The tectonic plates of media measurement are moving, as digital media behaviors upend decades-old approaches to measuring the world of TV audiences and advertising, giving impetus to the rise of the alt measurement movement, one of the biggest stories in our industry today.
I, for one, can’t wait to see sex- and age-delineated gross rating points take a back seat to more granular audience metrics for TV ad targeting and measurement.
It’s great to see Nielsen move forward with its next generation Nielsen One product. And it’s refreshing to see other alternative measurement suppliers doing deals with media owners and ad buyers to complement or supplant TV advertising’s historic primary and secondary currencies, as audiences continue to fragment across broadcast, cable, satellite and, most importantly, fast-growing streaming channels.
Connected, smart TVs are helping power many of these new measurements, and the availability of real-time viewing data from many tens of millions of viewers is a huge boost to the world. However, it is important that we don’t let the scale of this data blind us from some of its critical biases, particularly when we are talking about how it can help us better value and allocate the $80 billion or so of advertising spend in the U.S. this year on television and streaming properties.
advertisement
advertisement
The use of these smart TV datasets in TV advertising currencies is game-changing, but will need to be balanced and hybridized with significant amounts of viewing data from households that don’t have fixed broadband Internet at home -- the “return path” for the collection of smart TV data.
Pew research informs us that Black, Hispanic and lower-income households are significantly less likely to have fixed broadband in their homes. Many of them don’t have cable or satellite TV either, relying on broadcast TV signals, which have actually become much more robust over the past 10 years with the introduction of digital antennas prebuilt into new TVs, and many dozens of new multicast networks that offer programming free to air.
Thus, not only do so many non-Internet households get their local news and sports games for free over antennas, but they are also getting alternative language programming as well. As our industry is finally paying attention to the problems of decades of under-investing in programming and advertising for Black, Hispanic and lower-income populations, it is critical that we don’t forget them with our embrace of alternative measurements.
This isn’t hard to solve. Nielsen has been balancing its panel for appropriate representation of Blacks, Hispanics and lower-income. So too can those who want to supplement or unseat their measurements.
What do you think?
Dave, while I agree about properly representing all segments of the population, statistical sample balancing is merely a bandaid and not by definition a cure for the problem. For example let's say that Nielsen's panel has only 5% of a particular group when it should be 10%. Changing the projection value of each panel member in the group so it counts as if these folks were 10% of the total, not 5%, sounds fine, on paper, but the underlying assumption is that whom ever you got from that group represented a fair crosssection of the segment----only you recruited too few of them. But what if your sample of this particular population segment not only was too small but those who cooperated, in aggregate, were not a good representation of the whole? Sample balancing may give you a more correct quantitative audience projection---but not an accurate one.
A second point concerns smart sets and set-top-boxes and what they actually tell us. These millions of "respondents" do not give us any proper idea of viewing---how many or what kind of persons are watching. Their only value is to supply a huge sample of set usge data which will then be modified by projecting viewer-per-set findings from Nielsen's comparatively tiny people meter panel on to them. So if a particular TV show is found to reach 25,000 smart sets TV sets per commercial, its viewer projections---by sex, age, income, etc.---- may be based on only 100-150 people meter panel members whose "viewing" is ascribed to the 25,000 sets that were tuned in. What we are getting is primarily a large sample for device usage---not viewing. That's better than a small panel for both---but it's not the great breakthrough that some people are claiming. And, of course, the plan does not call for an attentiveness measurement---meaning that many of the commercial 'viewers" that will be reported are phantoms---they never saw the commercials.
Ed, thanks for the comment. I totally agree that using token data sets with sample balancing will NOT solve the problem here. It is critical that any new TV & streaming ad measurement approaches have significant data captured on the viewing behaviors of large pools of underrepresented broadcast and non- broadband homes, with true independence and verifiability.
Ed raises good points. But I'd like to expand on Ed's comment about the 5% which should be 10%. Yes you can raise that projection factor to get that individual vector closer to reality. However, doubling that cohort's projection factor would also double the results for every other vector (demo group). Also, the more granular you make your projection variables (e.g. Hispanic university educated males over 55 who have purchased a Ford in the past six months and who are married but the kids have left home) the more volatile and less accurate the overal results will be.