Incrementality Measurement for Mobile Advertising in the Era of IDFA and iOS14

Tim Koschella

Until now mobile advertising focused much on cost metrics (CPI, CPM, CPA) or revenue metrics like ROAS. But marketers are beginning to realize that there’s more to ad performance than just this. How do you measure the impact of ad campaigns?

The concept of incrementality measurement is not new: offline advertising has been doing this for a long time. How do you measure a shampoo being sold in a retail store? It’s all done through statistical measurement to connect ad impact with sales. Digital, on the other hand, has brought more one-on-one measurement into the industry. Digital – and the mobile industry in particular – has had the luxury to be able to measure the impact of ad campaigns very accurately. But with the growing industry concerns around the privacy issues of attribution, it begs the question whether mobile will go the way of offline to bring incrementality measurement as a standard practice? 

Through incrementality testing, advertisers can measure the impact of their ad campaigns by connecting the dots of how different variables affect the customer journey and path to conversion. By performing controlled tests, incrementality measurement calculates the “lift” by comparing outcomes from different scenario variables. This can help advertisers to determine which ad campaigns have the most impact. 

How incrementality works

At a very basic level, incrementality measurement starts with an A/B Testing of a random selection in a controlled environment. Take for example, as an advertiser you run a campaign for four weeks. With incrementality testing, you can select a set of users and see how users who are shown an ad (test) vs those who aren’t (control) convert. App marketers usually rely on Device ID to pool and split users into control groups. This difference gives us the incremental conversion and helps us better understand the impact of the campaign.

It is important to understand that “incrementality” and “lift” are not interchangeable here. While incrementality is calculated as: 

(Test – Control) / Test = Incrementality; showing the % of purchases

Lift is calculated as: 

(Test – Control) / Control = Lift; showing the % likelihood that a user will purchase or return to the app if the ad is shown

Advertisers can test both channels or campaigns to understand the incremental lift in performance. By having a holistic view of incrementality, marketers can allocate budget spend, improve audience selection as well as customer journeys. 

Incrementality measurement in UA

Advertisers have been using incrementality widely in the context of retargeting, but not so much in UA. One of the reasons being, it is easier to attribute the user identification to measure if the sales were impacted by a retargeting push. Mobile marketing relies on attribution to understand campaign effectiveness for UA, with MMPs and self-attributing networks (SANs) providing the last-mile data.

But that might be up for an upheaval since the announcement of iOS14. While IDFAs has provided for a convenient and precise tool to track a mobile advertising campaign on Apple devices, an increasing number of users have chosen to opt out of it, thereby limiting the advertisers’ options. iOS14 changes this by bringing in a new opt-in requirement, where app developers will have to get consent before they can share the IDFA with third-party monetization partners.

Increased user privacy is always a welcome step in advertising and companies that offer the most value are the ones who innovate their products and services while respecting the need for data privacy. The latest changes on iOS are a step in this direction, and I believe, we will see many companies offer disruptive solutions that will eventually lead to incrementality measurement in UA becoming a standard way of judging the effectiveness of mobile marketing channels. With the lack of unique identifiers for iOS traffic, there will also be a more level playing field between self-attributing networks (SANs) and non self-attributing networks. Up until today, SANs have been able to grade their own homework by having MMPs pass all IDFA based user data to them and returning their “claims” to the MMP APIs. This will fundamentally change and thereby create a more neutral comparison between marketing channels. However, this doesn’t mean that the data available will allow for better judgment of effectiveness. Quite the contrary, marketers will have to make more judgments based on incomplete or ambiguous datasets. Experience and “gut feeling” will become a more important asset in the mobile marketers’ skillset.

Looking at the future

Advertisers have been slowly moving towards incorporating incrementality models into their overall analysis and with the latest announcement from Apple, we are likely to witness this transition sooner than expected. The mobile industry as a whole will have to work together to adjust to this new ecosystem, and as the processes and standards change in the background, industry watchers will be keen to look at what developments come along the way in not just how ads are served but also anti-fraud measures. There are plenty of questions around the future of attribution. Apple is, for example, deliberating to use the SKAdNetwork API to receive metadata from ad clicks, and this will largely impact the future role of MMPs in the ecosystem. Since SKAdnetwork at its current stage has many limitations, How this will shape up the measurement on mobile and last-click attribution challenges is going to be interesting for all of us to watch.

Tim Koschella
An entrepreneur who appreciates high-end technology and data. Tim founded 3 other internet tech companies before Kayzen, each of which employs 100+ people today. He started the first one at the age of 23 and has learned a few lessons since. In his free time, Tim enjoys music, outdoor sports (water or snow) and traveling.

Recent articles