Last week, this column covered the concept of response rates for Nielsen radio ratings and how to understand them in the context of the diary service. Since July is a great time for radio people to be reading about basic survey concepts (is there ever an ideal time?), let’s move on to PPM which uses an entirely different metric to calculate response rate.
As anyone in a PPM market knows, the service utilizes a panel to create estimates. Unlike the “one and done” diary service (which is not truly “one and done” but we’ll save that for another time), panelists can stay in the PPM panel for up to two years (technically 26 months, but why quibble?). While I haven’t seen an updated figure in recent years, the average tenure of a panelist used to be just short of a year.
Measuring response rate in a panel is not a straightforward exercise, but then again, the diary service response rate is convoluted, too. PPM adds even more twists.
First, Nielsen must identify potential panelist homes. The sample frame is the same as the diary service, that is, all addresses in the metro minus dorms, military barracks, prisons, businesses, etc. Phone numbers are added to the addresses when available.
Once Nielsen has all the information, they pull a “pool” of potential PPM households and from this pool, households are selected for something called “sample prep”. This is where it gets interesting, which depends greatly upon your personal definition of “interesting”.
In the sample prep process, Nielsen Audio designates each household as either a “Basic” or an “Alternate”. Getting Basics into the sample is the key for reported response rates.
For example, if the vendor can add a phone number to an address in the PPM sample frame, a Nielsen interviewer will call and ask for demographic information. If the Basic household refuses, that household stays in the sample pool. If an Alternative refuses, see you later. By the way, Houston is handled differently from the other 47 PPM markets.
When PPM measurement began nearly 20 years ago, Arbitron reported a true response rate with no Basics or Alternates. Although I may be off just a bit as I was heavily involved in the diary service at the time, the response rate metric was essentially the average daily intab divided by all the estimated people that Arbitron had tried to reach to be in the panel. The problem was that response rate was a very important statistic for the Media Rating Council, and single-digit response rates didn’t help in the quest for MRC accreditation for the PPM service.
There was another option which Nielsen TV had used for some time, a concept called the Sample Performance Indicator. The MRC had blessed SPI and the suggestion was made that perhaps Arbitron should consider using SPI instead of a true response rate.
Here’s how SPI works at a very high level: Take the number of persons from Basic households and divide by the number of persons in the panel for that month. If it were only that simple.
There are many ways you can make your head spin, some of which are illegal. If you’d like to try a legal one, go to a Nielsen E-book and pick a PPM market. Click on “Market Summary”, then “Population Estimates & Sample Summary” and then “Sample Summary”. Now scroll down to the heading that says, “Sample Performance Indicator Results”. The first metric, the “Unified Persons Sample Performance Indicator” is straightforward. You’ll see an average daily and total monthly Unified Persons SPI (in bold print). It may be in single digits or somewhere in the teens.
Further down the page is where the head-spinning part resides. In the spirit of disclosure and probably MRC requirements, you will find all the components that comprise the SPI metric. Unfortunately, you probably won’t understand how Nielsen calculated these numbers even if you read all 31 footnotes (hello Baskin-Robbins!) that explain how Nielsen arrives at the results in the tables. And don’t ask your Nielsen rep because he or she won’t know either. For that matter, I probably can’t explain all of them, but I’d get close.
I’ll return to the same question as last week, which is “Does this matter?”. Yes and no. No, if the panelists generally behave like the people who suggested various anatomical places that the Nielsen interviewer could put a meter. If the panelists have very different behaviors than the refuseniks, then yes.
Survey response rates in today’s world are lousy. It’s not just Nielsen, but everyone, public and private. Government agencies spend billions on surveys and get better response rates than the private sector due to more time and money, but even those surveys are having problems.
For example, the Consumer Expenditure personal interview survey conducted by the US Bureau of Labor Statistics had a response rate of 67.9% in April 2015. In March 2025, the response rate was 39.1%. That’s a decline of 42% over ten years.
Nielsen could probably do a bit better on the SPI, but the cost and effort wouldn’t be worth it. Most of us can offer ideas of better ways for Nielsen to invest in the audio service (or not invest and lower their pricing). But now you have an understanding of response rate and what it means.
Let’s meet again next week.
