Opinion | Thoughts on audience profiles and segmentation...

Cimeon Ellerton considers how profiles and audience segments are built ...

I don’t know about you, but when it comes to audiences and how to reach them, it feels like it’s never long before someone pulls out a profile or describes an audience segment.

But do we ever stop to consider what profiles and segments do and how? It’s really important to think about how a segmentation model was built if we’re going to make best use of it.

Let’s start with profiling.

So what is an Audience Spectrum profile and how does it work? An Audience Spectrum profile simply groups and counts all of your bookers or survey respondents according to typical behaviours, interests and preferences of the UK population as a whole. This takes into account all the natural variation in audiences across geographical, attitudinal or even artform ranges and groups them in a way that is both actionable and statistically significant (using data that meets the five R’s of Data Quality).

But what is significant? It is generally not a good idea to waste effort in describing people according to characteristics that are not relevant to what you are going to do. Or worse, guessing which category they fall into - getting ID’d because you’ve retained a youthful glow is flattering, but that sort of categorisation mistake based on a personal assumption can be less flattering or even highly upsetting in a different context.

The only way to know this information for certain is to obtain data on it. So asking (through surveys) is the only safe way to do so. There are, of course, algorithms that do make quite good guesses (Google Analytics provides a male:female ratio in this way). There are even companies who claim to give you a breakdown of likely ethnicities from a list of names, but what they’d make of my name and particular spelling is anyone’s guess. Unless you have a team of data scientists building and testing such a model, keep away from this, or you may fall into the trap of ‘common sense assumptions’ masquerading as data science - assuming everybody on your database who has a traditionally male name identifies as male or vice versa is just such a trap.