Home Mobile Verve Group Capitalizes On The PET Trend With On-Device Cohort-Based Targeting

Verve Group Capitalizes On The PET Trend With On-Device Cohort-Based Targeting

SHARE:

Google’s decision to delay third-party cookie deprecation in Chrome until next year dominated the ad trade headlines last week.

But when it comes to signal loss, mobile developers have already been there and done that.

Apple released its AppTrackingTransparency framework in April 2021 (after a delay of its own, mind you). App developers and their ad tech partners have been attempting to mitigate the impact of ATT on iOS addressability and measurement ever since.

Verve Group was early out of the gate in May 2021 with the beta release of an alternative targeting solution called ATOM (which stands for Anonymized Targeting on Mobile). It uses on-device and contextual signals to aggregate audience cohorts without relying on user-level identifiers, such as the IDFA.

On Monday, after three years of tweaking and running tests with partners, including both publishers and advertisers, Verve Group rolled out ATOM 3.0, an updated and more feature-rich version of the tool.

Long-term investment

Verve had planned to release the solution sooner, but it took a little longer than expected to get it market ready.

There are a few reasons for that.

One is because deploying ATOM through an existing SDK, which is what Verve did – ATOM is bundled directly into the HyBid SDK (formerly PubNative) – is technically complicated, although better for adoption in the long run, because the SDK is already integrated into the app.

But there were also other internal and external factors to contend with, said Anish Aravindakshan, Verve’s director of product marketing.

Cohort-based targeting and monetization was a relatively new concept in 2021, and publishers were hesitant to test an unproven technology.

Subscribe

AdExchanger Daily

Get our editors’ roundup delivered to your inbox every weekday.

Meanwhile, ATOM uses its own machine learning models to classify users into cohorts. Training those models is a complex, time-consuming and data-hungry process, Aravindakshan said.

This created a bit of a chicken-and-egg scenario whereby Verve needed adoption in order to get data to feed its models. But in order to get more adoption, ATOM’s models had to be as precise, effective and performant as possible.

Now, ATOM 3.0 is fully integrated into the HyBid SDK, which has a footprint across more than 10,000 app publishers with a collective reach of over 1.5 billion users.

Cohorts and context

Because ATOM 3.0 is shipped within Verve’s SDK, the solution automatically activates any time a user opens an app within the network.

If there is an ad request, ATOM runs machine learning on that user’s device to analyze in-app contextual signals (gestures, session length, session depth and so forth) along with device signals (keyboard language, battery level, whether the person is connected to Wi-Fi, etc.).

ATOM combines these signals to infer how to classify a user into a cohort (or multiple cohorts) for targeting. The process is 100% probabilistic, and no deterministic user-level information ever leaves the device, Aravindakshan said.

To start, Verve is offering a handful of broad categories of cohorts with more specific segments under each umbrella.

There are behavior and interest-based cohorts, for example, such as loyal or casual users, night owls, active gamers, sports fans or people interested in a healthy lifestyle. “Personal finance” cohorts categorize people by income level, while “mobility” cohorts segment users based on the likelihood of whether they’re at home or work.

ATOM can also deduce age range, gender, marital status, if there are children present in a household and whether someone is likely on a business trip.

Ask app not to track?Privacy push

One would be forgiven if the word “fingerprinting” were to pop into their head at this point.

But ATOM 3.0 is not a fingerprinting solution, Aravindakshan said.

“There is no use of any proxy ID: no IDFAs, no mobile IDs, no IP addresses and no hashed emails,” he said. “We are only reading contextual signals for a certain period of time restricted within one app with no cross-app information being shared.”

The whole point of ATOM 3.0 is to preserve addressability, but to do it in a way that passes the privacy sniff test. There’s been more interest recently in this type of privacy-enhancing technology (PET), said Gaylord Zach, Verve’s head of mobile product, spurred in part by the large platforms.

There’s Apple’s ATT (of course), SKAdNetwork, privacy manifests and the Android Privacy Sandbox.

As these tools and related platform policies are gradually enforced, Zach said, the ad industry is casting about for alternative targeting and measurement solutions, like ATOM, for example.

Beyond user-level targeting

But that does beg the question: Does the alternative perform as well as what it’s replacing?

And that question is not academic for an app developer like FunCorp, which publishes iFunny, an entertainment app with more than 60% of its users on iOS.

Shortly after Apple released ATT along with iOS 14.5 back in 2021, iFunny’s revenue dropped noticeably, according to FunCorp CRO Sergei Efimov. The app’s IDFA opt-in rate plummeted from 50% to around 15%, which iFunny managed to increase to 25% after making numerous product changes.

The opt-in rate is “still insufficient,” Efimov said, which is why iFunny focuses its user acquisition efforts more on Android than iOS these days, despite non-IDFA users generating two to three times lower revenue. But at least they’re addressable.

The iFunny team first started testing ATOM when it was in beta and has continued to experiment with the technology over its multiple iterations. Within a couple months of first implementing the tool, Efimov said, iFunny began to see an uptick in iOS revenue and monetization.

Even so, the economics on iOS are still out of whack, Efimov said. But cohort-based targeting is helping.

“We believe that the only chance to recover iOS revenues and return our users to the platform,” he said, “is by moving away from user-level targeting.”

Must Read

play button with many coins isolated on blue background. The concept of monetization of the video. Making money on video content. minimal style. 3d rendering

Exclusive: Connatix And JW Player Merge To Create A One-Stop Shop For Video Monetization

On Wednesday, video monetization platforms Connatix and JW Player announced plans to merge into a new entity called JWP Connatix. The deal was first rumored in July.

HUMAN Raises $50 Million

HUMAN plans to build a deterministic ID from its tracking of more than 20 trillion digital signals per week across 3 billion devices, which will aid attribution for ecommerce.

Buyers Can Now Target High-Attention Inventory In The Trade Desk

By applying Adelaide’s Attention Unit scoring, buyers can target low-, medium- and high-attention inventory via TTD’s self-serve platform.

Privacy! Commerce! Connected TV! Read all about it. Subscribe to AdExchanger Newsletters

How Should Advertisers Navigate A TikTok Ban Or Google Breakup? Just Ask Brian Wieser

The online advertising industry is staring down the barrel of not one but two potential shutdowns that could radically change where brands put their ad dollars in 2025, according to Madison and Wall’s Brian Weiser and Olivia Morley.

Intent IQ Has Patents For Ad Tech’s Most Basic Functions – And It’s Not Afraid To Use Them

An unusual dilemma has programmatic vendors and ad tech platforms worried about a flurry of potential patent infringement suits.

TikTok Video For Open Web Publishers? Outbrain Built It.

Outbrain is trying to shed its chumbox rep by bringing social media-style vertical video to mobile publishers on the open web.