This article is part of a mini-series ahead of Mobile World Congress and the launch of our Modeling as a Service offer for DSPs. In the series, we'll be exploring the headwinds and opportunities for DSPs in a fast-evolving mobile performance advertising space.
A person’s smartphone provides a uniquely intimate snapshot of who that person is and what they care about. For advertisers, smartphones present unparalleled opportunities for considering a person’s context, tastes, interests and needs at any given moment. It’s no surprise that in-app ad mobile ad spend is up over 25% YoY to $45B at the end of 2017.
The challenge for DSPs is that although the opportunity in mobile is accelerating, it’s challenges are equally substantial for two major reasons: the average mobile user evolves quickly and tying their identity to a given device is becoming an increasingly fluid endeavor.
Consider the challenge presented by the nature of device IDs. One may assume that IDs would be the sole static identifier in the constantly-shifting mobile landscape. In reality, there are several events that trigger device ID churn including OS updates, manual resets and of course the purchase of a new device. With every reset, old IDs become stale and the previously known-identity is once again anonymized.
The other, more complex, challenge is that context surrounding an individual is in constant flux. Their needs, interests, and affinities are always evolving and shifting. In fact, the average app experiences 80% churn over a 90-day period. In short, an indicator that a specific user fits into a target segment today, may be stale by tomorrow. Conversely, someone that did not fit into a particular segment yesterday, may qualify today.
The Opportunity Cost of Static Audiences
In essence, static pools of device IDs are quickly becoming invalid, many of the ones that remain valid are quickly becoming irrelevant, and ones never before considered relevant are quickly qualifying for a given audience segment.
The question before us then, is to what extent churn impacts a given audience segment. In other words, how much does that audience change as new IDs qualify and existing IDs go stale?
To get at that answer, we analyzed a DSP data stream and tracked the unique device IDs over a 4 week period. Each week, we took a snapshot of device IDs and analyzed a.) how many disappeared each week and b.) how many net new IDs appeared as newly qualified users.
Though we can’t parse the volume of device IDs that became invalid from the number of apps that simply were no longer being used, we get a clear picture of the opportunity cost associated with relying on a static pool of device IDs.
After the first week, we see that about 2% of the original device IDs had already gone stale. That number increased 4X by week two and at the end of the month, nearly 1/3 of the IDs were no longer valid. Depending on the timeline it takes for a DSP to segment an audience and run ads against a pool of device IDs, a significant portion of the original audience are either irrelevant or invalid.
Newly Qualified Users
The rate at which new users qualify for a given segment is even more jarring than the rate at which they transition out. After the first week, 24% of the total audience was comprised of newly qualified device IDs - users that fit the targeting parameters of the original audience.
Quantifying the Total Opportunity Cost
When we consider both factors - newly qualified users and disqualified or invalid IDs - we get a clear look at how quickly a static audience becomes irrelevant. After the first week, 26% of the audience is either newly qualified or has churned out of the segment. By week 4, only 14% of the original audience is still relevant for the campaign. After just one month, the original device ID pool is essentially irrelevant.
What does this mean for DSPs?
Data freshness is paramount for high performance mobile advertising. By limiting your optimization algorithms to a static pool of device IDs, you limit the efficacy of your technology. To close the gap on performance with the Facebook and Googles of the world, it makes sense to emulate what makes those platforms so effective: dynamic audiences.
Appnique uses proprietary Machine Learning and AI technology to help DSPs access timely user data from app stores and integrate it into your own technology to amplify what makes your solutions unique and deliver better performance for advertisers.