Radio Ratings and those Shifting Sands

In radio, few numbers carry more weight than the latest radio ratings. In Australia, the numbers are the GfK Radio 360 survey results. Stations live and die by them, Content Directors change line-ups, decide to renew talent contracts, or not, reallocate marketing budgets and the commercial team adjust advertising rates — all off the back of share¹, cume² and average audience³ figures. Yet if you don’t understand how statistics work, those all-important numbers could be shifting sands.

Shifting sands is a metaphor for an unstable, unreliable foundation on which important decisions are built. In the context of Australian radio ratings, and indeed other developed radio markets like the UK, USA, Canada and New Zealand for example, it refers to making major programming, talent, marketing, or commercial decisions based on a single survey result, or misinterpreted metrics (treating share as overall popularity, cume as loyalty, or one survey’s movement as definitive proof of success or failure).

I had 115 surveys over 15 years while I was at ARN and many others in the preceding years working at Coast Rock FM, PMFM, Austereo, Village Roadshow, GWR and Bauer Media UK and what was interesting to me was how many people placed far too much emphasis on a single survey result. Of course, there are the journalists who need a headline and grab one after each survey release, then there’s the survey spin by the Content Director and then there’s the strategic point of view which is usually not shared with journalists or even staff in some cases, and that should be looking at trends.

What many forget is that the numbers are estimates, they are not actual headcount, and the numbers are subject to statistical variations, regression to the mean as it’s called in statistics, external events, survey tactics, marketing and limitations of the methodology. What looks like solid ground in one survey can shift in the next survey, often without any real change in audience behaviour. The result can be expensive, reactive decisions built on shifting sands rather than sustainable trends. Generally, small movements in share or cume, especially in narrower demographics or dayparts frequently fall within normal statistical variations. Regression to the mean is where extreme results in one survey (very high or low) tend to moderate in subsequent surveys. When I was CCO at ARN, I would expect any share movement above a 1.5 either up or down to correct over the next survey or two. The exception is when there was something obvious that caused the result. For example, Kyle & Jackie O moving stations or in the most recent example, leaving the market and you need to factor in marketing campaigns as well for they bring a station front of mind and given it’s a recall methodology that campaign will have some impact on a survey result. If the product isn’t right though, the residual effect of marketing tends to be short lived. Under normal circumstances audiences growth or decline happens gradually and without trend context, a station might celebrate or react over movements that are largely statistical.

While a clear advancement, GFK Radio 360 in Australia is not a fully passive system (like Portable People Meters used in some U.S. markets). The diary component still exists, so some recall bias, sampling variations and response issues still exist, the wearable panel is relatively small, and integration and deduplication of multiple data sources add complexity, therefore results are modelled estimates rather than pure head count. So single-survey volatility has not been eliminated.

Single surveys are useful for timeliness and competitive benchmarking, but they are inherently noisy as they say, whereas trends provide a strategic perspective that single survey results lack. Trends across three to five surveys smooth statistical variability and reveal whether movements are sustained or not. Examining share, cume, TSL, and average audience together provides far richer insight than any single result. As I have said context is essential: external events, competitive activity, programming changes, and promotional tactics all influence results as does school holidays, big sporting events etc. A movement that looks dramatic in isolation often appears far more moderate when looking at trends. Trends distinguish between temporary noise and genuine strategic progress. They avoid the most expensive reactive mistakes kike abandoning the current strategy after one soft survey or doubling down on unsustainable tactics after one strong survey.

“One book does not make a trend”.

However, from my experience few people in the industry including those in senior management positions possess a nuanced understanding of radio ratings methodology. Content Directors are frequently required to respond to immediate, often irrational reactions to a single survey result by the CEO or General Manager. Following one average survey, I was directed to prepare a detailed list of action points to correct the average result. The document was compiled and presented, which reassured the concerned CEO. Yet the result was abnormal. Ongoing strategic research continued to support the current programming strategy, and my experience over the years told me that the survey result was more likely an outlier than evidence of a systemic problem. A correction in the subsequent survey was therefore expected and it happened. The CEO later congratulated me on the successful execution of the action plan, crediting it with reversing the prior result. In reality none of the proposed measures had been implemented.

What is important, is to understand what’s working and what’s not working before a negative trend develops. Normal audience behavioural changes are gradual, but when they reach their tipping point and move from a P1 to a P2 listener it requires more effort and time to get them back which is why investing in your own audience research is worthwhile. Market studies use a different methodology to that used by GFK in Australia and Rajar in the UK for example, so they are not generally used for tracking audience numbers like radio ratings, but they provide qualitative insight into the various elements that make up the product including music cluster analysis which tracks the music cycle over time, which for music stations is essential. Taking action to correct any issues that the study reveals is important if you want steady growth rather than a declining trend.

“Statistics are essential for winning the battle” Winston Churchill.

Radio ratings remain one of the most rigorous and standardised audience currencies available to any medium in Australia and other developed radio markets. When interpreted correctly, share, cume, and average audience provide invaluable insight into competitive position, reach, engagement etc. But when misread, the data becomes shifting sands and decisions built on that foundation carry real financial costs in terms of wasted investment, damaged relationships, lost listeners, impacted careers and competitive disadvantage that takes multiple surveys to recover.

The station’s senior executives and Content Directors who succeed treat ratings with respect but not reverence. They look beyond the headlines, insist on multi-survey context, and pair the numbers with qualitative insight. They understand that audience building is gradual and that sustainable success is built on trends, not snapshots. In an industry where every survey brings fresh numbers and fresh pressure, the real skill is not just winning the survey, it is understanding what the numbers are truly saying. When there is survey day pressure to act on an average result the best decision might be to not to act at all and that requires courage and conviction which is often more difficult than you think when everyone’s a programmer, or so they think they are.

Footnotes

1. Share (%): The percentage of total radio listening time in a market (or demographic/daypart) that is attributed to a specific station. It measures relative dominance among people who are listening to radio at all during the period (typically Mon–Sun 5:30am–midnight or specific dayparts). It is not a percentage of the total population. ↩

2. Cume (Cumulative Audience): The unduplicated number of different people who listened to the station for at least eight minutes in any quarter-hour block during the survey period (or specified daypart). It can be expressed in thousands (‘000s) or as a percentage of the population (or available audience). It represents reach—how many unique individuals sampled the station at least once with meaningful duration. Cumes cannot simply be added across stations or dayparts due to listener overlap. ↩

3. Average Audience (Average Quarter-Hour Audience): The average number of people listening to the station in any given 15-minute period during the selected time frame. It is the foundation for share calculations and serves as a proxy for typical “in-the-moment” audience size. It is usually much smaller than cume because listeners tune in at different times and for varying durations. Related derived metric: Time Spent Listening (TSL) = total listening hours attributed to the station ÷ cume. High TSL indicates loyal, heavy listeners who contribute disproportionately to share. ↩