Callcredit Blog

Attitudinal segmentation

Consumer Marketing Data

One of the common requests we receive from clients relates to the mapping of attitudinal segmentations to the wider customer or prospect base. Often the client has engaged with a marketing or research agency that has undertaken the research on their behalf. The research agency will perform the fieldwork, collate the results, apply statistical techniques, e.g. clustering analysis and produce an attitudinal segmentation. The segmentation will be accompanied by a series of pen portraits that describe each of the segments in detail. A lot of the solutions I have seen are exceptionally well presented, glossy and insightful; these guys know how to sell their solution, far better than us data monkeys!

Solutions of this nature, done well, can provide valuable insight into the attitudes, motivations and preferences of a client’s customer, which can be employed to direct future marketing messages, product development and further research. Due to the insight provided by such solutions, the client often then wants the segmentation mapped across all their customers or prospects so that it can be employed for instance, as a method of selection or to tailor the offer or communication more appropriately.

If this need has not been considered at the outset, this is where the issues arise. There are a number of limitations in being able to map solutions of this nature to a wider base, these can include:

  • Inappropriate sampling: When you get under the bonnet of what has actually been done, you can sometimes find that the solutions are based on very small samples or heavily biased samples, which do not generalise to a wider population. We have seen solutions for blue chip companies that have been based on samples of no more than 200 consumers and others that only account for a fraction of the different behaviours in the customers base, e.g. a financial institution where nobody was surveyed if they were over 65 or had savings less than £1000.
  • No bridge: As well as a robust and sizeable sample, in order to map an attitudinal segmentation to a wider base you need a bridge, i.e. a way of linking the survey responses to the data you have on the entire population, which could be demographic or behavioural data. Typically this bridge would be a name and address or e-mail. Quite often this bridge does not exist as the survey has been undertaken in an anonymous fashion or is subject to confidentiality and therefore cannot be provided. In such cases you could try to use the pen portraits to create a mapping, which is far from ideal, especially if the descriptors do not exist on the base population. We once did this and could only classify 30% of the customer base as every pen portrait basically described the same demographics – very useful!
  • Mapping data: It is often the case that only basic information is available on the wider base for mapping the attitudinal segmentation solution, e.g. Simple behaviours, gender, age, marital status and property characteristics. In this circumstance we would have to assume that every married male, aged 40-45 living in a 3 bed semi has the same attitude towards a particular issue, which is obviously nonsense. If you look at yourself and your work colleagues (try focussing on the ones of a similar life-stage), can you honestly say that they have the same principles and attitudes as yourself. Unlikely.

Believe it or not, I do think there is value in doing attitudinal research and mapping it to the wider base, it just needs to be considered at the outset and a practical approach to the mapping adopted.

Clearly the ideal but extremely impractical solution is to undertake research on the entire base and map the attitudes directly using known information. This never happens; it is just too expensive, so our approach needs to consider what we have readily available and how we can use this to understand the relationship between behaviour, demographics and attitudes/motivations.

Assuming we have behavioural information as well as a series of demographic characteristics on the customer base, we can use these variables to create a segmentation solution that can be applied to all customers. The segments are therefore defined by known behaviours and demographics that result in distinct groups of consumers that interact with the company in different ways.

These segments are used as the sampling mechanism for research – this is the key; thus individuals from each segment, displaying these non-homogeneous behaviours, are proportionately selected to take part in the research. This ensures that the findings of the research can be generalised to the wider population very easily, under the assumption that individuals within a particular behavioural segment, have similar attitudes and motivations. Using this approach we have created the link between behaviours and attitudes.

Foul play, I hear you cry, and you’d be right – what I’m suggesting is really no different to the approach that I criticised only a few sentences ago, and it carries similar risks. I’ve just gone about it in a different way – a more practical way. Instead of using behaviours and demographics to predict attitudes, I’ve assessed the attitudes for different behaviours and demographics. It is subtle, but there is a world of difference in the ease of application, usability and robustness of the solution.

In all honesty there is no easy way to map attitudes and motivations to your entire base without doing very extensive research. Trying to map a small sample to a wider base is difficult and there are lots of assumptions that you are going to have to make along the way. Many of these relate to the relationship between attitudes, behaviours and demographics. It will never be exact, but if we think about it in advance we can make it easier.

So get your qual team and your quant team together at the outset of the project and discuss: What do you want to achieve? How are you going to use it? How will it be applied? What solutions are already in place? What does a representative sample look like? etc. etc. It’s good to talk, but in this case, get your teams talking before you talk to your customers.

Author: Gary Childs

Leave A Comment

Your email address will not be published. Required fields are marked *