Last week, I covered the motivation behind the new measurement system, Radio Local+, introduced by Numeris in non-PPM Canadian markets. This week, let’s dig into the details of how it works in a post-diary world with input from Dr. Derrick Gray, the chief of research and operations.
Unlike Nielsen, Radio Local+ is a hybrid system, meaning data comes from multiple sources. Nielsen is either PPM or diary. Yes, the Nielsen diary service now includes both paper and MSurvey (online), but those are the same instrument used in different modalities.
Radio Local+ incorporates data from StatsRadio, based in Quebec City. The company measures streaming from log files, much like Triton does here. However, StatsRadio goes a step further to create overall listening estimates.
How the Hybrid Model Works
StatsRadio conducts phone surveys in the relevant markets to collect “day-after” audio listening. Per Derrick, the survey design is solid, and asking for one day of listening is likely to be more accurate as well as a much lighter burden for the respondent. Some of the respondents are drawn from panels, and some are random dialing.
Using phone interviews to understand the ratio of streaming to over-the-air listening allows you to gross up the streaming data to get to a total listening estimate. Well, sort of. Through some modeling and adjustments, you have a summary-level database of audio listening by station. The data has daypart-by-daypart data for various demos, just as we’re used to seeing today. But there’s a fly in the ointment.
To do reach and frequency calculations for ad schedules, you need respondent-level data. Meters and diaries supply that information because these are collected at the individual level.
Virtual Respondents and the Role of Modeling
That brings us to the concept of “virtual respondents” or a “synthetic population.” Derrick says Numeris uses Bayesian probabilistic models to assign listening to virtual individuals. Once completed, the aggregated respondent-level data will match the summary data. But that’s not the end of it.
Numeris adds more “special sauce” to the mix based on the PPM data in the five largest Canadian markets, mainly for co-listening. This helps to improve the model even more.
Derrick went on to explain, “The role of modelling is not to replace measurement, but to enhance it. By extrapolating listening from Radio Local+ to virtual respondents, we can provide better insight into reach and frequency, resulting in better value for the radio industry.”
Why Costs Drive the Change
If I lost you at “Bayesian,” let’s consider what’s going on here. First, as I wrote last week, keep in mind that Numeris is a JIC, not a private company like Nielsen. Their “shareholders” are broadcasters, agencies, advertisers, and a streamer or two, in other words, the entities that use the data. Costs had to be reduced, and Radio Local+ does just that. Less data collection, less respondent burden, and more statistical manipulation.
If you think this isn’t happening today in the US, just look at how video is measured. Nielsen and their competitors use high-level statistical machinations. The local TV diary has been gone for eight years.
As part of the larger audio strategy, Derrick also explained how Numeris expects to stabilize PPM estimates, which the organization is testing now. This gets statistical, but let’s just say that if you make a couple of reasonable assumptions, at least for statisticians, you can create a “synthetic” population of PPM panelists. Put as simply as possible, if you take the listening from PPMs with the variance of the streaming data, you can generate estimates that combine the best of both worlds. You’ll get the minute-by-minute PPM data with the stability of the census-level streaming information.
This is being tested in one market to see how it works in practice. The concept makes sense.
Lessons From Arbitron
When I was at Arbitron, any sort of modeling would raise hackles with the radio clients. “You’re making up data” was the typical refrain, and to some extent, that was true. It wasn’t as if we were throwing darts or rolling dice, although we were accused of that.
In today’s very complicated and cluttered media environment, it’s a matter of taking available information and using it in ways to effectively measure audiences while keeping costs down. The days of measuring everything via surveys (meters, diaries, phone calls, in-person) are numbered. We can’t get large enough sample sizes at a reasonable price.
Yet we still have the radio diary in this country. Numeris debuted its online diary in 2016 and killed it this year at approximately the same time Nielsen Audio decided the time was right to introduce an online diary. What’s wrong with this picture?
A Look Ahead
Numeris is showing us one path to the future of audio measurement. If Nielsen isn’t already working on a similar system or talking to Numeris and other JICs and companies, I’d be shocked. It’s far cheaper to model a limited number of datasets than to collect data everywhere, which means lower costs to broadcasters and a future for Nielsen. The first reports for Radio Local+ will be released in December, and the results should be interesting.
To see the future, we can paraphrase Horace Greeley: “Go North, young man.”
Let’s meet again next week.
Barrett Media produces daily content on the music, news, and sports media industries. Sign up for our newsletters to stay updated and get the latest information right in your inbox.
