Data-science in Radio, “it’s just a tool, isn’t it?”

A few years ago, a video from visionary David Bowie got viral. In an interview on BBC, from 1999, Bowie predicted the impact of Internet. The perplex interviewer,  Jeremy Paxman, asked “It’s just a tool, isn’t it?”. Bowie replied “No, it’s not. It’s an alien life form” and laughed. I believe both of them were right.

The usage of tools is defining for human kind. Literally.

Making, using and refining tools is one of the main elements that constitute the origin of early humans. The first hunting tools made of stones helped our ancestors ascend in the food chain. Fire enabled us to survive extreme climate conditions and to extend our lifetime by making our eating healthier.

It wasn’t just the utilisation of those tools that represented the evolutionary leap towards homo sapiens, though. Actually, other hominids used stone tools even before humans. Some other primates also do.

What is exclusive to humans is our capacity to understand the causal relationship between tools and the result of their use. It’s learning, understanding and transmitting socially our knowledge through generations what makes us unique.

The history of technology starts with those first tool inventions that, because of human understanding, weren’t products of magic to our early ancestors. Today, sometimes we need to remind ourselves that there’s no magic in technology. It’s our capacity to identify and understand causality versus casualty, the science behind the technology, that helps us survive, extend our life expectancy, improve our conditions and thrive in many other ways.

Data-science in Radio as a (non magical) tool

Enough playing anthropologist! I believe I made my starting point. Time to land on our field. Let’s get back to our team’s (Voizzup) point of focus in technology: data analysis at the service of on-air content evaluation in Radio industry.

Guess what, data aren’t magic either. By just collecting them, we don’t produce an impact. Data analysis tools help us find, observe over time and understand cause-effect relations. Again, it’s the science behind the technology that generates improvement.

By (data) science, we mean scientific method: testing causality, documenting results and sharing learning to be challenged in the future. We won’t go very deep into scientific method in this article. Instead, we will tackle the difficulty radio professionals seem to have comprehending data analytics as just tools. We intend to describe the different stages we have encountered in the learning curve when using on-air content evaluation tools. From making sense out of collected data, to empathetically understanding the psychology behind audience’s listening behaviours.

1. Darkness

Several professionals in the teams we work with have described the time before having analytics for daily on-air content evaluation this way. They rely on gut feeling. They have to, given the lack of insights. Often, assumptions, inherited absolute truths and undisputed beliefs drive content related decisions. Intuition is a highly valuable skill. However, limiting to intuition as your only prompter when deciding what to bring, keep or remove on-air, is far from ideal.


Most stations we worked with had strong but wrong assumptions before they started evaluating content with Voizzup.

  • “Commercial breaks have the worst performance, songs are safe”
  • “We are a news&talk station with a focus on business, our listeners have no interest in sports”
  • “Playing a brand new song, previously unheard, every day is a strong feature in our programming”
  • “Listeners are tired of this topic, we’ve been covering it for way too long”

You can be convinced and wrong.

2. Burden

A sudden source of light in the middle of darkness can be blinding, at least for a while. Moving from a very silent reality dominated by our intuition, to a noisy new scenario full of data points and daily second-by-second insights is baffling, for everyone. Every new show team we activate, has this initial reaction. The ability to find cause-effect relations between what you do on-air and how listeners respond needs to be trained. Everybody’s first attempts are usually dragged by insecurities, fears and obsessions. Resistance is a very human attitude when our core beliefs, purely based on intuition or unchallenged conventions, are potentially refuted by data. Our colleague Tommy often mentions Star Wars’ Jedi code for explaining this: Fear is the path to the Dark Side. The Force requires discipline and training, which never end.


It’s probably not an exaggeration saying most radio professionals we’ve worked with suspected that daily content evaluation would constrain their freedom and creativity. At the same time, the first impulse for some of them when starting to use data was to examine (almost torture) themselves. Soon after seeing listeners reacting second-by-second, they acknowledged the power they have on-air. That made them more responsible (Tommy calls this the Spiderman effect) and, paradoxically, also freer.

Once, the host of a music morning show learned of the death of an artist from local underground music scene when she was on-air. She cried during the show. And she deeply regretted doing so. Until she saw that such demonstration of authenticity was actually the most engaging moment for listeners since we started measuring their reactions.

Relying on impressions which come from fear is not a good idea.

3. Ownership

First step to overcome the burden is taking the driving seat, conquering the dashboard, owning the data. They are collected, processed and displayed for you to use. Remember, it’s just a tool.

To clarify, this piece of technology won’t replace the team of the radio station making decisions, implementing changes or dictating strategies. Nonetheless, it provides radio professionals with abundant and detailed information for them to answer their questions. The on-air or programming team needs to formulate questions first and figure out how to sort them, based on relevancy and priority.

Raw data or even processed and beautifully shown data on a dashboard are not insights. Insights need to be produced from data. By formulating questions previously, having a schedule of what and why we need to evaluate, finding patterns with a possible impact worth exploring and filtering out non-causation correlations.

Also, some show teams start getting rid of their fears by now. They begin challenging core preexisting beliefs, confirming some and denying others. In both cases, gaining knowledge.


Super popular radio host in his market. He has been directing and presenting this morning show for more than ten years. Always based his decisions purely on intuition. It’s in his nature, he’s a radio beast. From the first day using Voizzup he embraced it and, most importantly, owned it. Soon, he started testing new elements for the show and being confident about their results, keeping some and discarding others, after two weeks of tests on-air. 

In his words, Voizzup is “a radio-eye opener”.

4. Experimentation

Attitude becomes methodology at this stage. Losing fears and questioning beliefs turns into confirming and rejecting assumptions, systematically. In other words, the scientific method for experimentation is our new framework.

At this point, the team of the station formulates hypothesis, refutes or validates them through experiments, defines and manipulates independent variables, measures dependent variables, reach and document conclusions.

Believe it or not, most times this takes place with the team avoiding the use of such laboratory terminology or even ignoring the scientific nature of the methodology in the first place.


The teams at the stations we work with, together with us, have conducted tens of experiments in recent years. We have introduced isolated modifications in the on-air content or structure of radio shows and measured whether the impact on Time Spent Listening, in most cases, was positive. Some of those manipulated variables were:

  • Duration of travel/traffic information.
  • Placement of teases.
  • Length and structure of contests/games. 
  • Pushing breaking news notifications.
  • Categories of guests for interviews. 

Anything you put on-air is being tested. You just need an extra bit of methodology.

5. Empathy

For understanding this stage, two facts of Voizzup’s data-science applied to content in Radio are key:

  • Before aggregating events from thousands of listeners, collected raw data consist of individual listeners’ actions. That allows us to see that “audience” is an artificial concept that over-simplifies the comprehension of listeners’ reactions. The audience doesn’t exist as an entity with an homogenous behaviour. At most times, both engagement and disengagement are happening simultaneously.
  • The high level of granularity in our data allows us to observe extremely clear correlations between content and listeners’ reactions. By displaying these reactions second-by-second, the dashboard reveals changes in the flow of listeners happening very fast as we move from one programming element to another on-air.

As we’ve already seen, at this point the station team has familiarised with the methodology for detecting correlations between on-air elements and listeners’ behaviours that are really causal. This, combined with the two facts just mentioned, triggers a shift in the perception of radio professionals working with Voizzup.

This is the dimension of context, of psychology even. It’s the impressions, the feelings, the motivations behind the listeners’ acts that centres the conversation.

All of a sudden, the thinking and the discussions within the team transcend the dashboard, its indicators and even the mere numeric presentation of data. They reach a higher state of mind-visualisation of listeners’ behaviours.

At this stage, knowing that a programming element triggers tune-outs is not sufficient. We aim at understanding what feelings or emotions were linked to those tune-outs.


Recently we published a blog post about breaking news push notifications. We showed a graph that displayed the spike in Reach (number of accumulated listeners) generated after a “special coverage” notification was pushed on the mobile apps of a radio station. Generally, we want to make that spike as wide as possible (yes, high as well…) in order to increase listening time. As we explained in the article, often those spikes are very high, but not very wide. For multiple reasons, the listening time of the listeners who tune in triggered by the notification is usually very short. 

Precisely those “multiple reasons” are our point of focus in this example. Our most mature customer (more advanced in the learning curve) is at the “Empathy” stage. Concluding that short listening sessions after push notifications are not optimal would probably be correct. That wouldn’t be sufficient learning gained, though. 

We are currently working with them in understanding the correlations between the call-to-action (tune-in live now, in the notification) and the listening behaviours. On an almost psychological level. For those listeners who tune-in triggered by the notification on their smartphone, when tuning-out after a few minutes… Is their expectation, their need for an update, satisfied? If that’s the case, the radio station and its mobile app have played their role successfully, happy listener. Or on the contrary, listeners expected more information and didn’t see the value in listening longer? If there’s disappointment behind the tune-outs we are measuring, we’ll likely see a negative impact also in listener loyalty (and even market positioning for the radio brand) in the long run.

Data-science amplifies understanding

We started this article with the video of David Bowie predicting the future of Internet, discussing with Jeremy Paxman whether Internet was just a tool or a revolution. Another viral video might be relevant at the end of this article. Steve Jobs, also in an interview, said: “We, humans, are tool builders. We can fashion tools that amplify these inherent abilities that we have, to spectacular magnitudes.”

Bowie thought Internet was a revolution. For Paxman, Internet was just a tool. Jobs understood that the tools we create and use magnify our abilities which, ultimately, enable transformation. For Steve Jobs, a computer is a bicycle of the mind. For us, data-science in Radio is the bicycle for the understanding of listeners’ behaviours.

Please, get in touch if you think you could use our help riding it! 

VOIZZUP - 2022