Have you seen our latest webinar with AppsFlyer at MAMA Connect? Back in December 2020, we talked about remarketing measurement in the new era of privacy, covering incrementality and remarketing measurement in the no-ID world.
In case you missed it or want to watch it again, we've included the video along with a summary and recap below.
Key Points and Video
In this one hour video, AppsFlyer Senior Product Manager Roy Yanai and Product Marketing Manager Pavel Dimshiz are joined by Remerge’s own Eugen Martin to cover the following topics:
- Incrementality measurement: what it is and why it's important
- The importance of remarketing and the COVID-effect
- iOS14, remarketing, and user privacy
They also tackled questions from the audience:
- The typical amount of users who would have converted anyway?
- The biggest challenges with incrementality measurement
- iOS14: What's going to happen in the future?
- How do I start with incrementality testing?
Summary of Key Points
Incrementality - what it is and why it is important?
Incrementality measurement has become a widely used term in our industry. Through incrementality, we can measure and understand "how many of my retargeted users would have converted anyway?"
Incrementality measures how many of the retargeted users each ad brings incrementally, on top of the ones that are actually already going to convert. The concept applies a scientific method of testing in which users are grouped into control or test groups, where the control group is unexposed to ads and the test group is exposed to the ads.
How remarketing has progressed over time
Remarketing has continued to grow in the past few years and even in 2020, as more and more marketers are interested in retaining the users they've already acquired. With the growth of this channel, questions about organic cannibalization have also increased and marketers want to understand how effective their marketing efforts really are. Remarketing proves many benefits and use cases for different apps such as onboarding, driving repeat purchases, or reactivating lapsed uses.
The COVID Effect and the remarketing uplift
Despite the steady growth, we've seen a 12% drop in retention rate of the average app in 2020 contrary to the usage of apps: people are installing more apps but are less loyal to their apps. Those who use remarketing in their strategy experience a 70-85% increase in retention rates (source: AppsFlyer).
iOS14, remarketing and user privacy
The future or remarketing measurement in post-IDFA world: AppsFlyer touches on their incrementality roadmap which includes measuring the incremental effect based on aggregated data for both UA and retargeting.
Questions from the audience
Conversion
What is the typical amount of users that would've converted anyway?
This depends on the segmentation targeting strategy. In gaming verticals for example, the goal is to resurrect long-gone users (such as 30 days+ inactive). In this situation, organic conversions will be low and most conversions would be incremental.
It's very different from shopping or food delivery verticals. Retargeting those who recently purchased and upselling them to use the app more frequently by giving them recommended products means that the organic conversion would be much higher. In this case, incrementality measurement is much more required.
For example: Uber is running a campaign in San Francisco. Showing a few ads to a few people will likely show amazing attributed KPIs but their organic conversion would be really high.
Incrementality Measurement
Why do you think advertisers are not yet there in terms of measuring incrementality?
When it comes to learning anything, there is still a steep learning curve in adapting this methodology. We can compare incrementality as that extra phone in the drawer that you're not sure when you want to start using. Incrementality measurement is a resource-heavy process that requires experts being involved. However, we see good development within the market to have an approachable and easy-to-implement solution on incrementality.
The challenges of Incrementality measurement
What are the biggest challenges of incrementality testing?
- There are different methodologies of measuring incrementality: randomized control trial based (RCT-based) experiments target the test population with a campaign and don't show ads to the control group. Then the question is, how do you look at the results? Some methods have biased approaches.
- Long-term testing: most experimenters use the frequentists method for statistical significance. Running an experiment until the P-value is below 0.5 is a flawed method because it doesn't let you look at results through time, but only once, while Bayesian statistics are more suitable for the long-run.
- Think carefully about what you're going to look at: if you ask how incremental something is and what the lift and conversion rates are, these aren't necessarily related to the business.
For example: "Is a 1% lift good or bad?" If 1% means 2.5M incremental revenue at a cost of 1.5M, it's good incremental ROAS, but looking at the lift of 1% won't be good - so have a look at the KPIs you're looking at.
- How critical it is to connect the process of incrementality to your business goals: the first thing marketers need to figure out is "What do we want to test? What is the hypothesis we want to validate?" This will be the basis of the process and the data that comes out of the measurement will either validate or reveal their marketing efforts.
- Engineering effort in experiments: There's a lot of data in between - sending the audiences, receiving the ads, calculating the lift - which gets harder to do across multiple networks.
- Cannibalization and making sure it doesn't happen: A lot of customers have overlapping audiences that are competing - the control group that shouldn't be exposed to ads are sometimes actually exposed in other campaigns that end up being similar. Prevention can be done by suppression lists and making sure your audiences are in place.
iOS14 and the IDFA
iOS14 and the ATT framework impact - thoughts about the future? What's gonna happen if there is no IDFA, and how do you measure accordingly?
1st point of view (Android): Google IDs are still active and available for Android which is 70% of devices. Although iOS cohorts are more valuable as we know them, there's still a huge market for Android devices.
2nd point of view (iOS): AppsFlyer is building a solution to be based out of aggregate data. The industry is moving where the measurement, products, and mobile apps are going to change based on what they can attribute and collect. It's not only about the measurement but also the ecosystem and how the products are working on that.
From a remarketing perspective: the future of mobile apps themselves will change and the future of advertising products around them will change. In the past, we've had luxurious situations where the product only appeals to the 0.0001% of the users - the super-whales - and marketed to them through user profiles. This won't be possible in the future, at least not at scale. This means that business models will have to change towards building products that are much more appealing to the broader population.
ID-Free Measurement
ID-Free remarketing world - how can you still measure the impact it has?
In a post-ID world, both isolated retargeting and user acquisition will no longer be possible. Using the same example of Uber running a campaign in San Francisco and showing ads to everyone: it's very likely that this will be a re-engagement campaign as it is safe to assume that a majority of the users have been or are existing customers of Uber. In this case it's important to measure both the UA and remarketing impact. Currently, Apple doesn't have an infrastructure to do that, so Remerge is building a new incrementality solution to measure that part.
Starting with incrementality
What would be a good but simple tactic to start incrementality testing?
Whenever you start with a new channel, partner, vendor, ad network, suggest running a test right in the beginning. Test it at target scale - a short test but high scale. Testing small might show great performance in the beginning, but performance might not be the same when scaling up. By doing so, this holds the vendor accountable as marketers can hold them accountable based on the results. If the results are good, then the new channel (and such) can be continued.
