The fundamental nature of digital privacy is to limit the amount of personal information that’s widely available by third parties. Considering the incredibly successful — and effective — world of ad retargeting, it seems like privacy and retargeting couldn’t co-exist.
To help marketers understand, I’ve asked Loïc Anton, Chief Product Officer from the mobile-app retargeting company Adikteev, to sit with us and share the current state and the future potential of retargeting in mobile advertising.
— Adam Landis, Head of Growth at Branch
First, can you tell us a little about what Adikteev does and your role with the company?
Adikteev is a mobile growth engine focused on user retention. As a retargeting demand-side platform (DSP), we help you maintain and grow your user base through retargeting campaigns. We also supply user churn prediction technology to help app marketers prevent users from leaving the app. Adikteev started out just running branding campaigns for French companies but switched its focus to performance marketing for apps in 2017.
As Chief Product Officer, I’m in charge of the definition and execution of the company’s product roadmap. I’ve been at Adikteev for eight years, starting out in Paris, then moving to Berlin, then returning back to Paris. Adikteev was my first experience in ad tech, so all of my app marketing experience has been on the user retention side.
Can you give us a brief summary of how retargeting worked before and how some of the more recent privacy policies have made this more difficult?
Historically, mobile app retargeting involved getting a list of device identifiers from an advertiser who wishes to re-engage its valuable users, and then looking for those device identifiers in advertising opportunities available on the market. So the whole concept relied on “matching” these device identifiers between the advertiser’s analytics solution and the advertising inventory offered by publishers. In 2021, Apple introduced a change for iOS that limits the access to the identifier for advertisers (IDFA) for both advertisers and publishers, therefore, making it more difficult to perform this matching. Google’s Privacy Sandbox is set to be a much more developer-friendly initiative while still prioritizing user privacy.
So this seems like retargeting would be — well, “dead” is a strong word — but certainly made ineffective. Is this the case?
Initially, app retargeting has seen its reach reduced on iOS because of limited availability of IDFA, which now covers about half of the users it used to back before Apple implemented the changes. However, to complement this partial coverage, we’re now seeing alternative solutions such as churn prediction or probabilistic retargeting rising to address the same need: keeping users engaged and maximizing revenue potential of the existing user base.
Interesting, what is probabilistic retargeting? And how does churn prediction function as an alternative?
Probabilistic retargeting involves gathering a lot of contextual information about a user such as other apps and publishers they’re using then making predictions to determine whether or not the user has the app installed. DSPs then bid on the user under the probable assumption that they’ve already installed the app and are an existing user.
On the other hand, churn prediction can be used in tandem with a deterministic retargeting strategy. Churn prediction models analyze an app audience to determine the likelihood that users will churn. Marketers can then use these lists of likely-to-churn users to run retargeting campaigns, adjust in-app messaging during certain moments in the user journey, or even run paid social campaigns. Using churn prediction models like ours is a way app marketers can maintain a robust user base without being as afraid of having fewer device IDs. Knowing which users will churn and when is valuable information that allows them to target them at the right time to keep them engaged before they leave.
So how exactly DOES retargeting work today in the Apple ecosystem?
The deterministic approach relies on the principle of double consent: when iOS users download an app they’re asked whether they agree to share their IDFA with the developer of this app. Retargeting works as soon as users consent to sharing their IDFA both in an advertiser’s app and in a publisher’s app. This double consent allows ad buyers to recognize the same users across two apps and potentially retarget them.
The probabilistic alternative when device IDs cannot be matched is to build probabilistic models built on supply-side platforms’ (SSP) available data points (such as location, publisher, app, etc.) to go after users that have the highest likelihood to use your app.
What about Apple’s recent announcements of SKAN 5.0, how will that change things for retargeting?
Even though SKAN 5 features have not been officially detailed yet, Apple did mention during the latest WWDC that it will support the measurement of re-engagement: app publishers will be notified not only when a user installs their app like today, but also when users reopen their app thanks to a marketing campaign (usually after a threshold of inactivity time). This means that retargeting campaigns, which mostly drive app opens and not app installs, will be measurable with SKAN much more easily. However, it will not ease the challenge of identifying — at bidding time with a probabilistic approach — who already has the marketed app installed in order to retarget existing app users as much as possible.
Any recommendations you’d give to your clients to make retargeting more effective, today and in the future?
For apps within the same vertical, we’ve seen that opt-in rates can vary to extremes (i.e. the apps with the best opt-in rates can have double the opt-in rate of the apps with the worst opt-in rate). There is no real consistency, only that the most successful apps have a strong pop-up strategy. On iOS we always advise clients to optimize the way they ask for user consent in order to maximize the opt-in rate: for instance, by properly explaining to their users why they need to access their data. This can already have a drastic impact on the scale and effectiveness of their retargeting campaigns. Overall we’re seeing this consent rate slightly growing with time, so it seems like app developers are moving in the right direction here.
In addition, we usually recommend exploring complementary solutions to extend user retention such as churn prevention campaigns through churn prediction or cross-promotion campaigns targeted to likely-to-churn users, sort of a portfolio-level retargeting network.
What do you think will happen for the future of retargeting? How will it work?
Looking at the future, Google is bringing breakthrough innovation to the app marketing industry with the Android Privacy Sandbox project, which includes new tools for app marketers to show ads to specific audiences (including retargeting ads) without having to use device identifiers. Even though technical specifications are not set in stone yet, the concept of this ID-less retargeting is to let the device itself manage the audiences it belongs to, instead of sharing the device identifier to third parties who would maintain audience lists on a remote server. This is great because it reaches both goals of enhancing user privacy and maximizing the revenue generated by apps from their existing user base.
In other words, it is very likely that Google will succeed where Apple initially failed (i.e. allowing publishers to effectively retarget their users while enhancing data privacy).
What would you most like to see in future market developments?
Android’s innovative approach to user privacy is making iOS solutions look outdated already. For example, the latest version of iOS SKAdNetwork still doesn’t support a major feature like event-level reporting, whereas Android Attribution Reporting does support it. So I would definitely like to see a renewed strategy on iOS regarding privacy-friendly marketing features, and generally an approach more open to feedback from the app marketing industry which is a critical source of revenue for app developers.
What are some things you’re worried about that will happen in the market?
Device identifiers. With all the privacy concerns they were raising they have always been useful to detect and filter out undesired practices in advertising such as bots and fraudulent behaviors in general. So even though I definitely believe that users should have a better control of their personal data, I sometimes worry that hiding this data completely from the market, and increasing the opacity of the advertising chain as a consequence, will leave more space for fraud that manages to stay under the radar.
Still have questions about retargeting in a post-privacy world? Branch is here to help.