Derrick Gray: Radio's New Way to Count

Tara Sands (Voiceover)  0:02  
The Sound Off podcast starts now.

Matt Cundill  0:09  
For decades, radio ratings have been compiled by asking people to remember what they listen to and write it down. But as audiences scattered across FM to apps over smart speakers and connected cars. That system has become harder to trust and harder to sustain. Numeris in Canada is now replacing its online radio diary in 22 Canadian markets with Radio Local Plus. It's a hybrid system that combines streaming information with online and telephone surveys to estimate total listening, including over-the-air radio, the ratings will also arrive four times a year instead of twice, giving programmers and advertisers a more current picture of the audience. My guest today is Derek Ray. He is the chief research and operations officer at Numeris, and he's going to explain how all this works, what counts as listening,

Derrick Gray  1:00  
what the testing uncovered and why the first results may look different. Here's the disclosure piece and logic. The audience analysis and software arm of Numeris has been a longtime supporter of this podcast, and we're excited to welcome them back for another year and have them on the show when the big stuff is unveiled, Like now, and now Derek Gray joins me from Toronto. So I am the Chief Research and Operations Officer at Numeris. So I oversee both the research team, which includes data science, methodology, research operations, as well as the day-to-day operations. So the field team, the data production team, all all that.

Matt Cundill  1:45  
How long has Radio Local Plus been in the making?

Derrick Gray  1:48  
I would say we first started thinking about it or talking about it in the second half of 2024, 2025. We did a proof of concept using two different suppliers. We tested three markets: Kelowna, Quebec City, Kitchener, Waterloo, Cambridge. We took that through the committees. Committees were quite excited. We sort of focused in on one particular vendor. We worked with them. We did some extensive testing in the spring of 2026 against all 22 markets that are currently measured. Well, we're currently measured by the online radio diary, and we did the analysis, and we went from there. And actually, I should note between the POC and the parallel run, there was a few methodological enhancements that either the committees brought forth, or we brought forth, or our auditors, we used an audit both from a certification process. They did a certification certification audit in the fall of 2025, and they did a technical audit in the spring of 2026. So we did all those methodological enhancements between the POC and the parallel run, and we're still working on it. It's ready to launch, but like anything, it's living, it's breathing. There's going to be further enhancements as it evolves, and it plays a big part in the audio strategy, which of course will also cause a little bit more evolution.

Matt Cundill  3:14  
Can you share some things that happened in the testing of those three markets that you know that you together as a group said, "Well, we didn't think of that, or we didn't see that coming.

Derrick Gray  3:23  
I think the big thing was reach, and and it's funny because in hindsight, it was like, yeah, of course. Why didn't we think of that? Reach changed dramatically with the the hybrid methodology of Radio Local Plus. It's passive measurement, right? With diaries, there there is a tendency you're relying on people's memory, so they they remember the stations they really like, the stations they listen to all the time, the stations they listen to for long periods of time. But it's human nature, right? You sometimes forget those stations you you listen to for short periods. So we saw that also with PPM, so reach went up. We were a little shocked at first, going, "Whoa, where where is this coming from? But as we rationalized it, it made a little bit more sense. The original version of the product didn't have deduplication across stations either. So when we were looking at total market reach, that of course told us, "Hey, there's something crazy going on here because you had reach over 100 in some cases, and as we dug in, we recognized there wasn't a mechanism in the system to do that deduplication across stations. So that that was a major enhancement that we worked on between 2025 and 2026, getting that right. I guess the other thing that was a little surprising was the way that the data is presented. So it comes out as aggregate data, summary data. Reach frequency, of course, is something that's very important in the large radio markets, and reach frequency works on proprietary algorithms. That the third-party processors have in their software, but they need respondent-level data for those mechanisms or those algorithms to work. So it was also okay. How are we going to go about this? The underlying methodology was sort of created for smaller markets that tend to trade on spots as opposed to reach frequencies. So that was something Numeris had to really step into and work to figure out a way of converting that data into virtual respondents per se, so that the reach frequency curves would continue to work and the market can continue to trade the way that they always have.

Matt Cundill  5:36  
So there are some people who are approaching a fall ratings period, and it's quite possible they took the summer off and didn't even know that this was coming because maybe they started in new jobs, went to a different market, and they're just waking up to this now. So, how does this new system work?

Derrick Gray  5:54  
It's it's pretty technical, but let me give you the overall understanding of what's happening. So, Radio Local Plus is a hybrid approach, and by hybrid, it means it's a combination of various data sources. The baseline is streaming data, so the streaming data comes in from the platforms. That is census-like data; it's from all the devices that are streaming the the radio station, of course, it's device level, so that doesn't give you any information about the demographics, who's listening, co-listening, that sort of thing, or even location. Then there is two sets of surveys that are conducted: one which is online, and the other is over the phone. And I think the easiest way to explain these surveys is they're like a day after recall. A lot of people are familiar with that. So you sort of speak to the individual and you say, "Okay, so it's it's Wednesday today. Tell me yesterday what did you listen to? What would you typically listen to on a Tuesday? And we get that information to understand the demographics across the various stations, the co-listening that may be happening across the various stations. But the other important part is streaming is only a portion of total radio listening. It's growing, but today it's roughly around 10% Some stations it's higher, some demographics it's higher. But on a two plus five A to one A level, it's about 10% You need to understand what is that proportion that that streaming actually represents, and as I said to you, it varies dramatically by station, by time of day, by demographic. So that day after recall like survey helps us understand what those various proportions are throughout the day for each station across the various demographics. So all that information feeds into a statistical model, and then that statistical model is applied against the streaming data to create an over like total radio being streaming plus over the air audience estimate. So that that in a nutshell is how it works, and I I hope that makes sense.

Matt Cundill  8:08  
What happens if somebody is listening to an Ottawa radio station and they happen to be doing it from Calgary? What sort of counting is involved there?

Derrick Gray  8:17  
So when we do the surveys, naturally the surveys are geo-targeted, so Ottawa we're only surveying individuals in the audio Ottawa area, and at that time, the streaming data has IP addresses attached to it, which we can do geolocating on. So when I produce a number, or sorry, when Numeris is producing a number for Ottawa, it's going to be in market tuning, so any tuning from Calgary to an Ottawa station will not be included, and that's not that dissimilar to how it is today or how it was. Sorry, keep talking presence tense because this is new. Not that dissimilar to how it was in in ORD, right? We only measured the tuning in Ottawa to the Ottawa stations.

Matt Cundill  9:04  
So, in podcasting, we only get two pieces of information. You know, when this show is downloaded, we're going to know your location through your IP address, and we're going to know the user agent that you use to listen to this particular podcast, this does not count YouTube, by the way. And in that process, there's going to be some filtering. There's going to be stuff taken out that is just what would be considered a fraudulent or garbage download, where there's no real person behind it. Do you use the term filtering quite often? You know, let's say we're going to take the Calgary list out. That would be filtered.

Derrick Gray  9:38  
Yeah, I mean, there's a whole set of filtering that's going on, and you're right. The first thing is removing the the streaming consumption that's occurring out of market. You hit the nail on the head with terms of bots. There's various mechanisms that are in there to identify bot tuning, sort of fraudulent tuning that's not real individuals. We're going to be working really closely to look at the IAB standards around that to make sure that our protocols match the IAB standards perfectly, and perhaps even more. Right in many cases, we we typically do things beyond what the standards are. There's also some filtering. I would call a little bit more like capping when you have data that's coming from servers, and that you don't know if there's an individual there, and and frequently you see that for extremely long tuning sessions where it's just the player is just going for 24 hours. I mean, when you look at the raw data, you see cases where it can run for like a week using information we have from PPM. We looked at some of the information in ORD. We got to understand what the tuning distributions were as well. So there is a capping algorithm that comes into play that removes excessively long sessions. So there is a fair degree of data prep that goes into it, or I actually like to call it data hygiene as opposed to filtering because it's really cleaning things up and making things fit for purpose, and the purpose being measurement.

Matt Cundill  11:11  
You mentioned a little bit about attribution when it comes to streaming. I know in podcasting we don't get demographics of male or female. You can if you you know go in through Apple or if you go in through you know through Spotify to to get this stuff, but because there's no you know gender age attached to the to the streaming, does that present any form of of challenge or how do you handle that just a little bit differently?

Derrick Gray  11:36  
Yeah, so that that comes from the surveys. So the surveys we conduct form that source of truth of what the demographic profile is for each station across the various parts of the day. Now, some of the streaming data that comes into us has Google Analytics data attached to it, so there is some demographic data that comes into us, but that's a starting point. All of that demographic data is then calibrated and adjusted so it matches the source of truth, which are the surveys themselves. And there's a fair amount of confusion in some of the people that I spoke to about this, where they said, "Well, you know, what happens if one station has a Google Analytics but the other doesn't? Does that disadvantage the station that doesn't absolutely not because again the source of truth is the surveys and that's conducted against all the measured stations and irrespective of we have the demographic data from Google we do adjust it and align it to that source of truth of the surveys

Matt Cundill  12:36  
yeah it's one of the pieces of advice I give podcasters have a website and attach it to Google Analytics so you can get a better idea of who your audience is. But it's nice to know you're not going to penalize anybody who doesn't have it necessarily at their fingertips. You know, there's a lot of chatter because the same talk is going on in you know with Spotify, with YouTube, with podcasting in general. In in what counts as a listen. What counts as a play? What counts as a download?

Derrick Gray  13:04  
Yeah.

Matt Cundill  13:05  
So, from your perspective, what counts here?

Derrick Gray  13:07  
So, we we do have a minimum tuning threshold. Any statement that's below that minimum threshold is discarded. So, provided that you're reaching that minimum threshold, it does count as as a tune, similar to ORD, you needed five minutes of listening to win the quarter hour. We do also have that minimum threshold to win the quarter hour. Now, I I did mention earlier about Radio Local Plus being a foundational piece of the audio strategy. Now that we have the approval and we're rolling it out, the one thing that we're going to start looking at is moving towards an average minute audience, and also producing impression-based metrics. So, going from a 15-minute AQH down to a one-minute AMA, of course, we're going to have to revisit what counts as a listen. So that's something we'll be working with quite closely with the the governance committees, the the research advisory committee, as well as the audio executive committee this coming broadcast year.

Matt Cundill  14:08  
Spotify is trying to get into it at 30 seconds. Radio in the U.S. went from five minutes to three minutes. Podcasting's at one minute download. So with many metrics out there. Advertisers, I think, are going to want something a little bit closer in the future, and I'm really happy to see that you that you've kept this very flexible. So, how much patience would you advise radio programmers because they're the ones who will become the most impatient? How much patience should they allow for this for the system?

Derrick Gray  14:40  
Well, I I think the system that's rolling out this fall is is well developed for what it is for an average quarter hour audience. It's well developed. There will be less demographics; those qualitative demos that we had versus ORD. We're still working on getting. Better identification of in-home versus out-of-home listening. We are capturing all listening, but reporting correctly what's out of home, what's in a home. That's something we're working on. So I would say a little bit of patience because we are improving the product. This is not a one-and-done. This is going to be an evolution. I would hope that the industry would be patient with us and give us the feedback on what are the things that you need and what are the things you need us to focus on. But the evolution of the audio strategy that that is you know several year roadmap and sort of it's it's three phases. Phase one. This was part of phase one. This was very foundational, and phase three is where we're hoping to have a national reach for radio, and there's a lot of impatience around that, and I can understand why people really miss that. It speaks extremely well to the value proposition of radio, but we're we're still a ways off from that. We've just started the journey. You know, stick with us, and we'll work with the industry, and we will get there.

Matt Cundill  16:09  
I'm excited because it's I think 10 years, maybe 1213, now. I've been sort of saying, how can we don't really count the people who are streaming, like, and how they behave, and and getting them counted into the ecosystem, I think this is a good way to count to have this, you know, this this hybrid. But what about advertisers? Are they up to speed on on on this? And did they have feedback?

Derrick Gray  16:35  
On our research advisory committee, we have agencies who are representatives of advertisers. There's no direct advertisers on on the committee itself, but the agency certainly gave their input based on their clients who are who are the advertisers. So I would say that there is input from both sides of the marketplace, both the buy side and the sell side. It's interesting you talk about counting of streaming. If I go back to the fall of 2024 when we first started talking about replacements for ORD, that was something that was very loud and clear in both the Research Advisory Committee as well as the Audio Executive Committee. Something that includes streaming. Streaming is the future. It needs to play a role in the solution and the replacement for ORD, so I was really happy to hear you say that because that that was an important starting point for us.

Matt Cundill  17:30  
I'll use an example. I use the iHeart app. I never thought I would use it for podcasting, but it's because I listened to it on radio. So if I were to be streaming my local station on the iHeart app, will that be detected, and would the radio station that I'm listening to get that information? That there's a there's a guy who's got a podcast called the Sound Off Podcast who is streaming on the iHeart app.

Derrick Gray  17:51  
Absolutely, absolutely. Provided that your local radio station is a participating station or subscriber to Numeris, they will be included. Obviously, in the markets, there are some stations that aren't clients of Numeris. You know, the smaller community stations and college stations and and whatnot. But if it's one of the commercial stations that subscribes to the Numeris data, they will be included.

Matt Cundill  18:13  
Was there any consultation with with other countries, say like Australia, to to see how they did it, or anyone in Europe, perhaps?

Derrick Gray  18:21  
We spoke to Australia. Their system is quite different. It's still heavily based on other forms. Streaming is just a small input that goes into it. It's not the primary baseline like our solution is. I'll tell you a little bit of a story. So, November of every year is the ASI conference advertising seminars, international. It's it's probably the biggest global conference when it comes to measurement, radio and television measurement. And we put a submission to talk about the POC, and it was accepted. And we were kind of thinking, "Wow, this is awesome. We're we're going to be the only ones talking about this. This is so forward. The panel I was on, there was several countries actually, and Australia being one of them that had some sort of hybrid approach. So many countries are moving towards hybrid for radio measurement. We are talking to them quite a bit. We're getting a lot of phone calls ourselves. We've been asked to go back to ASI and speak to it again this year. I would say what we're doing is very unique, so we're kind of forging the way ahead in this particular approach. But like anything, measurements a very small community, and we're all very connected. You know, both Alicia and I have worked across a few different countries and continents. We're both very well connected with all the other jigs around the world, and we do speak to them frequently and share ideas and get feedback and thoughts and and their experiences on things that they're doing.

Matt Cundill  19:55  
I think one of the harder things for anybody who does research or. You need to reach out to people to participate in this sort of thing. Is recruitment a lot of people stopped answering their their landlines? It's pretty easy to do do not disturb. We have an entire generation who don't even know how to answer the phone. What are some of the challenges with recruiting people to join the panel?

Derrick Gray  20:16  
The panel is is a little bit more straightforward. We actually the panels are actually outsourced, so they are with large panel companies, established panel companies. They're high quality panels. They follow SMR standards. So SMR is the global standard organization for market research. So they are really robust and well managed. Not a lot of issues around that. The phone surveys are a little more challenging, but we've had good success with that. But you're right, what you're talking about. I mean, that was one of the biggest challenges with online radio diary because not only do you need to get the people to answer the phone, then you need to get them to agree, then you send them the diary, and you need to get them to fill out the diary and send it back. And that was becoming very very challenging. A lot of people don't realize Canada actually has one of the lowest participation rates when it comes to surveys in the world. Our rates are just significantly lower than many many countries, and that really poses a challenge. So the volumes that we needed to contact for online radio diary was no longer sustainable. These volumes are a lot easier to achieve, and using access panels is is a big help.

Matt Cundill  21:30  
I'm surprised by that that the Canadians are just a little bit more reluctant to to you know to take the survey. I've always thought people in the Maritimes are quite excited to pick up the phone and have a chat, people in Quebec. I know Quebecers want to be counted, but once you hit the Ottawa River, it's a different set of results.

Derrick Gray  21:48  
Yeah, no, it is, but it's. I think a lot of it with Canada is we're a highly privacy centric society. Got it. If you think, I mean, if you go back to our privacy legislations in Canada, Pipita was one of the first, and a lot of people talk about GDPR in Europe, but quite frankly, GDPR was mirrored a lot on a lot of the Pipita principles. So Canadians are are a little more risk averse. Canadians are much more privacy centric, and it does make it quite quite difficult to recruit people here.

Matt Cundill  22:23  
This is not a question. This is a comment. But I'll let you comment after my comment, and that's my level of excitement for measuring four times a year and doing it on a quarterly basis. You know, getting a survey back twice a year never really told the story of your audience. It only told the story twice a year, but you know, with with some form of quarterly measurement, we can now see some of the activities that go on, you know, year round, including summer.

Derrick Gray  22:49  
Yeah, no, absolutely, and and and it's even better than that. The releases are going to coincide with the PPM releases, so all 22 plus six markets, the data will come out at the same time with the ORD. It was like a 48 week rolling average. This is going to be a 13 week rolling average, so it's going to be a lot more recent, greater recency in the data. It will give a better understanding of the impact of programming changes or behavior changes out there. And yeah, absolutely, four releases a year. That's that's great. We have the possibility to do it even more frequently, and we were told by the industry that four is what they can handle today. But that's not to say in the future we can't go even more frequent, which could be beneficial, perhaps.

Matt Cundill  23:35  
Through all this, Eric, like what's going to change and what's going to stay the same?

Derrick Gray  23:39  
So in terms of what's staying the same? We're launching with the same 22 markets that were in ORD, so that'll be the same. Hopefully, we'll add more to it with time. The same software and tools that the radio broadcasters or the agencies or so forth currently use will remain the same. We're launching with the same key radio metrics and the same ability to evaluate audiences and station performance will will hold true, and Numeris's role as the industry currency for local radio measurement will also hold true. Now improvements we spoke about the four data releases a year instead of two. I think the end users are going to see there's a lot more stability in the data, which will be different than ORD, and a very positive change, will have much better comparability across metrics and a more representative degree of measurement. Of course, the incorporation of the streaming data is something new. About I said about 10% today on average, but it's growing, right? And as it grows, this product's only going to get more precise and more accurate, and you know, of course, a much stronger innovation, or sorry, a much stronger foundation for future innovation than no RD could ever allow. So, to me, those are the big, the big differences.

Matt Cundill  24:59  
So, if we look ahead. One year, what does success look like when we have this conversation again next year?

Derrick Gray  25:05  
I think next year we will have further enhancements that will be out there in the market, and I hope everybody's going to be excited about them. I hope by next year we'll be close to moving towards average minute and audience and impression based data. That would be a huge success. If we move towards the average minute audience, we can start to compatibilize Radio Local Plus with the meter data in the six metros. We can start to look at combining a database and having a single data set out there, which I think would be a huge win for the industry. And I also think bringing on more markets beyond the 22 that we're going to be launching with would be a huge win as well. And we're committed to work with the committees and see which markets we can also bring on into the Numeris umbrella. We've heard loud and clear what a lot of the pain points are in the industry, and we do want to address them,

Matt Cundill  26:01  
hey Derek, thanks for taking the time to to do this, and we'll follow up again in one year.

Derrick Gray  26:06  
Yeah, sounds great, Matt. Great speaking to you.

Tara Sands (Voiceover)  26:13  
Another Sound Off Media Company podcast.