Customer Value Segmentation

With finite budgets and a large customer base a segmentation was needed to help understand the shape of the database. The primary objective was to understand the value of an individual customer so that it was possible to compare them against others and use this position in the hierarchy to include or exclude the customer from marketing campaigns based on the objectives of the activity.


Understanding a customer's rank enabled decisions to be made around the level of investment made and in some cases, whether to not invest in a customer at all. The investment was not only limited to the marketing spend, but sales representative and branch staff time.


The starting point for the project was to look at existing techniques for VIP and engagement models. Recency Frequency Monetary models are a staple in direct marketing, helping marketers find their ‘best’ customers. These models typically rely on the designers assigning a weighting to each parameter, with customers falling into one of 3 or 5 boxes for each, giving you a code that assigns the box they sit in.


Taking into account the need for the segmentation to be easily understood and actioned by branch staff and sales representatives in addition to marketers, a decision was taken to move away from typical RFM modelling in order to meet the objective of using the segmentation outside of the marketing department, where concepts such as RFM are more easily understood.


A dynamic calculation was chosen as the method of segmenting, which gave the sales and operations staff a much easier concept to understand, the closer to 100 the score is means the customer is good, the closer to 0, it is bad. A second angle was added in the form of a Red/Amber/Green (RAG) status to show the progression of the customer from the last time that the score was calculated.


The results of the model development were tested not only with the marketers, but also with branch and sales staff to gauge the level of immediate understanding and usefulness. The time spent testing the model output with non marketers made it easier to understand how much education was needed and which direction to take the internal communications so that it effectively supported a roll out of the segmentation.


Clustering analysis brought out the groups of customers within the scoring and enabled profiles to be created which helped people across the business understand more about the customer base. This provided the most support to the marketers through understanding the behaviours within the clusters and became useful in excluding customers for direct marketing activity.


The final model was automated so that the scoring rebuilt on a regular basis and could be accessed by analysts within the business to support the marketing, sales and branch activity. The RAG status became especially useful as a method of highlighting sales opportunities for branch staff to use as it prioritised their outbound sales call lists. This became a key component in a branch support programme and was well received by branch staff and well supported by commercial and operational management within the business.