For decades, the practice of determining which TV shows people watch has depended on a brew of hard data and educated guesswork. Fill-in-the-blank written diaries, electronic meters that gauge what’s on TV, and an assortment of statistical models cranked out by computers combine to produce the audience-ratings currency that a $70 billion industry uses to assign value.
The deeply entrenched presence of a TV-ratings infrastructure orchestrated by one company, Nielsen Media Research — the dominant TV-ratings and information provider — means it’s unlikely the current way of doing things will be supplanted anytime soon. But a new technique for divining viewing patterns is starting to emerge in the TV business, with surprising velocity. It measures television usage by tracing exacting indicators stored within the electronic innards of modern cable TV set-top receivers.
As the population of these digital set-top devices and associated measurement software increases, and as numerous companies collaborate to coax detailed tuning information from the boxes, TV industry researchers believe an important new era of audience-measurement is dawning.
“In some ways, it is very much the Holy Grail, because it allows us to capture information at a granular level that’s just not available anywhere else,” said 27-year television-research veteran Jonathan Sims, a vice president at Comcast Spotlight, the advertising-sales arm of the largest U.S. cable operator.
From tests Comcast is conducting to evaluate information captured from its digital set-top boxes, Sims said it’s possible to discern minute but important viewing patterns that tend to be obscured or ignored by prevailing ratings and measurement techniques.
Particularly interesting to him is the ability to track how viewers are using interactive features like video on demand and specialized advertising applications that attract highly interested viewers, albeit often in small numbers. Collecting information from set-tops “allows us to capture finite usage levels that are just flying under the radar now,” he said.
CLICKSTREAM COUNT
Researchers get excited about deriving viewing data directly from cable’s digital set-top population because doing so gives them a glimpse into precisely how people tune into and out of channels and content. Buried within the software stacks that populate many digital set-tops are programs called middleware, which faithfully record every interaction, or “clickstream,” between a user’s remote-control and the digital set-top unit.
When viewers raise the volume, change channels or order on-demand movies, those electronic transactions are noted by built-in middleware applications present in digital terminals. By using a return path — the phone line or built-in Internet connection that sends data from the home back to the cable company’s data center — operators can collect information about what’s going on inside those set-tops.
The reason the software that tracks clickstreams is inside set-tops in the first place has nothing to do with TV ratings. Instead, the software is intended primarily to keep watch over the health of the set-top box and to report back diagnostic information that can help alert system technicians to problems. But its ability to capture and forward information about the instructions sent inside the box has prompted software developers to devise new extensions that are well-suited for collecting channel-tuning information.
Those efforts are only beginning to take hold. “There aren’t a lot of people that are using this yet,” said Tracy Geist, senior vice president of advanced digital-TV marketing at OpenTV, one of a few set-top software providers that provides a set-top activity measurement application. (Scientific Atlanta, Navic Networks and Microsoft also make telemetry software that measures set-top activity.)
But that seems likely to change as more cable companies and research firms warm to the idea of collecting detailed TV-viewing information directly from set-top boxes. Geist, whose company supplies middleware to direct-broadcast satellite provider Dish Network and has announced a forthcoming deployment with Time Warner Cable, believes the cable and research industries are serious about turning data captured by digital set-tops into the foundation for a new approach to TV viewing measurement.
“This is the second or third wave of that kind of buzz, but I think now it’s on the serious side of the wave,” she said.
NIELSEN ROLE
Not surprisingly, the growing interest in figuring out what people are watching by collecting massive amounts of set-top instructions has captured the attention of Nielsen Media Research. In February, Nielsen established a new service unit, Nielsen DigitalPlus, that’s devoted to mining data from cable’s digital set-top population.
According to a Nielsen statement, the new service “will work with set top box data from cable-system operators and satellite providers to create new insights and services for clients by integrating set top box data with other Nielsen information.”
Jed Meyer, a 15-year Nielsen veteran and senior vice president of the DigitalPlus unit, said Nielsen executives have believed for a long time that set-top receivers offer extraordinary promise in producing precise TV-viewing information. But limits to the internal computing memory associated with early-generation digital receivers made it difficult to embed the software needed to cull detailed usage data, Meyer said.
That has changed as cable companies have embraced more muscular versions of digital set-tops and expressed more willingness to explore ways to collect information on what’s going on inside them. Now, Meyer says it may be possible to combine some of Nielsen’s intelligence about viewer demographics and other attributes with raw clickstream data derived from digital boxes.
“We think set-top boxes provide a mechanism to provide electronic measurement on a more cost effective basis,” he said.
Nielsen’s establishment of the DigitalPlus unit has rippled across the TV research world in more ways than one. Shortly after launching the unit, Nielsen hired a well-known cable-advertising industry researcher, Jeff Boehme, as senior vice president of business development for the new unit. Since 1992, Boehme had worked for National Cable Communications, the national advertising-representation firm owned by large multiple-system operators.
Word of Nielsen’s new set-top endeavor also caused consternation to a Florida firm, ErinMedia, which had hoped to raise investment capital to support a new set-top TV research business. ErinMedia, which is embroiled in a lawsuit with Nielsen, said it laid off most of its employees to conserve cash after the Nielsen announcement scared away prospective investors.
ErinMedia contends that Nielsen’s method of tracking the viewing habits of 10,000 is anachronistic and antiquated.
“Let’s pretend that you’re an Asian woman in a certain demographic and you have a bad dinner,” ErinMedia CEO Frank Maggio said in April. “Bad Chinese food and you throw up and you go to the bathroom and you shut the TV off. Well, there’s 10,000 Asian females that had bad food that did the same exact thing at the exact same time in Nielsen’s panel. It’s severely flawed. It’s not true.”
“Having looked at set-top box data, we have yet to find any two set-top boxes that have behaved the same,” said Maggio. “Tens of millions of [pieces of] clickstream information over hundreds of thousands of boxes over hundreds of days, and we have not yet found any two set top boxes that behave the same.”
ErinMedia propounds the benefits of using set-top box data and of using logical algorithms based on hundreds of thousands or millions of users’ behavior to decide when, for instance, a viewer has stopped watching, even if a TV set is left on. This is a process Maggio calls “inverse mathematics,’’ which can take the mounds of information stored in large databases, based on set-top activity, and compare the viewing habits of TV watchers in different ZIP codes with different age, income and ethnic backgrounds.
That will give a better picture of what large portions of the populace actually watch — and what is different about those audiences — Maggio contends.
The presence of the TV industry’s dominant research firm in the emerging set-top research category is a sign of how seriously the category is now being taken by researchers who long for more precise indicators of how people watch television.
With set-tops that record every remote-control click, “most definitely, there’s better data than Nielsen can ever hope to offer on household tuning,” said Dave Zornow, a software developer and TV-industry research consultant who chairs the Committee on Cable Network Audience Research.
But Zornow and others caution that collecting viewing data from set-tops doesn’t alone provide a complete picture of how people watch TV. For one thing, researchers recognize there’s a gap between which channel a set-top box may be tuned to, and what, if anything, appears on the television screen. Because few cable customers ever power down their set-tops, a digital receiver will commonly report that it’s being tuned to a particular channel for long durations of time, even if the TV set was turned off hours before.
“A lot of homes leave their set- top on all day,” said Nielsen’s Meyer. “If you’re just measuring whether it’s on, you’ll think viewing is going out the roof.”
Research companies are attacking the so-called “on-off bias” by developing statistical models that attempt to account for the set-top vs. TV set gap partly by tossing out elongated viewing sessions.
Another concern is that while set-tops spit out a verbatim record of which channels have been tuned, they know nothing about who’s doing the tuning. Research executives say the information won’t be useful unless it’s paired with some sort of demographic or behavioral overlays that attempt to provide a more complete picture of the relationship between what’s being watched and who’s watching it. There’s also the matter of whether information gleaned from set-tops will comply with systems and processes advertising agencies use to evaluate media.
“We’re not looking for any paradigm shifts here,” said Zornow, who is the founder of TNG Research. “It has to sort of fit into all the flow charts and plans and expectations that agencies already have.”
Privacy, too, is a big issue cable companies must contend with as they navigate the line between collecting highly detailed viewing information and protecting consumer rights. OpenTV’s Geist notes that complying with rigid state privacy-protection laws — California, for instance, demands a “double opt-in” affirmation from customers before companies can collect their data — may be a challenge. Generally, cable companies are working to make sure data from set-tops are aggregated into collective statistical buckets that make identifying individual viewing behaviors impossible.
THE OTHER END
At the same time that cable companies are beginning to yank viewing data from set-tops, they’re also exploring ways to cultivate viewing information from the other end of their networks — in headend facilities where large digital video servers spit out on-demand TV programs that are unique to particular households.
Every night, computers maintained by the Portland, Ore., data-monitoring company Rentrak hash through a dense thicket of information culled from cable video servers across the U.S. to produce reports indicating viewing levels for tens of thousands of video-on-demand programs, many of them containing advertising.
The job is part science and part real-world intuition. Cathy Hetzel, senior vice president of Rentrak’s on-demand research service, said one of the challenges is accounting for differences in the way some cable companies and their content providers describe the same programs. To accurately aggregate data across programs, Rentrak’s software has to interpret minute variations in the spelling or abbreviation of movie titles, for example.
As in the set-top research category, however, it turns out that just counting things isn’t enough. Rentrak is working now to color its video-on-demand tracking data with a new product, AdEssentials, that incorporates demographic profiles derived from ZIP code-based overlays.
Similar efforts are under way elsewhere. One large local cable-advertising company has enlisted the Kansas research firm Griffin Media Research to fill in the blanks associated with the demographics of on-demand TV. Based on telephone polling across local Comcast markets, Griffin Media is developing profiles of on-demand viewers that can be presented to advertisers to help round out the viewing picture for a new medium.
“What we find cable ad sales organizations want is more information about the on-demand viewer than they can get simply from counting program views,” said Griffin Media president Bennett Griffin.
The same demand will likely prevail on the set-top side, where a cadre of companies now is angling to play a big role in marrying the raw data of clickstreams with educated guesses about who’s watching. On one side are the middleware providers that supply the essential information-technology ingredients for set-top measurement and reporting. On the other are research analysts including Nielsen, Rentrak — Hetzel said the company is working now to develop a process to aggregate set-top viewing data — and TNS Media Intelligence.
NEW COMPETITION
The onset of a set-top research category could give Nielsen rivals a legitimate chance to upstage the big TV-research firm, some believe. “Because it’s still so early on, and because we are definitely vendor-neutral, I think the field is open to a few key players,” said Comcast’s Sims.
As the category rises, it may produce new benefits for local cable advertising organizations that long have been critical of Nielsen’s diary-based methods of estimating viewing. A long-standing cable-industry complaint is that diary-derived ratings understate or even ignore viewing to niche cable channels that attract small but highly engaged viewers.
Nielsen’s Meyer thinks a big benefit of set-top research will be the ability to report viewing for narrowly targeted networks. “Local cable advertising folks want to have a number for those niche networks … not just the hash marks they see today,” Meyer said.
Just as important going forward will be finding ways to report viewing to on-demand and interactive-TV sessions that reduce television to a highly personalized event in which a single household gets a unique program stream. Sims says the hybrid nature of cable television, which now delivers blends of traditional linear television with interactive viewing sessions, demands a more precise mechanism for understanding viewing patterns.
“The ultimate solution is to capture both the linear advertising and how it works to attract people into interactivity,” he said. “We need unified, holistic measurement to capture that data. The whole space requires that set-top box.”
