A key component in Nielsen’s plan to convert from paper diaries to electronic data collection in small and medium-sized markets is still being tested, even though the conversion is already under way in 14 markets.
The component is the process by which the ratings firm hopes to produce demographic data for its diary households after they have been equipped with passive, electronic “code readers” — that is, by deriving it from demographics data collected from Nielsen Local People Meters in large nearby markets.
That the demo component was still very much a work in progress was revealed Thursday morning in New York by the Council for Research Excellence, an organization of media research professionals — funded by Nielsen — that studies issues in TV audience measurement.
Last month, when Nielsen confirmed that the gradual dismantling of the paper-diary system had begun in 14 markets, the company described the demo-data collection process in terms that were far more certain than they were at Thursday’s CRE meeting.
“It remains to be proven,” says CRE facilitator Richard Zackon, who is leading the 10-market test of the demo system.
Unlike local people meters, the code readers proposed for the diary markets are incapable of identifying which members of a household are watching TV at a specific time. Instead, these passive devices merely track which shows and channels are being viewed by “listening” to the audio portion of any shows that are being watched, and by detecting a Nielsen audio watermark embedded in the signal transmitted by a TV station.
Thus, the code readers might provide a more accurate picture of household viewing than the hand-written diaries did, but information about who’s watching — a crucial element in TV audience measurement — will be lost. To make up for that, Nielsen is proposing the derivative demos.
“This is a predictive model of local TV ratings using supervised ‘machine learning’,” explains Zackon, who is spearheading the group’s study of the data collection method. Supervised machine-learning, he says, “is an artificial intelligence technique in which the computer is presented with inputs and outputs in order to learn a general rule or algorithm that connects the inputs and the outputs. The inputs are data from markets all over the country and the [outputs are] the ratings in a particular local market.
“So the question is, can we develop a machine-learning or artificial intelligence algorithm to use the data from other markets with all the household characteristics to forecast the ratings in a separate market? That’s the idea,” he says.
But some at the meeting questioned how data taken from one market could accurately measure viewership in another, especially of sports, news and other local programming.
That’s the big challenge,” Zackon admits. “News and sports and any local programming are going to be the issues. How good will [the process] do? I don’t know.”
He suggested that the test might yield some ideas on how to apply other “variables” to determining demographic viewing patterns for local news such as how long a set of anchors has been on the air.
Also, he says, “It will be easier to predict the network news than the local news, but we might use the network news, then once we have that prediction, [use it] as a hook for the local news.
“We’ll see. If we knew the answers in advance, we wouldn’t do research.”
Zackon says the CRE’s test involves “taking data from people meters all over the country” and using the data to “forecast” ratings for 60 stations in 10 markets. The test markets are 10 of “the smaller of the top 25 markets.”
He says the test is on schedule to yield a final report on the feasibility of this demographic data collection method by the end of the year.
Ironing out the demographic-data issue is crucial to Nielsen’s plan to eventually convert all 150 paper-diary markets to code-reader collection, starting next year in 14 markets where Nielsen is already recruiting households and installing the new devices. Nielsen said in September that sample sizes will be increased under the new plan.
Zackon’s presentation was given in the context of a half day of presentations and talks organized by the CRE on the subject of “Big Data,” the catchall phrase that refers to the mountains of data available to television programmers, advertisers and marketers today, and how to organize and exploit it.
The centerpiece of the event was the presentation and release of a wide-ranging report commissioned by the CRE — a 37-page document styled as a “primer” on Big Data and “intended to help orient advertising and media executives around the opportunities and challenges that present themselves as we go from megabytes of data on PCs to petabytes of data in the cloud,” the CRE says.