Networks, Advertisers Consider Set-Top Data For Audience Guarantees

A top CNBC researcher suggested Wednesday that set-top-box data can offer more stable measurement for upscale viewers, since it is drawn from a larger sample size than a Nielsen panel. The matter is of particular interest to CNBC as it looks to attract advertisers willing to pay a premium to reach wealthier viewers.

Sheryl Feldinger, a senior vice president at NBCUniversal, offered a snapshot of Nielsen measurement for CNBC that showed considerable volatility over a four-day period with the data coming from about 14 homes.

Comparatively, data captured from Rentrak -- which uses set-top-box (STB) data -- showed less volatility, as it was culled from the much larger group of nearly 16,000 households tuned to the network.

Rentrak says it has access to data from STBs in millions of homes served by cable, satellite and telco TV operators. And Feldinger said using the mass can lead to ratings that are "highly predictable."

By contrast, she said the small Nielsen sample can yield a scenario where "the behavior of (a) few can disproportionately affect ratings."



At a gathering of MPG's Collaborative Alliance, Feldinger also presented data from the non-NBCUniversal network BBC America, which showed a similar pattern to CNBC. Rentrak's STB data from about 3,000 homes had less dramatic swings than the Nielsen measurements, which came from three households.

Drilling down into the CNBC sweet spot, Feldinger used data for homes with incomes of $125,000-plus. During a five-day period in February, Nielsen figures for CNBC's "Closing Bell" produced daily indexes ranging from 56 to 227.

It is frustrating for CNBC that the 56 index represents a notably "downscale" home.

By comparison, the Rentrak daily indexes on "Closing Bell" then were less variable, stretching from 181 to 193 (the figure was 193 for two of the days).

Feldinger said that "according to Nielsen, the income profile for CNBC programs fluctuates dramatically from one day to the next ... most of us would probably agree that a lot of the same people watch 'Closing Bell' every day, and the income profile is pretty consistent from one day to the next."

Data for BBC America's "World News America" offered a similar pattern, with the daily indexes in the $125,000-plus homes from Nielsen stretched from 18 to 227 (in the Feb. 14-18 period). The 18 marks a notably "downscale" home. Rentrak's STB data showed a less variant 156 to 182 range.

The volatility in the Nielsen ratings can make it difficult for a network to predict ratings performance, which propel its sales process from pricing to guarantees.

It is unlikely, of course, that STB data will replace the Nielsen panel as the broad market currency -- at least in the short term -- at a CNBC or other network, but when negotiating deals for campaigns that are targeting an affluent segment, Feldinger said STB data "emerges as an attractive alternative."

Tom O'Brien, CNBC's chief revenue officer, said the network would be willing to engage in "conversations" about using STB data in deal-making, perhaps for some type of secondary guarantee.

Peter Sedlarcik, a senior vice president in research at MPG, joined Feldinger in making the presentation Thursday and said if it makes sense for both sides then "we absolutely want to push" for an alternate type of guarantee for a high-income audience and "make that part of the dynamic." MPG represents Fidelity, which frequently looks to hone in on particularly affluent audiences.

At the very least, Sedlarcik said monitoring the Rentrak data alongside the Nielsen ratings can help identify troubling fluctuations that may need to be addressed with a network.

Rentrak is not the only provider of STB data, with companies such as Kantar and TRA also offering it, but Feldinger indicated the data is more stable regardless of the provider.

She cited an example of a March Thursday where CNBC saw its Nielsen ratings drop 20% "for no apparent reason."

"We looked at set-top-box data from three different sources," she said. "Tuning on that day was basically the same as it was on the other Thursdays that month."

Pursuing the matter further, CNBC discovered that the 20% drop-off was due to a single viewer "who was out of tab" in the Nielsen panel, so his viewing went uncounted. "He wasn't necessarily not watching CNBC -- he just wasn't counted that day," she said.

The viability of STB data has been questioned partly because it may not take into account when a set-top box remains on, but there is no viewing going on. Rentrak, for its part, says it has algorithms to account for the potential issue, and Feldinger indicated she was satisfied with its methods.

3 comments about "Networks, Advertisers Consider Set-Top Data For Audience Guarantees".
Check to receive email when comments are posted.
  1. Darrin Stephens from McMann & Tate, October 6, 2011 at 9:17 a.m.

    Just because set top box data is more stable, does not make it better.

    If the data is not indicative of who's really watching, then it's potentially more damaging for a brand to use it than using unstable data that over the long haul is more accurate.

  2. Charlene Weisler from Writer, Media Consultant:, October 6, 2011 at 11:16 a.m.

    Hi Darrin,
    I respectfully disagree.
    Just as the current measurement system requires viewers to continuously indicate that they are viewing by pressing a button, STB data has the capability to indicate viewing vs tuning by activity on the remote. And this may become even more of a common indicator as we evolve into connected TVs.

    With larger samples, passive data collection and the ability to merge with consumer and demo databases, STB data is a logical solution / extension to the current TV measurement.

  3. John Grono from GAP Research, October 6, 2011 at 4:59 p.m.

    Just a few thoughts:

    * STB data from large opt-in sub-sets of households will clearly produce more robust and stable tuning data.

    * We know that tuning data is more stable than viewing data, and we also know that tuning is not equal to viewing.

    * Here in Australia we use TV remote control usage as an adjunct that there is definitely someone (we don't know who) in the household definitely viewing.

    * I note in the example cited that because a single person went "out-of-tab" on a single day and the ratings went down by 20% implies that the active sample size for that programme went from n=5 to n=4 (assuming weights were relatively similar). I think the question to ask is what was the total sample size? If it was from the national sample - which I think is around 25,000 homes which would be around 60,000 people (and there is insufficient detail provided in the case cited to be sure) then I think that whether it was 4 or 5 becomes a moot point - we'd be arguing over just how miniscule the audience actually was.

Next story loading loading..