Nielsen is sharing some results from a test last summer in three markets in which it used set-top box data from Charter Communications to refine ratings at the local level -- and in some cases found that it could virtually eliminate the dramatic statistical variability for individual dayparts.
The Nielsen test analyzed data collected in July 2010 in three Charter markets: St. Louis, where Nielsen offers its Local People Meter service; Greenville, S.C., where it uses set meters; and Reno, Nev., where it collects ratings data via paper diaries.
A key finding: By interpolating set-top data, Nielsen was able to eliminate between 50% and 90% of the variability for ratings in a diary market. The information and research company used a "weighting and factoring" methodology to create a hybrid rating that combined National People Meter data with local set-top data.
"One of the criticisms of existing ratings in diary markets is that they're too variable," said Pat Dineen, Nielsen's senior vice president of local audience measurement, adding, "We find that usage of television increases when we look at set-top box based ratings versus diary ratings."
Nielsen's efforts to improve the accuracy of local ratings come as competitors like TiVo and Rentrak have launched new research products -- based on set-top data -- aimed at local markets.
What's next for Nielsen? Dineen said that "this is the beginning of a journey for us to figure out the role of set-top box data" in the broader context of the ratings it reports. The Nielsen proof-of-concept used set-top data from 46,000 Charter subscribers in St. Louis; 12,000 in Greenville; and 6,000 in Reno.
Whatever role set-top box data plays going forward, it will be used to supplement Nielsen's core panel-based ratings, not replace them, Dineen noted. Currently, set-top box data is available commercially for fewer than one in five U.S. households, he said.
With supplemental set-top box data, Nielsen should be able to provide ratings for programs that previously would not have had sample sizes big enough to produce a statistically meaningful number, Dineen said.
"This is the initial lab work that allows us to have really good conversations with our customers," Dineen said. "We want to make ratings changes that will be good for our clients."
As for using set-top data on a national basis, Dineen said the sample size in the 20,000-household National People Meter is "sufficient for most viewing." He added, "Don't get me wrong, most national advertisers would crave more data. But that's not on the short list" to incorporate set-top box data into the NPM ratings.
