Why Mobile App Marketers Are Struggling To Measure CTV Ad Performance

One of the many changes that Connected TV brought about is that it democratized television advertising. Companies that previously wouldn’t dream of promoting their products on the big screen now regard CTV as a regular part of their media mix. 

Among these companies there are a significant number of major mobile app providers. During the pandemic, they started to move substantial parts of their video advertising budgets to CTV inventory - in 2022, the number was estimated to be 60% in the US alone. The main driver was that CTV enabled advertisers to combine the reach and storytelling abilities of traditional linear TV with the precise targeting of the digital space and more recently, the rise of interactive CTV ad formats.   

While it’s undoubtedly true that app marketers can achieve great things by connecting the big screen with the small one, they face real limitations when trying to understand how CTV performs in the context of their broader media mix. Some of these limitations are technical, others are political - and nasty. Anybody who makes money selling ads on CTV should care about them. 

Limitations From Going Cross-Device

Let’s discuss the technical limitations first. Most of them are rooted in the fact that when mobile marketers promote their apps on CTV, the user journey is cross-device. In a single-device user journey, for example when someone clicks on an ad on Instagram or Google Search, performance is assessed using a method known as “deterministic measurement.”

Deterministic measurement involves an attempt to match the unique ID of the device on which the ad is displayed with the unique ID of the device on which the app is installed. Measurement providers in the mobile space initially used device IDs to make these matches, until mobile platforms introduced advertising IDs in order to give their consumers more control over the ways these IDs were used. So, whenever a match is found, the app conversion is attributed to the network that served the ad. 

In the case of a CTV ad, deterministic measurement is impossible, simply because there are two devices involved and the unique IDs will never match. Therefore, mobile marketers who want to assess the performance of CTV ads for their business rely on what is called probabilistic modelling. Probabilistic modelling is a somewhat embarrassing euphemism for fingerprinting. 

Fingerprinting was developed as an alternative for deterministic matching in case unique IDs were not available (or, in a desktop setting, cookies could not be used). Measurement providers try to make a match between ads and conversions using certain device characteristics, such as the screen resolution, operating system and most importantly, the IP address. In a cross-device situation, probabilistic modelling aka fingerprinting is reduced to just using the IP address, which is the only common identifier potentially available on both the CTV and the mobile device. 

Obviously, this method is not watertight. For a match to be made, the CTV device and the mobile device need to be connected to the same household wifi. But people can respond to CTV ads when they are not at home and even if they are home, their mobile phones may not be connected to their household wifi. It is difficult to assess the size of this issue.

Research I conducted myself while working at Adjust in 2021 (a mobile measurement company), showed that on average, 85% of mobile app installs occurred via household wifi. A reasonable score, but not a perfect one and moreover, one that puts CTV ad networks at a disadvantage compared to social media, search and other inventory providers. 

The Politics of Attribution 

Attributing CTV ads to mobile conversions is not a question of measurement alone. The measurement providers mobile advertisers use (AppsFlyer, Adjust, Singular, and Kochava being the main ones) apply so-called attribution waterfalls to define what ad should be given priority in case multiple ads match. By default, these waterfalls prioritise the deterministic approach over the fingerprint approach and clicks are prioritized over impressions. While this makes sense in a single-device context, particularly for mobile and desktop, it puts CTV ad networks at an unfair disadvantage in two regards, as they can only be measured probabilistically and their ads are not clickable, at least not in the way mobile ads are clickable. 

To make matters worse, the most commonly applied attribution model also does not work in their favor. Ad attribution takes place using the last-touch model. This means that credit is given to the ad people saw or clicked on last before converting. Obviously, the ad people see before converting is much more likely to be an ad on social media or search than a CTV ad, even though it plays a pivotal role in getting people excited about mobile apps. 

The last-touch model has been challenged for as long as it has existed. When I was working for one of the first major affiliate networks back in 2004, we were already aware that it was biased towards certain types of publishers. The fact that it is still around can be explained by the fact that the social media and search ad networks which it advantages are the most powerful networks out there. Without their data, measurement providers in the mobile space do not have a product anymore, which empowers these networks to dictate what attribution models to use. CTV ad networks looking to get fair credit for their ads are fighting an uphill battle. 

Potential Solutions

But there is good news for CTV ad networks too.

There are plenty of solutions in the market to tackle the limitations of fingerprinting. Leveraging location and other data, household device graphs can map sets of devices and thus tell, with a certain amount of certainty, whether CTV and mobile devices belong to the same household or not. CTV ad networks can then leverage these device graphs to inform their own measurement and at least correct the image that mobile measurement providers are offering. They should alsopressure mobile measurement providers to use these solutions, even though this might come with compliance challenges, particularly in the EU. 

Right now is a good moment to do so, as measurement based on fingerprinting has made a revival ever since Apple usage of their advertising ID (the IDFA) dependent on the consent of their users. As opt-in rates are roughly around 20% globally, measurement providers and mobile app marketers cannot rely on deterministic measurement alone anymore. In fact, even the major networks have started putting more emphasis on fingerprint-based measurement, even if nobody in the industry dares to call it that way. 

On a more substantial level, there are two developments that will support CTV ad networks in the long run. The first is that despite the pressure from the major networks, more and more players in the space are starting to offer alternatives to the last-touch model - Sweden-based Funnel is a good example.

Even mobile measurement providers are trying to develop alternatives. At Adjust, I launched the so-called assist dashboard, which enables advertisers to figure out to what extent people installing apps had engaged with ads on multiple networks. This means that even if app conversions are attributed to social media or search, advertisers can still see when CTV ads contribute. 

The other development is that we are rapidly moving into a world where apps are being used on multiple devices. Gaming companies, after taking a ‘mobile-first’ approach for over a decade, are now going to great lengths to push their users to install their apps on their PCs as well. The fact that Google introduced a Windows-emulator to facilitate exactly this says it all.

Interactive ads on CTV are pointing towards a nearby future in which the mobile phone and the CTV device are part of the same experience. In fact, CTV has the potential to become the household’s central digital hub, used to buy, play and do the same things people now buy, play and do on their mobile phones. 

The future of apps is cross-device. All CTV ad networks have to do is enable consumers on the big screen.