I am stunned. Thirty years after I started covering the broadcast networks as a young Wall Street Journal reporter, the demo still forms the foundation of the $80 billion a year in TV advertising in the U.S. and billions of dollars more overseas. Michael Jackson, Ronald Regan, Nirvana, Pets.com, Bear Stearns—all gone. Yet advertisers and programmers still think that the most important information to have about the audience is their age, sex, and race. It's time to go beyond the demo, beyond look-alikes. It's time for act-alikes: to apply what we know about the science of what people do, how they behave, the interests they share.
The approach, which is called advanced behavioral targeting (ABT), goes far beyond the demo that has defined American audiences since the early days of broadcast television. Where the casual observer sees a melting pot of viewers, ABT classifies viewers based on shared interests, minute-by-minute TV-watching habits, buying behavior (both online and off), social media engagement, and more.
"We've been using act-alike modeling for promos, not just look-alike," says John Curran, director of media analytics at RSG Media, a New York-based software and consulting firm. "Moving beyond age-and-sex demo targeting is essential to growing ratings in this ever-shrinking linear landscape." RSG Media, the company that sponsored this series, helps content companies harness Big Data and sift through dozens of data feeds to unearth customer insights.
Welcome to the Club
With act-alike modeling, networks that want to attract viewers build "reciprocal affinity" groups—communities of viewers who watch the same networks, the same shows. "It's like love for each other. You like me and I like you, so let's find other people who like both of us," explains Curran, "only the love is for what our group thinks is great television."
Earlier this year, RSG Media launched a breakthrough platform, Media Mantra, and already several major cable networks are seeing audience growth using 10–30% fewer promos. Powered by proprietary, machine-learning algorithms, Media Mantra is a "campaign optimizer." It uses proprietary algorithms to constantly analyze incoming viewership data from Nielsen AMRLD (All-Minute Respondent Level Data), the underlying linchpin of television. Media Mantra uses AMRLD's insights to recommend promo schedules for new shows, allowing programmers to reach more of their target viewers with fewer spots.
RSG Media's clients keep many of their real-life examples secret. So here is a fabricated example that illustrates the capabilities of Media Mantra:
Viewers of HBO's Game of Thrones are of a particular stripe (probably covered in animal fur), a community of some eight million who watched the season five finale last year. Now let's say 30 percent of them also are fans of The Walking Dead on AMC, whose season five finale last year garnered 15.8 million viewers. By inference, given that 2.4 million viewers of The Walking Dead are also fans of Game of Thrones, the other 13.4 million zombie-watchers on AMC might be ripe for converting over to Thrones.
"Our machine-learning algorithm learns from past successes, but more importantly, past failures, and automatically adjusts to those successes and failures when moving forward," says RSG Media's Curran. "What's particularly beautiful about it is, Media Mantra takes all this knowledge and does the work for you.
Media Mantra is working its machine-learning magic for various networks, and it has delivered sometimes surprising increases in reach and conversion rates, getting high percentages of viewers to tune in to a show whose promos they saw on other networks.
One factor driving these advances is the surge in TV viewing on handheld devices. Suddenly programmers can know far more about viewer behavior, about how much they watch and where, about which shows hold their interest the longest and which shows they stop watching after a few minutes, and about which viewers tune in to a program after watching a promo.
The aim, in this era of fragmenting audiences and the rise of millions of separate Internet video streams, is not as focused on trying to reach more total viewers as it is on targeting more precisely the right viewers on linear TV—those who are most likely to watch a promo for a new show and later "convert" and tune in. Remember, not all GRPs are created equal.
The RSG Media platform can also break up these audiences into quadrants delineated by viewing behavior, an approach proffered by Nielsen itself. "Gold" viewers are those who tune in to your network often and stay tuned in for a long time; "Silver Sliders" are those who visit your network often but flit about and don't stay tuned in for very long; "Occasionals" are people who don't visit your network often, but when they do they stay tuned in for a long time; and "Lights" are viewers who don't watch your network much at all and leave quickly when they do deign to pay a visit. Using these quadrants, networks know not only to whom they should promote their shows, but they also know on whom they should stop wasting time.
So when Media Mantra runs its machine-learning calculations to schedule a network's promo spots, it knows to target the middle two groups: the Silver Sliders and the Occasionals. What makes the system extra clever is that it realizes that one network's Lights are another network's Golds. The system tracks these lighter viewers to learn where they land, because this information can be useful for other shows or other channels.
The optimizer machine's ultimate objective is to transform your Occasionals into Golds by identifying and targeting Occasionals who are Gold on other networks, and then using the power of data science to entice them to your outlet. And, as soon as they have turned to Gold on your network, Media Mantra adapts, ignoring these newly converted Golds as it starts to search out still more fans elsewhere.
People tend to like what they know; it's comfortable. This is often the case with new technology. What we're seeing now is a few select networks, sprinting to the profit line, turbo-charged with data-fueled insights.
Big Data Projects on the Rise (But Data Use Could Be Better)
Big Data and the Future of Business
How the Super Bowl Uses Big Data to Change the Game
Big Data Is the New Ground Game: How Ted Cruz Won Iowa
Cloud, Mobile, and Big Data Analytics Drive Latest Rounds of Megadeals