risk not being viable anymore.
[💎@16:20] If products can't survive or thrive through advertising, then those products become more dependent on the platform operator for distribution.
Apple doesn’t track users with its own ad network, but that’s based on Apple’s definition of tracking.
What Apple does is group users into interest-based segments based on usage of its first party products. What recourse do users have?
[💎@17:17] Apple gets to define tracking, and then it gets to define the privacy spectrum. With its own approach at the extreme protective end, and tracking at the extreme exploitative end. But in fact, Apple’s privacy rules are at the middle of the privacy spectrum.
Privacy, as it relates to digital products, is a tradeoff between data collection and usage and product functionality. Apple’s position is that the user data collected belongs to the developer of the product but may not be exchanged with a 3rd party. That’s ok, but it is the middle point of the spectrum and does not represent total privacy.
Very efficient and very targeted advertising should, in a sense, remove any need for a user to ever search for content. And that would make the App Store a middleman.
The context here is that the device on which you consume digital content is becoming irrelevant: users expect all their content to be available across all their devices.
[💎@20:01] Users now expect all of their content to be available across all their devices. The idea of a direct relationship between a platform and a proprietary hardware (e.g. the App Store and the iPhone) is becoming anachronistic and ATT is Apple’s attempt to make this last longer.
Apple wants to tie content distribution and hardware. If Apple loses this, then they need to compete with developers for content. And so Apple has been trying to preserve this by measures like blocking “app stores within the app stores”, preventing game streaming services from publishing to the app store and forcing the use of its own payment gateways.
Cf. article “The profound, unintended consequence of ATT: content fortresses”.
[💎@21:15] If ad platforms can no longer efficiently connect supply and demand via serving ads, they'll try to keep content interactions on their own properties so that they can leverage first party data and sell advertising. They’ll become content fortresses. Example: OTT apps in China and Facebook with instant games, instant articles, FB shops, etc.
Advertising is just leasing eyeballs out. If that is not cost-effective then platforms will ask publishers to move interactions onto the platforms’ own properties, making it possible for them to sell advertising there.
[💎@21:56] If all the publishing happens directly in Facebook’s services instead of Facebook serving as a routing service between publishers and users, it increases the dependency to Facebook and other platforms. It could result in less real choice for consumers and a much less competitive content ecosystem on mobile.
[💎@22:27] The SKAdNetwork framework that Apple is providing to advertisers is inadequate, unnecessarily punitive and injures advertising beyond what’s required from a privacy standpoint.
Google’s privacy sandbox is thoughtful, and being rolled out slowly. But SKAdNetwork interested no one when it came out in 2018.
[💎@23:04] SKAdNetwork doesn't include parameters that are necessary for optimizing ad campaigns and that wouldn't reduce privacy protections. One example is the absence of a creative ID, which wouldn’t help attributing a specific user to a campaign yet is critical to advertisers (as Apple knows since there is a creative ID in the new Apple Ads attribution - AdServices).
[💎@23:44] As of March 2021, SKAdNetwork is completely dysfunctional: only a few networks are allowing advertisers to run SKAdBNetwork verified campaigns and even for these there are problems with reporting (receiving only a portion of postbacks, incomplete parameter information like conversion values and publisher site IDs)
SKAdNetwork is broken and advertisers can not rebuild their ad serving and ad targeting infrastructure.
[💎@25:00] Connect to existing consumers with first party data in as many ways as possible, so you can retain them.
Although you can’t use email addresses for targeting, there’s nothing stopping an advertiser to use them to communicate with their users.
Acquisition won’t be something an advertiser can rely on to yield a 90 day or a 100 day LTV.
[💎@25:50] As targeting gets less precise, acquisition will become less efficient and to stay viable you need a higher LTV over a longer lifetime.
Personalization increases the degree to which users monetize.
[💎@26:40] Ad platforms use to do the filtering and provide qualified users. With less qualified users, the product will need to segment users into buckets and optimize the experience of these buckets to better monetize. It will need to go beyond early stage offers and content recommendation.
As traffic becomes less targeted, many products should prepare for a dramatic collapse of product metrics like early stage retention, ARPU, ARPDAU, RRPU, etc.
You want to present to users the content they are the most likely to engage with. This will become a significant competitive advantage.
Targeting broadly gives an ad platform as much room as possible to find the right fit between users, products and the ad creative. This relies on the feedback loop between ad impressions, clicks, engagement, and monetization.
If the feedback loop between ad impressions, clicks, engagement, and monetization is broken, ad platforms will no longer be able to build proclivity profiles of users that can be used for granular user level targeting.
[💎@29:01] It will be hard to build niche products and advertise to small audiences, and it will impact B2C products in many app verticals that saw success with low DAU, high ARPDAU.
[💎@29:40] Just because impressions become cheaper doesn't mean all advertisers are able to maintain or increase spend. The entire funnel degrades (CTR, retention rates, monetization rates) and this impacts ROI.
[💎@30:00] Products will need to be appealing to a broader audience to reduce the negative impact of losing granular targeting.
Little is known about how the platforms are going to accommodate ATT, and it’s impossible to give guidance now.
Big changes to think through:
Change #1
[💎@30:29] Most ROI models will need to be completely recalibrated because the composition of cohorts will change as targeting becomes much less granular and LTV projections will change.
Change #2
[💎@31:00] Only D0 can be used to project out to some target LTV, as you can only transmit data back to the network within a limited timeframe
Change #3
[💎@31:30] Creative testing will be slower, because there are a limited number of campaign IDs you can use and most of them will be used by ad platforms to model ad group and ad creative performance.
You will most likely have to lower the number of creatives and variants you produce, if you want to get a sense of performance.
Change #4
You will have to be very good at assessing and reporting monetization based on early stage product behavior. It can be defining a “value path” that creates a combined event so you can provide context.
ATT replicates LAT on an app-by-app basis.
Initially, advertisers believed that the only practical consequence of an ATT opt-out was that an app could not access the IDFA and that you could use other data for app tracking.
[💎@02:05] Advertisers thought you could just not use the IDFA (i.e not show the prompt) and instead use other identifiers. But Apple is not allowing for any loopholes (e.g. fingerprinting) that would undermine its new policy.
The scope of iOS 14 became much clearer in December when Facebook revealed that it was told by Apple that it must expose the ATT prompt in its apps or risk having them removed from the app store.
[💎@03:15] Back in December Facebook also revealed that the ATT prompt applies broadly to all mobile advertising, including mobile web campaigns.
Apple sees ATT as a policy, not an IDFA gating mechanic and adherence to ATT guidelines is expected to be total.
[💎@04:00] No mechanism may be used for aggregating data at the user level for the purposes of tracking in the case of an ATT opt-out.
[💎@04:18] For all campaigns (app and web), other identity options like device fingerprints, email addresses, et cetera, as well as back end conversion APIs like Facebook's, can't be used as they are now if the user has not given consent.
Ad networks leverage user-level tracking to get full transparency into what users do/buy on websites and in apps in order to build personal profiles and know which ads to expose users to.
[💎@05:50] SKAdNetwork is how attribution will be done for mobile app campaigns for users that have opted out of ATT. Private click measurement (PCM) is how web events will be attributed for opted out users.
[💎@06:08] Both SKAdNetwork and PCM aggregate advertising data at the campaign level and allow for one conversion event (or “instrumented prioritized onsite event”) to be transmitted back to the advertising network in a postback, for every acquired user.
What’s being lost with ATT is both measurement and targeting.
If a user has opted out of the ATT opt-in window, their IDFA will be inaccessible and any data used to attribute them to a source campaign will be transmitted at the campaign level, not the user level.
This will be done using SKAdNetwork for mobile app campaigns or some variant of PCM for mobile web campaigns (such as Facebook's AEM, which was modeled on PCM).
Facebook is only accepting data via the conversions API in accordance with its AEM structure, and it isn't matching users that have opted out for custom audiences.
Only up to one event will be attributed to a campaign per user.
There has been no good news since June on ATT.
[💎@10:03] ATT is the future, and the most radical interpretation of how severely ATT will be enforced is probably the correct one. Advertisers need to adapt to that (rather than grieving or bargaining or conceiving of workarounds).
ATT fits into the broader privacy transformation being applied to digital advertising on the browser with Apple's ITP launched in 2017 and the upcoming deprecation of third-party cookies on Chrome that catalyzed Google's privacy sandbox. They are the signs of a future of interest-based targeting and aggregated campaign level measurement.
Apple’s motivations for ATT:
[💎@12:22] What might be the primary motivation of ATT for Apple is to regain control of content distribution on iOS, and move away from app discovery being driven by advertising.
It gives Apple more leverage in battles like the one with EPIC. If Apple is seen as the primary point from which content is discovered, then it has a much better case for demanding that developers must use app store payment processing.
It's impossible to really assess how Apple will regulate and enforce ATT without first considering what it is trying to achieve with ATT.
[💎@13:45] Apple is most likely using ATT strategically to strengthen its market power, with respect to operating the app store. Because of this, Eric expects that ATT policy will be enforced to the maximum extent.
Ads personalization is a public good. But the IDFA has been abused and does need to be retired. Cf. more in “The IDFA is the hydrocarbon of the mobile advertising ecosystem”.
But this is not what ATT is designed to do, and ATT is problematic.
Apple doesn’t provide real consumer choice and implies that the difference between being tracked and not tracked is the difference between the data being collected by developers or not.
Freemium economy is based on in-product personalization and ad personalization. If that’s not possible the monetization model needs to change or risk not being viable anymore.
[💎@16:20] If products can't survive or thrive through advertising, then those products become more dependent on the platform operator for distribution.
Apple doesn’t track users with its own ad network, but that’s based on Apple’s definition of tracking.
What Apple does is group users into interest-based segments based on usage of its first party products. What recourse do users have?
[💎@17:17] Apple gets to define tracking, and then it gets to define the privacy spectrum. With its own approach at the extreme protective end, and tracking at the extreme exploitative end. But in fact, Apple’s privacy rules are at the middle of the privacy spectrum.
Privacy, as it relates to digital products, is a tradeoff between data collection and usage and product functionality. Apple’s position is that the user data collected belongs to the developer of the product but may not be exchanged with a 3rd party. That’s ok, but it is the middle point of the spectrum and does not represent total privacy.
Very efficient and very targeted advertising should, in a sense, remove any need for a user to ever search for content. And that would make the App Store a middleman.
The context here is that the device on which you consume digital content is becoming irrelevant: users expect all their content to be available across all their devices.
[💎@20:01] Users now expect all of their content to be available across all their devices. The idea of a direct relationship between a platform and a proprietary hardware (e.g. the App Store and the iPhone) is becoming anachronistic and ATT is Apple’s attempt to make this last longer.
Apple wants to tie content distribution and hardware. If Apple loses this, then they need to compete with developers for content. And so Apple has been trying to preserve this by measures like blocking “app stores within the app stores”, preventing game streaming services from publishing to the app store and forcing the use of its own payment gateways.
Cf. article “The profound, unintended consequence of ATT: content fortresses”.
[💎@21:15] If ad platforms can no longer efficiently connect supply and demand via serving ads, they'll try to keep content interactions on their own properties so that they can leverage first party data and sell advertising. They’ll become content fortresses. Example: OTT apps in China and Facebook with instant games, instant articles, FB shops, etc.
Advertising is just leasing eyeballs out. If that is not cost-effective then platforms will ask publishers to move interactions onto the platforms’ own properties, making it possible for them to sell advertising there.
[💎@21:56] If all the publishing happens directly in Facebook’s services instead of Facebook serving as a routing service between publishers and users, it increases the dependency to Facebook and other platforms. It could result in less real choice for consumers and a much less competitive content ecosystem on mobile.
[💎@22:27] The SKAdNetwork framework that Apple is providing to advertisers is inadequate, unnecessarily punitive and injures advertising beyond what’s required from a privacy standpoint.
Google’s privacy sandbox is thoughtful, and being rolled out slowly. But SKAdNetwork interested no one when it came out in 2018.
[💎@23:04] SKAdNetwork doesn't include parameters that are necessary for optimizing ad campaigns and that wouldn't reduce privacy protections. One example is the absence of a creative ID, which wouldn’t help attributing a specific user to a campaign yet is critical to advertisers (as Apple knows since there is a creative ID in the new Apple Ads attribution - AdServices).
[💎@23:44] As of March 2021, SKAdNetwork is completely dysfunctional: only a few networks are allowing advertisers to run SKAdBNetwork verified campaigns and even for these there are problems with reporting (receiving only a portion of postbacks, incomplete parameter information like conversion values and publisher site IDs)
SKAdNetwork is broken and advertisers can not rebuild their ad serving and ad targeting infrastructure.
[💎@25:00] Connect to existing consumers with first party data in as many ways as possible, so you can retain them.
Although you can’t use email addresses for targeting, there’s nothing stopping an advertiser to use them to communicate with their users.
Acquisition won’t be something an advertiser can rely on to yield a 90 day or a 100 day LTV.
[💎@25:50] As targeting gets less precise, acquisition will become less efficient and to stay viable you need a higher LTV over a longer lifetime.
Personalization increases the degree to which users monetize.
[💎@26:40] Ad platforms use to do the filtering and provide qualified users. With less qualified users, the product will need to segment users into buckets and optimize the experience of these buckets to better monetize. It will need to go beyond early stage offers and content recommendation.
As traffic becomes less targeted, many products should prepare for a dramatic collapse of product metrics like early stage retention, ARPU, ARPDAU, RRPU, etc.
You want to present to users the content they are the most likely to engage with. This will become a significant competitive advantage.
Targeting broadly gives an ad platform as much room as possible to find the right fit between users, products and the ad creative. This relies on the feedback loop between ad impressions, clicks, engagement, and monetization.
If the feedback loop between ad impressions, clicks, engagement, and monetization is broken, ad platforms will no longer be able to build proclivity profiles of users that can be used for granular user level targeting.
[💎@29:01] It will be hard to build niche products and advertise to small audiences, and it will impact B2C products in many app verticals that saw success with low DAU, high ARPDAU.
[💎@29:40] Just because impressions become cheaper doesn't mean all advertisers are able to maintain or increase spend. The entire funnel degrades (CTR, retention rates, monetization rates) and this impacts ROI.
[💎@30:00] Products will need to be appealing to a broader audience to reduce the negative impact of losing granular targeting.
Little is known about how the platforms are going to accommodate ATT, and it’s impossible to give guidance now.
Big changes to think through:
Change #1
[💎@30:29] Most ROI models will need to be completely recalibrated because the composition of cohorts will change as targeting becomes much less granular and LTV projections will change.
Change #2
[💎@31:00] Only D0 can be used to project out to some target LTV, as you can only transmit data back to the network within a limited timeframe
Change #3
[💎@31:30] Creative testing will be slower, because there are a limited number of campaign IDs you can use and most of them will be used by ad platforms to model ad group and ad creative performance.
You will most likely have to lower the number of creatives and variants you produce, if you want to get a sense of performance.
Change #4
You will have to be very good at assessing and reporting monetization based on early stage product behavior. It can be defining a “value path” that creates a combined event so you can provide context.
