Identifying highest customer value segment with no definitive, data driven, scoring available to identify the segment
Therefore the challenge was to identify the best performing segments along with underperforming ones, to prioritize company investment.
By analyzing Point of Sale SKU level transactions, along with key demographics from loyalty card registrations, and related census demographics (linked by postal code), a scoring system was created, and implemented in the analytic data mart, to produce KPIs on a quarterly basis to track revenue growth in the segments. In order to preserve high-valued strategic business objectives, the analysis also had to preserve natural groupings across verticals: foot traffic, on-line, special events etc.; a balanced system that would not favour one customer type over another, and utilized by multiple stakeholders within the company.
Using a combination of K-means clustering and discriminant analysis, nine segments were identified with varying characteristics and activities across verticals, with the segments not splitting or ‘washing-out’ key target behaviours. All segments had a story to tell, with product likes (and dislikes), days and times of activity, group behaviour, and demographic descriptions. These ranged from high revenue, young family, to occasional student.