fbpx
AdvertisingDigitalFeaturedInsightsMarketingOpinion

Overcoming ‘signal loss’ in a privacy-first world

"Privacy compliance may have bound our feet, but our hands are free. There are privacy-friendly ways to model data. And thanks to a new tool in the arsenal, AI, those models can be accurate enough to serve our need for targeting," AppsFlyer's Paul Wright says.

Paul Wright, General Manager for Western Europe and MENAT at AppsFlyer
Paul Wright, General Manager for Western Europe and MENAT at AppsFlyer discusses options of playing to a privacy-conscious audience.

Marketing professionals live in ‘interesting times’. In the mobile space, Apple’s App Tracking Transparency (ATT) framework presents a significant challenge by limiting the visibility of user data for marketers, often leading to what is known as “signal loss” — a reduction in the data signals that once guided campaign optimisation.

Experience-builders — marketers focused on crafting seamless user journeys — did the best they could. Worldwide, as of the first quarter of 2024, 84 per cent of gaming app developers and 68 per cent of non-gaming developers had integrated ATT by showing a consent prompt.

But given the option to share or shield, the majority of users chose the latter. Even the United Arab Emirates, while still more favourable than Western countries such as the US (44 per cent), has an ATT opt-in rate of 66 per cent.

When we factor in the announcements by major browser producers to phase out third-party cookies (Chrome is expected to follow in the footsteps of Firefox and Safari by year end), times are getting more and more ‘interesting’.

The era of user-level data has been all but consigned to the history books. Marketers may pine for the good old days but if they are to address the white noise of the modern age, they must chase every faint pulse and unite them to form a new picture.

It may not be the crystal-clear IMAX image of yesteryear, but with the right methods, marketing professionals can still weave the strands together into a sufficiently truthful rendering of campaigns to give them back confidence and control.

Find the gap

Marketers must return to data modeling. Privacy compliance may have bound our feet, but our hands are free. There are privacy-friendly ways to model data. And thanks to a new tool in the arsenal, AI, those models can be accurate enough to serve our need for targeting.

While Apple’s SKAN and Google’s Sandbox, frameworks designed to protect user privacy by limiting access to individual-level data, certainly create gaps in real-time visibility, AI-driven modeling can close these gaps and deliver real-time geo-level data and a range of LTV (lifetime-value) data.

Self-reporting networks arguably suffer most from the measurement problem, relying as they do on deterministic attribution. But by combining probabilistic modeling with their deterministic approaches, they too can prevail.

According to AppsFlyer analysis, Snap increased its share of the iOS install-base by 138 per cent, and its effective cost per install (eCPI) was reduced by 46 per cent. Google and Meta have also taken the modeling route to circumvent signal-loss challenges.

Google Ads now uses machine-learning models to fine-tune the bidding process, and Meta’s Ads Manager uses predictive modeling for ad targeting. Both companies also use modeling to personalise user experiences, which in turn optimises ROI per campaign.

These are just some of the techniques on hand to help marketers overcome signal loss. But even when using all available modeling tools, we can only eliminate some of the white noise. To de-fuzz our campaign picture, we must go a little further.

Funnel vision

The age of privacy has barred marketing executives from accessing the ‘action’ end of the funnel, so it may be time to turn their attention to the ‘awareness” end. Top-of-funnel events can yield a profusion of data that is tied to creatives and campaigns.

This data falls into two broad categories. The first is the rich, granular telemetry from creatives themselves. In a process made significantly easier by AI, we can identify user-generated content, animation, real-life footage, colors, scenery types, specific objects, text, and so on.

Second, we can sift out enriched-engagement types (EET): Prevailing attribution standards revolve around clicks and views, but these approaches give an incomplete picture.

Marketers looking to overcome signal loss should look at the overlooked data in between clicks and views, at interactions that occur between these two actions, such as swipes, hovers, or time spent on a particular element to provide fresh perspectives on performance and attribution logic.

Outside of top-of-tunnel data, we should reevaluate third-party data. Its value is now greatly diminished because it can no longer be shared, so marketers must use more first-party data. It follows that they must now make its collection a top priority.

They must establish agile, yet compliant, ways of asking users to share their info and give permission to use it. Marketing teams must also brainstorm processes for determining that data is clean, actionable, and fit for purpose in both owned and paid media.

First-party data can be stored in a data-collaboration platform rather than on a company’s own servers. The platform will serve as a trusted environment for first-party data monetisation, audience activation, and measurement. You will notice that this has reintroduced third-party data but in a way that is privacy-compliant. We have also reinvigorated the value of third-party data.

Continuing our search for new sources of data, we can revisit the Web, CTV, PCs and consoles, commerce media networks, and out-of-home campaigns. Each touchpoint is an opportunity to engage with users and by combining the data from these channels, marketers can make the campaign picture a little less blurry.

Analysing user behavior across platforms can shed fresh light on the ins and outs of cross-platform marketing and how it can drive business outcomes.

Additionally, we can look into “single source of truth” (SSOT), the ultimate countermeasure to the data fragmentation that has so paralyzed marketers. SSOT combines data from several sources and puts advanced technology to work on an array of tasks, from the identification and removal of duplicate installs to the automated population of missing data. A

ppsFlyer analysis has revealed that SSOT can deliver to the average app a 29 per cent boost in attributed installs, a 40 per cent reduction in eCPI, and a 62 per cent increase in attributed revenue.

Something’s coming through

Tired of trudging through a permanent sandstorm, marketers will gain fresh perspective when they realise they can restore visibility themselves.

Privacy empowers consumers, giving them greater ownership over their data. In time, that sensation will lead to confidence and a newfound ability among businesses to acquire and retain customers.

In the interim, marketers are not as tethered as they think. There are many options for playing to a privacy-conscious audience. User-level data is difficult, not impossible, to come by.

Driving growth is difficult, not impossible. Measuring with confidence is not beyond us. We need only put our heads down and innovate our way through the wind.

By Paul Wright, General Manager Western Europe and MENAT, AppsFlyer