Segmentation models – an alternative to targeted samples?

In a previous blog we’ve looked at the impact of GDPR legislation on the availability of targeted consumer data (hint: it resulted in fewer data fields available for selection). With this in mind, are segmentation models worth considering as a solution should targeting be lacking?

Let’s start by defining what we mean by segmentation models. These are – in the world of offline data at least – tools that divide the UK population into different categories and sub-categories. Based on a number of factors such as census data, housing, lifestyle data and other publicly available sources, the data is analysed and the segments then identified and named. There are two very well-known geo-demographic segmentations: MOSAIC which was developed and is owned by Experian and ACORN from CACI. Both offer a similar number (10-ish) of clusters, ranging from 5% to 15% of the population, and around 60 sub-groups.

So far so broad. However the modelling is already reasonably precise, letting us deduce media consumption such as newspaper readership. But can we be more accurate than that? How far can we target households for the purposes of market research sampling? There isn’t a neat answer, but rather a range of accuracy. As noted in another blog the past twenty years have seen consumer / household data become less readily available and with fewer selections when it is there. Can we use geodemographics to determine ownership of certain makes or models of cars? Although we can draw certain conclusions about life-stage and family, it’s unlikely that segmentation alone will reach even a popular car model such as the Ford Fiesta, despite the volumes of that car sold over say a 3 to 5 year cycle.

None of which is to say geodemographics can’t be useful or employed to target samples. Indeed, when used in combination with lifestyle data – such as we hold in-house – it can be powerful. Perhaps it’s best then, to temper expectations around being able to target brand or product usage and look to segmentation for category buyers. A good example here would be financial services. Blending a segmentation model with other criteria such as household income or social grade allows us to target households with private pensions or perhaps credit cards with a particular credit limit.

Another possibility is for sample records to be supplied with the segmentations, allowing for analysis by groups once data has been collected. Should you have clients who already work with geodemographics and require that completed interviews are flagged with e.g. MOSAIC, we can append this to any records you already have, or any records you may need to purchase. Sample Answers currently has 3 segmentation models in-house, with Axciom’s Personix being perhaps the most well-known.

In conclusion then, “geodems” are neither a magic bullet of precision nor the blunt tool that cynics might lead you to believe. Rather, they are, if used in the right context another valuable way to sample and reach parts of the population when data availability or cost might preclude a more targeted approach.

If you would like to find out more about our segmentation models, please click here.