Survey-based attitudinal segmentation maintains its strong position in the market research toolkit. It is highly valued by brand and marketing professionals. Market segmentation is considered an essential piece of research for setting the marketing strategy.

Segmentation, very broadly, allows marketers to identify groups of people sharing certain commonalities and use these insights to better target potential customers with the most compelling messages or products. In addition, better understanding what drives groups of customers enables more effective customer value proposition and better streamlined product and service offer.

Traditional Use of Attitudinal Segmentation

The segmentation profiles have been instrumental for decisions on messages appealing to the largest or the most valued segments of consumers. Additionally, segmentation helps building product and service offer targeting customers from preferred segments.

For those brands that do not have customer database and most communication has the form of mass media advertising - until recently that was the case for many FMCGs - attitudinal segments actually make sense and can add significant value.

Increasing Volume of Customer Data – Challenge or Opportunity?

In the past years, we have seen lots of changes on the customer data landscape. For many online-based start-ups, collecting data on all customer interactions is a critical part of their business model. Established companies increasingly invest into building and upgrading customer relationship management (CRM) systems and customer databases. Consumer brands selling through brick-and-mortar stores have been launching loyalty programs to capture variety of transactional data and to learn more about their customers’ purchase behaviour.

Some companies managed to bridge very well traditional market research with own customer data analytics. But there is a number of those that struggle. The effectiveness of their marketing activities does not reach its full potential as long as market research outputs do not help leverage the value of customer database for personalised targeting and messaging. The attitudinal segmentation is a good example of marketing tool that need to be linked to customer data to be more actionable.

There are several ways to go about it. They can be summarized in two groups. The first one is to link existing survey-based segments with customer database. The second approach starts with behavioural data to build segments. Below I briefly discuss both.

Mapping Attitudinal Segments to Customer Database

In the first approach the existing information about the surveyed customers is used to build a predictive model that assigns every customer to one segment (or to every segment - to each with certain probability). The model is based on the assumption that customers with similar attitudinal profiles share commonalities in their behaviour. The algorithm is then used for classification of all customers in the database.

The main objection to this approach is its mapping accuracy. According to Colin Tener from Knowledge-based Marketing [1] – the algorithm correctly classifies only around 65% of modelled records. In addition, accuracy of new record classification is often only around 50%. From my personal experience, this figure is indeed not far off. However, it is possible to achieve better results when the company collects information on a bigger variety of customer interactions and touch points.

There are several reasons for lower accuracy. First, segments vary in their homogeneity. Some segments are more clear-cut and these can be predicted better compared to more heterogeneous segments. Second, availability of behavioural data covering all aspects of customer behaviour has an impact on classification accuracy. Third, the objective of segmentation and the focus of survey questions can also affect the success rate of mapping.

There are some ways to mitigate this problem. First, a representative sample of customer data should be included in attitudinal segmentation. The customers selected in the sample should have a sufficient track record of past transactions and behaviour.

Second, it is better to use a predictive algorithm that generates for every customer a set of probabilities of classification to every modelled segment (these sum up to 1). This output is more informative than simply identifying one segment where the customer most likely belongs.

Another, more subtle, objection to mapping questions validity of the assumption that similar attitudinal profiles imply common behaviours. Correct, there can be segments where similar attitudes can be associated with a variety of actions. And the mapping exercise help reveal exactly these segments. This is important as better understanding of more complex segments can lead to more effective communication strategy and product offers.

In sum, segment mapping to database is definitely more useful than giving up on the use of attitudinal segmentation in database marketing. That said, this approach is more a necessary step to leverage actionability of existing segmentation. If a company plans for marketing database programs, attitudinal segmentation is really not the best way to start. Agree with Colin on that one. So what is next?

Behavioural Segmentation Supplemented by Attitudinal Profile

If segmenting all customers in the database is a priority, I see starting with behavioural/transactional data as a more sensible approach.

Behavioural clustering can be based on a larger volume of data. If appropriate, one can create a higher number of more granular segments. New customers that have not been included in segmentation are then assigned to the segment with the closest behavioural profile. Insights based on understanding of segment profiles can lead to actions improving customer experience, engagement and, in turn, their loyalty and spend.

However, as pointed by one market researcher, I discussed this matter with, ‘behavioural segments are unlikely to unlock growth opportunities’. A thorough analysis would show to what extent this claim is justified. And even if we are missing some angles, the marketer can always undertake attitudinal profiling study where every behavioural segment is equipped with an attitudinal profile.

Several behavioural clusters can share similar attitudinal profiles shedding more light on different types of need-, event- vs attitude driven behaviours

In this way, the survey questions can be focused on filling gaps in knowledge. The actual value and actionability of research can be much higher.

Industry practitioners speak also about holistic segmentation that takes into account a broad range of customer attributes or about a ‘segmentation toolbox’ that combines several segmentations, each based on a subset of customer characteristics. Such toolbox provides multi-layered segments that can be flexibly used to solve specific problems. These are all very interesting topics, that go beyond the scope of this post.

Are We There Yet?

Segmentation, similarly as other analytical research methods, should pay for itself by lower conversion costs, improved customer value and retention. For this purpose, being able to classify all customers into identified segments is paramount. Otherwise, the customer segmentation, that often incurs a significant investment, has very limited application.

Companies can choose the way that works best for them. Just one message at the end: Always start with your own data! You cannot really go wrong, when you base your strategies on what you know and understand about your customers.

Sydney, 26 April 2014,

Elena Yusupova, PhD

[1] Colin Tener, Opinion: Attitudinal segmentation a waste of money: While the process used to define the characteristics of specific sets of like-minded customers is statistically sound, the end result is all but useless, 1998,