Part 1: Major MMPs Discuss IDFA Deprecation, the Future of Fraud Detection on iOS, and  SKAdNetwork Conversion Values

IDFA_Blog-Header-03 (1)

While the mobile adtech industry has been expecting greater user privacy controls for some time, the announcement of Apple’s privacy changes shocked the mobile advertising world with its quick and decisive update back at WWDC in June 2020. Since then, the mobile advertising industry has been making preparations, but there are still open questions about the impact on certain areas of the ecosystem.  

As a brief background, one of the biggest privacy updates Apple will implement as part of iOS14.5 is the App Tracking Transparency (ATT) framework. ATT framework, which requires the user’s active consent before their IDFA can be accessed in each app for cross app tracking, fundamentally changes how mobile marketing performance is measured, fraud is detected, and advertising transacted. 

To get a full picture of the wider impact of this change, we will be presenting a multi-part series on the MMP, Exchange, and DSP perspective on how Apple’s updates will impact our industry.

To kick off the series, we interviewed thought-leaders from all major MMPs, including Adjust, AppsFlyer, Branch, Kochava, and Singular, to learn their insights about Apple’s privacy updates and how SKAdNetwork will change mobile measurement. We also tackled if Apple’s new policies and attribution solution open the door for fraudsters taking advantage of the iOS14.5 changes. 

Headshot of Katie from Adjust, Ben from AppsFlyer and Alex from Branch

Headshots of Garrett from Kochava and Eran from Singular

Question 1: Will fraudsters take advantage of Apple’s iOS14.5 privacy changes? How will this be prevented?

Adjust: While the new privacy changes brought about by iOS14.5 can open up opportunities for fraudsters, Adjust Fraud Prevention Suite remains as efficient fighting back fraud as ever. Because it is not reliant on the IDFA, our combination of Distribution Modeling, Anonymous IP filtering, and SDK Signature ensure that the traffic we receive is clean and correctly attributed.

When it comes to SKAdNetwork, fraud prevention is different. As the SKAdNetwork payload is signed and Adjust verifies it, it prevents any kind of SKAdNetwork install spoofing. However, there is still one angle of attack for fraudsters: conversion values. Conversion values are not part of the SKAdNetwork verification signature, which means that fraudulent parties could change the value, potentially to increase their perceived performance. Whether these potential exploits escalate into actual fraud schemes remains to be seen, as identification through Apple is going to result in exclusion from the iOS platform’s ad-ecosystem. 

Overall, marketers relying on Adjust to verify their traffic can rest assured that we are still leading the charge against ad fraud for all traffic, iOS14.5 included.

- Katie Madding, Chief Product Officer, Adjust

AppsFlyer: Fraud is as certain a reality as death and taxes. Fraudsters will always take advantage of changing realities to find new loopholes for propagating fraud. Even in unchanging realities, they’re constantly getting more sophisticated and more innovative with their methods.

We’ve already found several fraud loopholes in Apple’s SKAdNetwork, as the protocol that the methodology relies on leaves plenty of opportunities for fraudsters to manipulate postbacks in transit. SKAdNetwork is also at risk for click flooding and device farming fraud, which could potentially cost advertisers millions in ad spend per year. 

The changes introduced with iOS 14 mean fraud could be harder to detect, especially when it comes to aggregate data reporting methods such as SKAdNetwork. It’s practically invisible to the naked eye, which emphasizes the importance of a reliable measurement partner with access to extensive data and an unbiased position towards fraud protection. As difficult as it may be to detect fraud in this new reality, behavior anomaly trends are the key to identifying attribution fraud attempts - and these can only be done by reviewing vast amounts of attribution data.

- Ben Roodman, AVP, Business Development, AppsFlyer

Branch: Fraud detection and prevention is always cat-and-mouse game: something happens to upset the balance (for example: a new advancement in fraud detection capability or a major change to underlying tech standards, such as SKAdNetwork in iOS 14), fraudsters search for a profitable way to exploit the new situation, and then anti-fraud solutions triangulate against the new threat and move to contain it.

In the case of iOS 14.5, SKAdNetwork is a fundamental adjustment to how the mobile measurement stack works. We expect that it will make 'easy fraud' (things like click spamming) harder, simply because there are fewer places in the campaign flow where a fraudster can tap into the process. However, more sophisticated fraud techniques will likely still make it through (for example, device farms), and will be much harder to catch because SKAdNetwork reports those conversions with Apple's blessing as legitimate installs, and provides none of the additional data necessary to perform an independent analysis. 

It's great that Apple is already taking steps to help prevent obvious technical fraud vectors (for example, cryptographic signing of SKAdNetwork postbacks to prevent direct manipulation of the data, or the decision to enforce a minimum exposure time of 3 seconds for view-through impressions). However, those are really just table stakes at this point — hopefully Apple will move quickly to enhance the SKAdNetwork system with protection against the more sophisticated fraud vectors that have cropped up in recent years.

- Alex Bauer, Market Strategy Director, Branch

Kochava: Where there is a will, there’s a way–so yes, fraudsters will undoubtedly exploit the impending iOS changes to their advantage and try to continue stealing from marketers and publishers alike. There’s a long list of potential issues that fraudsters could capitalize on, but a few key ones are (1) SKAdNetwork’s lengthy attribution lookback windows that are at the moment unconfigurable, (2) ad networks being the sole recipient and disseminator of the postback data from Apple, rather than an independent third-party, (3) authentication loopholes that could allow for falsification of impression, click, and/or install data and the resulting attributions between them. 

Kochava, through its SKAdNetwork Partner Certification Program is working to achieve transparency with all partners running on SKAdNetwork. The goal is to create relationships of trust between advertisers and their partners as we all venture out into this new horizon. Partner certification requires partners to forward install notification postbacks directly to Kochava, with our systems performing independent postback verification with Apple. In addition, partners must send all impressions and clicks for SKAdNetwork-based campaigns. This enables Kochava to calculate final click-to-install (CTI) ratios, click-through-rate (CTR) and other key metrics that can provide directional indications of click spamming, fake impressions, and install reporting anomalies. Such insights will help marketers optimize away from fraudulent sources.

- Garrett C. MacDonald, EVP Strategic Sales, Kochava

Singular: With iOS 14.5, all aspects of user acquisition are changing; UA strategy, optimization techniques, measurement, and of course, ad fraud. While advertisers work to reinvent their UA strategy, fraudsters are working hard to reinvent fraud methods. 

We already identified different weaknesses with the new mechanisms, which require protection:

  • SKAdNetwork postbacks can be duplicated or re-used to claim multiple installs if their transactions aren't validated
  • SKAdNetwork conversion values can be altered since Apple does not sign all fields
We offer unique solutions, including Secure-SKAN postback validation and fake installs and bot protection, and are continuing to innovate as the industry evolves. Since data is still limited, our team focused on protocol research and mapping the landscape to ensure our customers will be safe.

- Eran Friedman, Co-Founder & CTO, Singular

Question 2: Can you talk about Conversion Value solutions?

Adjust: Conversion values are a key part of SKAdNetwork as they allow marketers to gather some information about the user behavior within their app.

At Adjust, our clients can enjoy full flexibility in setting up and managing their conversion values. We offer three separate approaches:

  • Event-based conversions - Our event-based solution allows marketers to link 6 distinct events to specific conversion values. It empowers them to have full visibility into their main post-install KPIs and ensures that they know as much of the user behavior as possible within their SKAdNetwork data set.

    This solution is especially suited to marketers who really want to use a wide array of events to define the importance of each user.

  • Value-based conversions - In our value-based conversions, Adjust seamlessly aggregates the user behavior across 4 different activities: session, events, revenue, ad revenue and gives marketers full command of their conversion value management by leveraging all 63 distinct buckets.

    This solution is especially well suited to marketers who want to focus on a specific KPI or, leveraging the AND logic, want to build very distinct in-app flows for maximum accuracy.

  • Server-to-server conversion value - Both solutions are not only supported by the Adjust SDK out of the box, but can also be utilized server-to-server for complete flexibility.

Conversion window and conversion URL 
The conversion value window allows marketers to set a defined time period after an app install, or reinstall, in which Adjust continues to update conversion values based on user activity. For example, setting this window to 24 hours, will provide a clean dataset as all users will have been measured within the same period.

- Katie Madding, Chief Product Officer, Adjust

AppsFlyer: The conversion value system in itself is very complicated, both to understand and to implement. Advertisers are facing three main challenges when it comes to the SKAdNetwork conversion value:

- Translating their KPIs and business needs to a conversion value map
- Tweaking and optimizing the conversion value map
- Sharing their map collectively with partners, and updating all partners when changes are made

Each one of these issues requires a lot of unpacking, but let’s take a brief look at the first point and the complexities involved.

Apple provides app developers with 6 bits to measure conversion values. There are a few ways advertisers can split the bits based on their KPIs:

  • Flat: Allocate all the bits to measure a single KPI (such as revenue). In this example, each possible value (0-63) would represent a monetary value generated by the user.

  • Split: Split the bits 2 ways (or more). Example: allocate 3 bits for measuring revenue, and the other 3 bits for measuring game progress (level 3 achieved, tutorial completed etc).

  • Combo-split: A combination of the above adding a 1-bit deterministic on/off signal for a conversion event (Example: male? yes/no)

The conversion value data runs on a timer, measuring user activity within a set time frame. This adds another level of complexity to the optimization of the conversion value map, as data is reported with a delay of at least 24 hours. Advertisers are tasked with solving this puzzle, which requires in-depth understanding of the inner workings of SKAdNetwork. We’re seeing a lot of advertisers struggling with this. We developed SK360 - our integrated suite for SKAdNetwork - to help them maximize the opportunity from SKAdNetwork. 

- Ben Roodman, AVP, Business Development, AppsFlyer

Branch: The conversion-value parameter is the only signal available for post-install reporting with SKAdNetwork, which makes it an extremely precious commodity. At the same time, Apple has placed such extreme limitations on this parameter that it simply may not be meaningful data for many marketers. In other words, gone are the days when user LTV could be reliably connected back to marketing efforts — we're now living in a world where campaign ROI is based on predictive assumptions and fuzzy estimations.

At the moment, all MMPs are providing (or soon plan to provide) similar out-of-the-box functionality. There's some fancy marketing spin going around, but ultimately it's a bit like lipstick on a pig: you can only do so much with 6 bits of data that are available sometimes. At the moment, the most common SKAdNetwork conversion-value strategies involve mapping existing in-app events (9 = user completed SIGNUP), bucketing revenue (9 = user spent $10-$15), and simple aggregation of event counts (9 = user performed VIEW_PRODUCT more than 5 times). These are also the strategies being recommended by major walled garden networks like Facebook and Snap.

More advanced advertisers are building their own in-house strategies based on econometric models, to varying degrees of success (Apple still seems to be tweaking the privacy thresholds for when the conversion-value parameter is available, which makes it very challenging to test). An example of this strategy would be looking at a user's in-app behavior within 24 hours of install, comparing that to down-funnel performance using a model trained on activity of users who took similar actions in the past, and then dynamically triggering conversion-value updates (for example, 9 = 'we predict this user is highly likely to purchase in future').

- Alex Bauer, Market Strategy Director, Branch

Kochava: The conversion value is Apple’s gift to marketers with the rollout of SKAdNetwork v2. It brought post-install performance and user quality insights to the SKAdNetwork for the first time, something that was sorely missing in v1. Kochava has developed the most innovative approach to maximize the value and insight that can be packed into this range of numbers between 0-63. Using a 6-bit protocol system, we created flexible conversion models that cover more use cases, more key performance indicators (KPIs), and more types of monetization strategies than any other measurement solution. We also pack in a configurable post-install measurement window that advertisers can scale between the day of install out to as much as the day of install + 7 days. Our data science and client analytics teams are helping our customers analyze their first-party app data to understand which conversion model and configuration gives them the most performance intel to work with in the most reasonable measurement window possible to be effective with their optimization. Our four models are Highest Value Event Completed, Revenue, User Journey, and Engagement, and you can learn more about them through our free guide here

It’s important to note that marketers can change their conversion model server-side from the dashboard at any time without having to set aside additional development resources for a complete app update This means more flexibility to support shifting KPIs because of seasonality or other factors.

- Garrett C. MacDonald, EVP Strategic Sales, Kochava

Singular: Conversion management is one of the most challenging components of SKAdNetwork. You only have 64 numeric values to represent post-install conversions, and you only get one chance to trigger a post-install conversion postback, without a guarantee when it will happen.

Experimentation with different conversion points is essential. You might want to play with different conversion models, such as revenue, retention, engagement, or predictive LTV. And you definitely don’t want to have to update your app code every time you need to test or optimize.

That’s where Singular’s Smart Conversion Management helps:

  • Singular will implement all the SKAdNetwork functionality
  • You can change conversion models in real-time without touching code or resubmitting your app
  • Singular will automatically postback your conversion events to your ad partners
We support multiple different models:
  1. Conversion events: Conversion events indicate if specific, meaningful events have occurred (things like sign-ups, purchases, etc.)
  2. Engagement: Measures how many occurrences a specific event had, and you can count this in different buckets for different events.
  3. Revenue: Measures how much revenue your users generated, based on both in-app purchases and ad monetization.
  4. Funnel: Lets you optimize campaigns based on a specific post-install user journey.
And it gets better. You can take your historical data and model it against multiple conversion models to simulate what might work best in iOS 14, giving you important insights into how to run user acquisition in iOS 14 right from the very beginning.

- Eran Friedman, Co-Founder & CTO, Singular

Question 3: How IDFA deprecation will impact marketers’ buying activities on mobile? What are your predictions? 

Adjust: While a certain portion of spend may shift to web and android, we believe our Conversion Modeling solution will allow for advertisers to still succeed on iOS. As we’ve seen with our global clients thus far, SKADNetwork is hovering at only around 18% of available inventory with over 30% traffic misattributed to Organic. Given SKADNetwork’s areas for growth and still around 60% of users falling into “Not Determined” through App Tracking Transparency Framework, we knew we had to find an iOS 14 privacy-compliant solution that would allow for performance marketing.  With Adjust’s Conversion Modeling, we bridge the gap between consented, device level users and non-consented, aggregated users. Our advanced modeling will allow clients to measure their campaigns performance while maintaining their user’s privacy. 

- Katie Madding, Chief Product Officer, Adjust

AppsFlyer: Disruption often leads to transformation and innovation. Marketers will have to find new ways to guide their decision making process based on data they can fully trust that delivers maximum accuracy and granularity, while still protecting user privacy.

We believe that the increasingly privacy-centric ecosystem will push marketers to test advanced measurement methods to deliver higher value and results. For example, validating marketing efficiency through incrementality tests which allow advertisers to measure the real impact of their marketing investments, and optimize their performance towards true incremental lift. Furthermore, managing all of their spend and cost models under one roof will be more critical than ever to gain the holistic insights they need to understand and optimize their marketing budget and ROI.

- Ben Roodman, AVP, Business Development, AppsFlyer

Branch: In the short term, so much of the industry is still so underprepared for iOS 14.5 that we might see a noticeable decrease in buying activity: many advertisers may basically be forced to take a time-out for technical reasons over the next month or two. We might also have a temporary return to more simplistic campaign measurement techniques — such as CPM or CPC — as a way to keep something flowing while everyone rushes to complete their tech updates. Advertisers that did prepare in time will likely consolidate toward the few ad networks that got SKAdNetwork up and running prior to iOS 14.5 (including big walled garden networks like Facebook, Snap, and Google), and we'll likely see a short-term rebalance toward Android.

Longer-term, it will be much more difficult to get hard data on ad campaign ROI for iOS. That could make it harder to justify a lot of investment, though it's probably still too early to tell. Owned and organic/earned channels will see a bump in importance, since they do not rely on IDFAs and are not in scope of Apple's new policies, and will therefore be relatively undisrupted.

Re-engagement campaigns, unfortunately, deserve a special callout: these are likely to experience extreme disruption (if they even remain possible at all) for two reasons: SKAdNetwork only supports installs today, making it technically impossible to measure conversions that happen when the user already has the app installed at the time of ad click, and without IDFAs there will be no functional way to generate targeting lists of users that already have the app.

- Alex Bauer, Market Strategy Director, Branch

Kochava: IDFA opt-in for user privacy - It’s in the marketer’s best interest to stop talking about deprecation of the IDFA and start focusing on communicating that value exchange with their customers. While the IDFA will undoubtedly be less available, it's likely that the actual availability will fall somewhere between what we're seeing with the opt-in rate for the General Data Protection Regulation (GDPR) and the predictions from the doomsayers. Invariably, some types of app categories will see higher opt-in rates and some apps will see better opt-ins vs. their peers. 

  • CPA is dead - Media buying will move to cost per mille (CPM) which means big spenders need to spend more than the hyper-casual apps that have historically grown via cross promotion. 

  • Consolidation - We've been talking about this movement since 2012, and it's finally here. The markets are white hot, the timing is right, and access to capital is looking good for those companies with solid business metrics. There are also lots of struggling companies looking for a new home which represents great buying opportunities. Companies that focus on app monetization will seek to acquire other companies with large audiences to leverage the IDFV for inter-company measurement and targeting.

  • Fraud will increase - If the industry has grown sick of talking about fraud, get ready to talk about it even more. It will move one level up to the impression and many of the online viewability companies are not equipped to make sense of mobile signals while the mobile measurement companies have been mired in the post-click world and will need to adapt.

- Garrett C. MacDonald, EVP Strategic Sales, Kochava

Singular: The most significant shift for marketers is going to be their inability to measure performance on the user-level. Although there are still many unknowns about exactly how user acquisition will differ in a post-IDFA world, there are certain strategies marketers are preparing in advance to adapt.

When it comes to creative strategy, marketers will need to optimize for broader appeal of their target market instead of hyper-granular segments. Hyper-personalized ads will become scarce on iOS with the loss of user-level measurement.

Another thing to consider with creatives is that as the targeting capability goes down, creative becomes even more important as a differentiator. And if you know your best, most valuable players well — and what they like — you can use creatives in ways to "target" more users like them without actually having targeting ability.

On that note, we think there will be an increased focus on first-party data. In particular, BI teams are looking into the best ways to leverage their internal data on users to understand who their high-value customers are. And with SKAdNetwork's limited timers mechanism, BI teams would gain an advantage by predicting the customer's lifetime value in a short time period.

Additionally, we've heard marketers will be investing more in web campaigns that take prospective users to an advertiser's owned website before prompting them to install their app. In this web-to-app flow, the user acquisition part is done on the web, so attribution isn't based on the IDFA, instead of leveraging cross-device identifiers and deep linking technology. Plus, web-to-app flows create yet another conversion path with good UX for users that you can also accurately measure, which is essential in such a time when there is a lot that is changing.

Lastly, we're also hearing that many marketers will double-down on user acquisition efforts for Android while they adapt to the changes on iOS.

- Eran Friedman, Co-Founder & CTO, Singular

Even without the IDFA, there are still solutions we can apply to rise above the limitations of ad identifiers. Keep an eye on our Insights page so as not to miss the second part of this article series. We interview experts from major ad exchanges to learn their predictions on how ad monetization and marketer’s buying activities will adjust because of privacy changes.

Topics: Marketplace Insights