As I continue to dig deeper into the advertising world, I’m fascinated by the rich history of advertising. Along with that comes an understanding of how past media habits influence today’s high-targeted, data-driven, digital media. The beloved television has a rich history in the heart of American folklore. But there’s an ugly truth behind how TV viewership is measured and priced to affect advertisers who want to educate an audience about their offering.
TV came into its own as the premier form of entertainment in the early 1950s, and quickly usurped radio from this throne of media dominance. With its wide broadcasting ability, and the visual impact it had that radio could not deliver, advertisers flocked to the new media.
But they had a problem. How were marketers to know how many people watched a particular TV show? Was it worth it for an marketer to pay for a spot in the show based on viewership?
Enter Nielsen — the then privately held company that still creates proprietary metrics on viewership. Nielsen started in 1923 and created a national metric to estimate radio audiences. With the emergence of TV, their metric was applied to television audiences as well. Over the past 70+ years, the $70 billion TV advertising industry has continued to rely on Nielsen ratings to determine fair marketing pricing for television commercials.
How TV Ratings are Measured
At the end of 2016, Nielsen estimated that there were 118.4 million TV households in the US. (We’ll focus just on the US market from here on out.) Of those households, Nielsen has created a “panel” of 40,300 households who steadfastly report what they wanted on a consistent and accurate basis. When you hear or read references to panel-based programming, that’s a reference to what the 40,300 household panel is reporting they are watching. Using the panel’s viewing history, the extrapolate the findings to the full 118.4 million TV households.
Significant Sample Size?
With just 40,300 of the 118,400,000 represented, that’s a 0.03% sample size. Is that enough? Nielsen certainly thinks so, but I wasn’t so sure so I turned to The Google to see what other, more statistically seasoned, people think.
Turns out that sample sizes can be quite small and still extrapolate to big audiences. Most recommended a sample size of 10% until the sample size reaches 1000. Once you sample more than 1000 then the precision of accuracy doesn’t increase significantly as you increase the size of your sample.
Using a sample size calculator from surveysystem.com, I found that for our 118.4 million TV households, we could obtain a 99% +/- 1% confidence level by surveying just 16,639 households.
Huh. I’m not a statistician, so that was surprising to me. Nielsen surveys almost 2.5 times the needed number, so we have to concluded that their sample size yields a statistically accurate representation of the viewing habits of all 118.4 million TV households.
These metrics do not include any kind of engagement factor. The TV could literally be on and with no one in the room. Or the viewer is focused on scrolling through Facebook as the TV drones on in the background. In advertising we refer to Reach — how many people were exposed to the ad. Reach doesn’t consider engagement.
Demographic Rating Projections
Nielsen knows the demographics of members of each of its sample households — sex, age, ethnicity, location. It can thus extrapolate those demographics to the whole population (of TV households). They also correlate demographics with the day of week, time of day, and show genre to create a matrix of demographic viewership. With matrix in hand, the demographic viewership of future shows can be projected based on show genre, day and time of broadcast, and other factors such as the market share the broadcaster currently enjoys compared to competing broadcasters.
With all that math, if a marketer wants to show their ad to a particular demographic then a TV advertiser can show what percentage of their demographic should tune in. Let’s say the marketer wants to reach women between the ages of 18 and 49. A viewership of a particular show and programming slot may include 30% of their demographic. If they air one commercial during that time then their GRP (gross ratings point) would be 30. GRP is a primary measure of reach in that industry. If they air the commercial twice, then their GRP would be 2 X 30 = 60; three airings and their GRP increases to 90.
With Broadcast TV in Decline, Is GRP Dying?
Below is a chart published a few months ago by AdAge. It shows how the 200 leading national advertisers are spending their money.
As you can see, TV advertising is decreasing. Does that mean that GRP buying is dying? No. It’s as strong as ever. Traditional TV buying and selling is entrenched in the industry.
As TV buyers have sought more targeted advertising online, metrics like Nielsen ratings and GRP are creeping into the digital advertising world. Social media is beginning to cater to traditional marketers. Both Facebook and Snapchat incorporate Nielsen ratings alongside their data-driven metrics. Facebook uses Nielsen ratings to create a buying program called TRP (“total rating point”). GRP is also used by Hulu and other streaming services owned by traditional media companies.
Mathematically Accurate, But…
Neilsen sampling, GRP projections…that’s a lot of projection of projections. Mathematically it may work. But personally, once we start estimating from estimates derived from estimates, I get suspicious.
But it is what it is. These metrics are so ingrained in the traditional advertising world that it will be a struggle to dislodge them.