Nielsen Unveils Plan To Model About Half Its National TV Ratings Beginning This Fall, Some Clients Surprised And Befuddled
- by Joe Mandese @mp_joemandese, September 4, 2014
In what likely is
the biggest methodological change in the way it measures national TV viewers since it introduced people meters in 1987, Nielsen Wednesday began informing clients of a new twist in its sample expansion
plan that will effectively utilize mathematical modeling to estimate viewing for about half of its sample.
While an email sent to clients Thursday implied that details of the plan were “previously discussed,” some clients were surprised -- and befuddled -- by what they said was the new twist, which Nielsen described as an enhancement of its “viewer assignment methodology.”
According to clients, and confirmed by a Nielsen spokesperson, that enhancement will essentially use mathematical modeling to double the size of its persons viewing and demographic estimates. Nielsen will do this by attributing data from its actual national people meter sample to its household tuning data generated by TV “set meters” in its local TV markets.
advertisement
advertisement
“Nielsen will apply a demographic assignment methodology informed by [national people meter] viewing behavior to report persons level ratings,” the spokesperson explained, adding: “We've been working on this with the industry for nearly two years.”
Nielsen said the new methods would begin contributing to its national TV ratings estimates effective Sept. 29 -- just after the start of the new season -- and that it would release data on the impact of those methodological changes to a client committee and to industry watchdog the Media Rating Council in the first quarter of 2015.
It was not clear at presstime whether the MRC has approved the changes, and whether they will impact its accreditation of Nielsen’s national TV ratings service, but it appears that these changes will be ongoing for some time as Nielsen integrates different elements of its sample expansion plan, which also includes the addition of some actual people meter households in its sample.
Nielsen’s national people meter sample currently is about 25,000 households, and Nielsen said it will add 2,200, but did not say when that expansion would be completed. It also said the addition of those homes, and the implementation of its new methodology would double the “effective sample size” of its national ratings panel, which implies nearly half of the effect will be coming from mathematical modeling.
One long-time Nielsen customer described Nielsen’s statement that it will double the effective sample size of its national TV ratings as misleading, because some customers may infer that it will reduce Nielsen’s sample error by half. But he said that would require Nielsen to quadruple the size of its sample, not double it.
Nielsen said it was making the changes to “address concerns over fragmentation in the television landscape by increasing ratings stability.” Another likely motive is that it will enable Nielsen to take on new customers whose audience might have been too small to be measured with statistical accuracy with via its smaller sample.
In a separate announcement made Wednesday, Nielsen also announced details of a local TV ratings sample expansion.
“As previously communicated in May, Nielsen plans to increase the sample in five Local People Meter (LPM) markets over the next several months,” the Nielsen spokesperson said, adding: “The five markets are Dallas, Denver, Houston, Miami and Washington DC. These LPM sample expansions will leverage the new Scrolling Text People Meter, which is the latest Nielsen Global Television Metering hardware platform. The first wave of replicates will begin to contribute to these local markets' samples with the data of September 29, reported on September 30. We plan to provide dates for the remaining replicates at a later point in time.”(Image provided courtesy of Shutterstock)
Were I a Nielsen client, the only way I would accept this plan would be if it was "validated" by taking the demo modelling approach against the peoplemeter panel as it now exists and seeing whether the resulting projections---- on a show by show/ telecast by telecast basis---- jibed with various subsets of the same panel. In other words, if you take a random 50% of the panel as now constituted, and project its findings to the entire panel, using the modelling approach, then compare these results to the actual findings of the full panel, how close do you come?
I wrote a series of blogposts last year predicting that Nielsen was going to go for fusion in a big way like this. The first of those posts: http://www.mediabizbloggers.com/bill-harvey/When-Will-It-Be-Time-to-Change-the-TV-Currency--Bill-Harvey.html - and I agree with Ed that the onus is on the purveyor of fusion to show that the actual specific sex/age program airing level data the client uses is well predicted by its fusion counterpart.
Pretty sure this is not accurate... The announcement said they're expanding the nat'l sample by 2200 homes in markets 56+, starting this Fall. That will help the fragmentation issue some, though not nearly as much as is needed. As I understand it the modeling is specific to the local sample, in set metered homes that do not have people-meters (markets 26-56), where they need to model demos from the HH-only data (can't speak to the reliability & impact on local data, but it is not half the national sample.)
Permit me to join the conversation.
David (Giles), You need to speak with Colleen Fahey Rush and your other distinguished Viacom research colleagues. As far as I can determine, Mr. Mandese has done his homework with Nielsen again. (For all we know, with news like this, Nielsen may have sent the media a Press Release to meet certain responsibilities it is know to have.)
Bill (Harvey), I think we need to be careful about technical language when these plans are discussed and evaluated. Is Nielsen really proposing "fusion" or simply sophisticated "modeling" as a solution to the stability/reliabity issue where small media entities are concerned? Perhaps this is a distinction without a difference. However, now more than ever, precision in research language is of the essence.
Finally, Ed Papazian has continued to demonstrate the priceless value of knowledge, experience and wisdom. On its face, it is hard, if not impossible, to see how this Nielsen plan constitutes an enhancement of Nielsen's National Service (NTI). If "half" of the demographic data are modeled, then -- IN REALITY -- "all" of the aggregate Nielsen GTAM (formerly NPM) data are MODELED. Once you add saltwater to freshwater, on average, the water is brackish -- and no longer fresh or salt-free. So too with TV viewing data. In 1963, Congressional Hearings (Congressman Oren Harris) were held when a Nielsen customer fabricated some tuning/viewing records. The Harris Hearings led to a number of improvements in the way ratings data are "produced" and "policed." Unfortunately, Congress needs to be about other matters at this point in the nation's history. However, if the Media Rating Council ( a byproduct of the Harris Hearings) was ever crucial to preserving the integrity and quality of media measurement, then now is the moment. Mr. Papazian's methodological research proposal must be vetted at once. Whether it becomes the ultimate test of validity, reliability and utility matters not. What matters is that tests comparable in strength to the "Coincidental" that was used in the 1987 People Meter Review conducted by Statistical Research, Inc in concert with the entire industry be conducted now. When all is said and done, the last thing the TV medium needs is a research travesty and a business abomination. A minor reduction in statistical volatility is no excuse for potentially a quantum loss in statistical validity.
[Important Reference Document for 1963 Harris Hearings: "Audience Ratings" by Hugh Malcolm Beville, Jr. - http://books.google.com/books?id=sALywz73eB4C&printsec=frontcover&source=gbs_ge_summary_r&cad=0#v=onepage&q&f=false]
Thanks. I meant more about timing -- to clarify and correct myself, they do plan to model those set meters, but not for a year, and needs to go through a lot of scrutiny & industry approval beforehand. Fully get the point about the dangers of the black box, The idea of telecast by telecast rigor is a great one. Author clarified here: http://www.mediapost.com/publications/article/233628/nielsen-clarifies-expansion-plan-says-modeling-wo.html?utm_source=newsletter&utm_medium=email&utm_content=headline&utm_campaign=75870
Nielsen says it will begin implementing the viewer assignment methodology this season, but won't include it in official ratings estimates until clients have had a chance to review the impact data and okay it. The modeling will not affect the sample, but the addition of the new national people meter households will begin impacting it on Sept. 29th.