This week’s guest blog post is contributed by Heather Robertson, Senior Manager, Partner Marketing, Oracle Data Cloud.
During the first half of 2018, marketers saw a growing number of headlines pointing to the impact of using audience data for advertising. Long story short: Audience data is creating new challenges—and opportunities—for marketers.
We spoke with Mike Schumacher, VP, Data Science, Oracle Data Cloud, about the value of audience data and key areas marketers should keep top of mind when partnering with data providers.
Mike Schumacher: As the industry has matured, we’ve seen increased scrutiny and demands for transparency in many facets of adtech, including ad units, open RTB, viewability, etc. As a result, we’ve seen common currencies emerge, along with improved understanding within the industry, enabling marketers to make better decisions.
Schumacher: Intuitively, we know that relevancy matters. Great creative, a viewable ad unit, or a compelling offer can’t overcome the wrong audience. Audience-centric media buys simply perform significantly better than ordinary campaigns. Whether the KPI is click, conversions, or other actions, the right audience data enables marketers to capitalize on relevancy.
Schumacher: Significant increases in data volumes, coupled with innovations in audience construction and reachability via cross-device, resulted in a dramatic increase in audience options for media buyers.
All data is not created equal, and it’s important that marketers work with their audience partners to understand their options and make fact-based audience selections.
Schumacher: First, marketers should demand fact-based evidence, ideally proof of efficacy of audiences. Second, marketers ought to request transparency of their audiences, including where data are sourced and how audiences are constructed. Finally, marketers should look for partners to activate audiences in many channels with high-fidelity integrations.
Schumacher: Audiences can be built in a variety of ways, ranging from “if/then” business logic to complex machine learning models. Business-logic audiences depend on manual definitions of desired behaviors. For example, “Users who read car reviews online might be in-market for a car.” These rules-based audiences are easy to understand and provide the ultimate control.
Modeled audiences may leverage the same underlying data, but use that data in an optimal, multivariate manner. They also score individuals using the collection of their data attributes with empirical assignment of each data element.
With all that said, a true test of an audience is not whether it uses declared vs. observed data, or whether it’s rules-based vs. modeled. The true test is whether it performs. We found that while methodologies matter, the underlying strength of the data used in the audience and an accurate ID Graph are the most important components for building an audience.
Schumacher: Syndicated audiences tend to cover the most commonly requested audiences, with a blend of data and methodologies, to enable sufficient scale with strong performance. Custom audiences, including custom models, typically take advantage of data-driven insights and technologies to yield the highest performance audiences with configurable scale.
Schumacher: Our general framework for audience assessment allows us to compare different audience reach techniques, including demographic data, purchase-based data, custom models, etc., in their ability to predict and reach future buyers. Minimizing the cost of reaching future buyers and giving brands an opportunity to influence a consumer’s future spend are the primary goals.
In research conducted in 2018, we’ve found that modeled or data-driven audiences tend to reach significantly more (25%, 50%, 100%+) future buyers than demographic targeting and many multiples of general, run-of-site based media.
What’s also interesting is that this data-driven improvement is consistent. That is, we’ve yet to uncover a situation across a wide variety of brands and categories where audience didn’t yield a significant impact.
About Mike Schumacher
Mike oversees several groups of data scientists at Oracle Data Cloud who build, validate, and deploy analytical solutions on behalf of advertisers and consumer platforms.
Mike has more than 15 years of applied analytics experience, including data science positions within advertiser, media publishing, and technology organizations. He has a deep understanding of audience modeling, campaign impact measurement, and media optimization algorithms.