CRM: Customer Segmentation
Through Cluster Analysis
Not all customers are the same: they have different needs, financial backgrounds, and buying patterns. It therefore makes little sense to have the same strategy for all of your customers. But at the same time, having an individual approach for each customer is time and cost intensive. The best choice is to combine customers into groups or segments that can be addressed in a targeted - and therefore more promising - way. Such groups could be, for example, "bargain hunters," "trend setters" or "technology freaks."
In order to define these groups, companies are faced with the task of identifying segments that are as similar as possible within a broad customer base. In this business case Fashionette GmbH, an online shop specialising in the sale of designer bags, was faced with precisely this challenge. The aim was to determine which theoretically-possible segments actually exist in order to develop a specific and efficient marketing strategy.
The basis of every customer segmentation is data that reflect the purchasing behavior of customers. This includes data on each purchase, such as the type and number of items ordered, whether there were returned items, the date and time of the order, etc. In order to assess what of this information could be relevant for the segmentation, Fashionette discussed which data could be used to characterize different customer groups before the actual segmentation began. This approach ensures that all potentially-relevant information is considered. In addition, further important information was identified to enrich the existing data on purchasing behaviour, e.g. whether or not the customer's place of residence is in a city.
The data-driven determination of the existing customer segments takes place with the help of a cluster analysis. Within the framework of this procedure, the statistical similarity between the customers is first determined on the basis of the relevant data. In a second step, the customers are grouped into segments so that the customers within a segment are as similar as possible, while the segments differ as much as possible from each other. In the case of Fashionette, eight customer segments were determined in this way, each containing 5% to 20% of the customer base.
The ability to interpret the clusters is essential for the acceptance and use of the segments in the business processes to be optimized. To this end, Fashionette and INWT held a joint workshop in which the segments were interpreted on the basis of their statistical characteristics and converted into “personas,” making them much easier to conceptualize and put into practical use.
Successful customer segmentation is the essential basis for Fashionette's differentiated customer relationship management. Customers can now be addressed in a targeted and demand-oriented manner with the aim of increasing customer satisfaction, and at the same time optimizing business results. Fashionette is now in a position to take targeted, optimized measures, such as:
- Sending a newsletter at the right time, with the right content, to the right customers.
- Sending a targeted voucher campaign with content tailored to the customer segments in order to transfer first-time buyers to long-term customers.
Additionally, new customers are automatically added to one of the defined segments based on their characteristics by the segmentation algorithm implemented in the BI system.