The history of television ratings is of great interest to me because—to quote Maya Angelou—”if you don’t know where you’ve come from, you don’t know where you’re going.”
Market research firm The Nielsen Company, founded in 1923, is perhaps best known for its ubiquitous television ratings, but Nielsen actually started out measuring radio ratings before moving to TV in 1950. As you might expect, its early forecasting tools were relatively simple. Audience shares were initially calculated using paper diaries where customers wrote down the shows they watched and when. Nielsen then used these diaries to compile statistical models that approximated viewing numbers.
Of course, as technology evolved so too did audience measurements, and just as paper maps gave way to GPS, Nielsen’s strategies similarly developed. The company now forecasts ratings using a variety of television meters and set-top-boxes that remotely monitor TVs to see what shows and ads people are watching and for how long. This anonymized data is then used by networks to calculate advertising rates for each show.
But the technological evolution is only part of the ratings measurement story. Much of our existing system is just as much a consequence of the industry’s business evolution as its technological development. For instance, Nielsen is no longer the only game in town. Comscore, founded in 1999, is also making a significant impact on the field of ratings measurement, resulting in a greater variety of methodologies for collecting data and, I would argue, the quality of ratings has benefited as a result.
A storm of screens
The ratings industry has a problem, however, and it is that traditional television isn’t the only way to view content anymore. For example, the premiere of the final season of the wildly popular show Game of Thrones saw a staggering 17.4 million viewers, but in a reflection of how much TV—and TV ratings—have changed, 11.8 million of those viewers watched the epic battles on linear TV, whereas the remaining 5.6 million people, which is about one in three viewers, streamed the show on HBO Go and HBO Now.
Measuring viewership on so-called OTT subscription streaming services like the above is quite problematic because there is currently no established method for third parties to verify audience numbers. We only know about the streaming figures for Game of Thrones because HBO told us.
One solution has been for participating networks to drop a digital tag, or identifying code, into their content that allows the rating firms to measure the number of views, but that requires the participation of each platform. Sites like Netflix that don’t publicize ratings (except what they selectively release) have resisted this effort thus far.
Another technological development that poses a conundrum for ratings is time-shifted viewing, like DVRs. How many people watch television on a delay through either a TiVo or DVR provided by their cable operator? I know I often do. Such viewership can be a very significant factor for overall ratings, but incorporating it into measurements is often tricky.
There have also been new developments with regards to the advertising business, some of them quite recent, that have impacted ratings. This year has been the epicenter of the streaming wars, with Disney and Apple joining an already crowded field led by Netflix, Amazon and YouTube as well as smaller players like Sling TV, DirecTV Now, PlayStation Vue, fuboTV and the aforementioned HBO Now. With each new announcement, the question of how advertisers can accurately determine ROI on OTT becomes more pressing.
Secondly, I still have Upfronts fresh on my mind and can’t help but ponder the future of linear television business through the lens of ratings. Although streaming appears to be the future, right here in the present, TV remains the most impactful advertising platform, and costs are somewhat counterintuitively rising despite declining audiences. EMarketer predicts Upfronts commitments to rise to $21.25 billion in 2019.
Simply put, the industry on the sell-side is consolidating. AT&T recently purchased Time Warner, Disney bought most of 21st Century Fox, and with fewer sellers, the price of inventory increases. A smaller number of business controls all TV advertising inventory, which means the ratings data advertisers rely on to assess prices is becoming more valuable.
Like before, some of the above challenges are being solved through technological progress. There are many efforts underway to calculate audiences with a campaign-based model, i.e., to holistically measure video advertisement delivery across traditional TV, connected TV, desktop, mobile and otherwise.
Relatedly, it seems clear that in the near future, ratings accuracy will be improved through advances in machine learning, which holds the potential to deliver cost-effective methods for tracking viewership patterns and inferring future trends. The NBCUniversal Context Intelligence Platform, for instance, analyzes show scripts and closed captions to identify the perfect moment to promote a product—so that, for example, an ad for beer would follow a party scene. These spots also appear on users’ mobile devices and social media channels as part of an omnichannel campaign.
NBC’s solution is an excellent indication of what advances machine learning can bring to our field, laying the foundation for fast, reliable, but also smart solutions that factor in new, unstructured data points. It also makes me wonder which industry players will rise to take advantage of these advances. Historically, the ratings industry has benefited from progress in business plans as much as any new piece of equipment. What is certain is that TV viewing habits have modernized, and there are many pressing reasons as to why the methods of evaluating audiences must follow suit.