A Debate on Standards Part 3: AppsFlyer Talks About Attribution Challenges

By Diksha Sahni | September 14th, 2017

Welcome to a new part in our series, A Debate on Standards, as we open up a space to discuss the problem that a lack of standards in mobile attribution poses. In the previous editions of the series, you've heard from AppLift's MD and CRO Maor Sadra on his insights into the cookieless environment of mobile attribution, and then Adjust's CTO, Paul Müller, on his views around the current mobile attribution space. Today, we bring to you thoughts from AppsFlyer's Director of Product Management Matan Tessler on the issue of mobile attribution.

Q: To start off, can you tell us where you see the greatest challenges in the current mobile advertising ecosystem when it comes to attribution?

A: This is a great question. Complexity and fractious connections are the greatest challenges facing the mobile advertising and attribution ecosystem. This complexity becomes further compounded when it enables fraud and pollutes omnichannel data, as we will discuss a bit later on.

Mobile attribution has become far too complex and difficult to maintain, often without good reason. Mobile attribution all too often requires far too much work from the marketer, including manually configuring postbacks, manually mapping fields in audience segmentation solutions and manually rebuilding offline pivot tables every time they need more advanced analysis. These inefficiencies, together with fraud, are two of the biggest challenges mobile marketers face.

At AppsFlyer, every solution is developed in partnership with the marketers that rely on our platform, as well as the ecosystem that relies on this data. We invest an incredible amount of time delivering solutions that make our data and platform more efficient, user-friendly and accessible. For example, a couple of months ago we released Audiences, an advanced audience segmentation solution. Rather than asking marketers to manually configure and QA each field with every supported media partner, we went the extra mile, dedicating significant resources to pre-configuring every possible element of the Audiences platform, delivering a true end-to-end solution for marketers.

Q: There’s a lot of talk about working with the right attribution models, but many over-simplify the approach. How should advertisers choose the right model?

A: The best way to explain attribution models, is to start with an understanding of how attribution data is used. There are two primary use cases for attribution data: marketing optimization, and performance payments to networks. While some of the larger networks offer their own attribution solutions that follow their own standards, strong, third-party unbiased and transparent solutions like AppsFlyer allow advertisers to define their own attribution lookback windows.

For example, let’s say a large US gaming marketer is running a CPI campaign on a major network, let’s call it Network A. Network A uses a 28 day lookback window. This US gaming marketer however, has set their AppsFlyer attribution to a 7 day lookback window. In this case, though the marketer will have to pay Network A for an install that occurred 10 days ago, their AppsFlyer reporting will not attribute this install to Network A. As a result, the marketer can see their actual cost-per-install based on their preferred attribution methodology. When looking at their cost-per-install in Network A’s reporting dashboards, they will see a slightly lower cost-per-install, whereas their AppsFlyer dashboard will show a slightly higher cost-per-install, as some of their installs did not fit their preferred attribution methodology.

While some have suggested that it would be easier for all marketers to agree on a single attribution methodology, we continue to see strong demand for customizable attribution lookback windows. For example, marketers running a limited-time promotion are unlikely to value an install 6 days after the promotion has ended, whereas marketers with a longer lead cycle will likely prefer longer attribution lookback windows.

At AppsFlyer, we provide every advertiser with the option to customize their click and view-based attribution windows, delivering the optimal amount of flexibility and transparency without distorting their data integrity. This desire to deliver dynamic solutions for data-driven marketers is why we provide the flexibility to define different attribution lookback window per tracking link, so marketers can determine what works best for their needs.

Q: How do you address the issue of attribution fraud?

A: Mobile fraud is a real and growing issue. Our unique market-share enables the world’s largest mobile marketing database, and have a dedicated team researching and building comprehensive solutions based on this data.

Increasingly, we are seeing new and advanced types of fraud that seek to trick attribution providers, and tap into lucrative CPI and CPA campaigns. While we are happy to see industry-standard solutions such as install validation, and CTTI/MTTI (click/mean time to install), these technologies are best suited for blocking basic fraud such as click flooding, install hijacking and bot-based installs. When deployed at scale, these techniques can identify problematic publishers, helping to address basic fraud at its source. However, these solutions fall short when facing more modern approaches to the more costly, install fraud.

The fastest growing type of install fraud is advanced, device-based fraud. Massive device farms are using real devices coming from “clean” IP addresses to generate “real” installs at remarkable scale, across every vertical and region in the world. Last year, we introduced DeviceRankTM, a proprietary technology built to combat the growing threat of device-based fraud. With of our unique scale and machine learning, we are actively blocking attribution of installs from devices that are known to perpetrate fraud.

For example, a major US-based e-commerce app recently found over half a million dollars in installs coming from a specific set of sub-publishers. Whereas their average sub-publisher had fewer than 10% of their installs coming from devices not yet rated by DeviceRank, a few dozen sub-publishers were sending over 90% devices that had not yet been rated. By quickly identifying these DeviceID Reset Marathons, the marketer recognized over one million dollars in savings.

Q: What trends are we seeing in 2017 and what are the current limitations of mobile attribution that must be fixed?

A: 2017 has been a fantastic year for mobile. Whereas in the past, many marketers were only just getting started with attribution, in 2017, we are seeing more experienced marketers looking to improve or upgrade to stable, reliable, unbiased solutions.

One of the biggest trends we are seeing is around data richness, data integrity and integrating mobile data into an omni-channel customer journey. We are seeing increased demand for integrated cost and ROI reporting, especially from leading networks like Facebook and Google. We are seeing incredible demand for integrated fraud solutions, and especially device-based fraud solutions. We are also seeing greater demand for advanced reporting, from agency transparency to multi-touch reporting and integrated pivot analysis. We have also seen a lot of demand for connections back into the marketing cloud, especially leading cloud providers like Adobe. This allows marketers to understand the role that mobile plays in the broader customer journey, as well as to use of mobile attribution data to improve their omni-channel ROI.

Unfortunately, there are also a few areas where some mobile attribution solutions leave much to be desired. For example, some attribution providers still deploy incomplete fingerprinting that is heavily reliant on IP data. This leads to inaccurate attribution, especially on shared IP addresses such as airport or transit terminal WiFi networks. Our NativeTrackTM platform automatically disregards these misleading IP addresses. In one recent example, a marketer that switched to AppsFlyer found a 15% difference in their fingerprinting attribution after making the switch, as their previous provider was heavily reliant on their IP fingerprinting attribute. Further research revealed that their previous provider had an incomplete fingerprinting solution, which had likely misattributed hundreds of thousands of installs.

Q: How important is harmonization of attribution measurement across different channels?

A: Every promotional channel needs the attribution logic that will work best for the marketer. For example, rewarded video often has a longer delay to install than standard display. Apps around events such as sports apps also have different average install times. In today’s market, marketers should pay close attention to their attribution settings, and use like-standards for similar campaigns. This is why it is imperative that all mobile marketers pay attention to their attribution lookback windows, and tweak them as needed.

Q: The first impressions is/should be the most valuable, yet, the last touch point with the user is what gets attributed - how do you think this problem should be solved?

A: This question makes a bold assumption, that the first impression should be the most valuable. While the first impression may be particularly noteworthy for video or playable ads, I don’t know that I would say that a first display impression is more valuable than a last touch playable ad or video view.

While as an industry, we have typically focussed on the last-click, we have long offered multi-touch reporting and recommend that all clients consider the value of each impression and click across the customer journey. Today, media is still bought in a last-click model. In this dynamic, though marketers need to pay attention to their last-click attribution to understand their spend, marketers should also consider the value of contributing networks across the customer journey.

Remember, mobile attribution is used in two ways: for marketing optimization and performance payments to networks. While you may have to pay networks based on your last click, remember to consider the broader, multi-touch customer journey when optimizing your marketing campaigns.

Thank you Matan for your time!

Stay tuned for the next part in series in the upcoming weeks. Our readers will get to read on mobile attribution in greater detail through interviews with some of the major attribution partners in the industry on their take on topic of mobile attribution as well as standardization. Our readers will also be able to get deeper insights into the topic with planned webinars with representation from key stakeholders in the industry to start an active dialogue.

Diksha Sahni
Diksha is a Content Marketing Manager at AppLift and is based out of our Bangalore office. When she is not behind her computer writing, you can find her binge watching her favorite movies, finding her happy place at a dance studio, and checking off places on her bucket list.

12063761 1030943930291273 1184227129633473729 n