This whitepaper first appeared on ANA’s Forward, May 23

Finding the Right Audience the Right Way

Data derived from digital media, in particular the consumer-controllable mobile advertising ID used as a central identifier, provides the ability to bridge the gap between platforms while still maintaining the integrity of individual data sets and the user’s anonymity.

The use of a publisher’s first-party data, particularly that of a publisher with robust consumer touchpoints and good notice and choice mechanisms, is the most reliable way to deliver advertising to a desired audience. While marketers can supplement with third- and second-party data, basing a targeting effort on the first-party data of the publisher that’s actually serving the ad is likely the best way to get optimal results. For instance, using the publisher’s first-party data as the foundation for defining an audience can help eliminate any drop-off between sources.

Still, the need for well-developed and tested machine learning models to expand the audience and reach new consumers should not be overlooked. Focusing too closely on only the immediate conversion piece of the marketing funnel may, in the long run, have a negative impact on audience expansion by oversaturating the same audience pool and limiting overall brand exposure.

Demonstrating Model Behavior

Given the breadth of linear television, modeling and segmenting at the broadest level, followed by even more targeted messaging through OTT and digital platforms, allows for a healthy marketing ecosystem. Taking advantage of the breadth of linear and the precision of digital media develops an ecosystem of the very best data and predictive models possible.

While an advertiser may know of a particular consumer or group of consumers predisposed to making a purchase within a specific category in the imminent future, there are likely many other high-value consumers with similar psychographic and behavioral profiles to those who have already been identified by the advertiser, of whom the advertiser is unaware. Using machine learning lookalike models to accomplish the dual tasks of reaching a broad audience and focusing on high-value audiences is essential, even for those with the widest reach.

There are vastly different perspectives and approaches to predictive modeling, as well as to where predictive modeling fits with regard to person-to-person marketing. Even when using lookalike models, thinking specifically about digital apps and OTT, the outcome is still very much person-to-person. The differentiation lies not in whether predictive models still yield one-to-one outcomes, but in whether the basis of the models is explicit user-provided interests, or implicit interests derived from permissioned firstparty data, or consented-to survey data matched to first-party data sets.

The benefits of using predictive models are rooted in extension-of-reach and the idea of conversion beyond the known pool, but this approach, capable of using anonymized identifiers, can also help shield consumers’ privacy. In that way, predictive models can become even more valuable.

The consumers’ experience, their privacy, their data’s security, and their trust have to be the top objectives when using data. In fact, these need to be the priorities for not only the brand and its marketers, but also data scientists, analysts, and anyone who comes into contact with consumer data. The focus of a data-science practice that uses anonymized data enriched with profiles relying on a significant amount of modeling, becomes what should be done on behalf of the consumer versus what can be done.

Turning it On, Turning it Up

When applying the same approach taken with digital and OTT to linear television, marketers should be focused on understanding the indexing of particular segments against specific shows, dayparts, and networks. Mapping a digital segment built through a machine learning model or derived from behavioral data that’s been tied to specific content against linear viewership data, for example, can yield an index of how people who intend on purchasing an automobile manifest in the audience of a particular show.

The ability to build an overarching graph across platforms and subsequently plan and optimize inventory accordingly can lead to more accurate targeting and deeper insights than marketing plans based on siloed thinking and guesswork. Ultimately, when used correctly, the path to robust sequential advertising begins to develop such that a true funnel approach, from reach to targeted creative to outcome messaging, becomes more accessible and more efficient through the bottom of the funnel.

Channeling the New Way Forward

As the line between media platforms blurs even further, and as consumers increasingly focus on the quality of the content far more than on the medium in which they consume it, the need to understand relevance, sequence, and interests grows progressively more important.

Permissioned first-party data, and associated predictive and vetted models can be used to deliver high-quality, trustworthy, contextually relevant advertising. Marketers who build pertinent attribution models and ensure meaningful, data-centered, cross-platform executions will not only improve their marketing efficacy, they’ll create a better consumer experience as well.


Dana McGraw is the VP of audience modeling and data science at the Walt Disney Company, a partner in the ANA Thought Leadership Program. For additional partnership opportunities with our performance solutions, please click here.