I’m a huge fan of Mark Ritson. He recently wrote a column for The Drum about a survey he received from Anthropic. The goal of the survey was to get human input into improving their Cowork product.
Kind of ironic, no?
But this led into a discussion about synthetic market research. This is potentially a huge industry because getting AI to answer questions is a lot cheaper than obtaining and reaching a live sample. As Ritson says, “Synthetic research gives you speed, scale and lower marginal costs…” And it can reach hard-to-find segments. Like radio listeners.
The scary part about this is that a couple of studies cited in the article say that the synthetic respondents agreed with the humans at an 85-95% rate.
Now, it goes on to say, “Every synthetic respondent ultimately depends on human-generated data. Without fresh human evidence, the synthetic layer risks drifting away from reality.”
So, we’re safe… for now.
The question is, how can radio use this to our advantage?
Can AI analyze hit songs to find similarities in their tonal structure and lyrical content that make them more likely to be a hit?
Can Xperi, for example, throw all their data into Claude and find out what tends to cause someone to change a station? Some of this is intuitive, but with their sample base we might learn something.
Can AI analyze your stream data and determine common behaviors your listeners exhibit?
Can we throw 100 perceptual questions at AI and find out which five or ten are worth including in our next perceptual study?
Honestly, I don’t have the answers to these questions. But someone does. Radio may not attract enough attention from tech experts to warrant a deep dive, but our industry has a vested interest in pursuing this line of questioning.
We’ve toyed with AI-generated voices and copy. Maybe we use this tool to find out what resonates with current and potential listeners. We might learn something useful.
You can read Ritson’s article here.
Feel free to tell me I’m full of it: [email protected]
