1. Representative target group segmentation
First, a representative consumer survey is conducted, e.g. online. This forms the extensive data basis for representative target group segmentation. The actual segmentation is then carried out on the basis of factor and cluster analyses and a detailed description of the personas, including their buying behaviour, is created. Segmentation characteristics of the personas can be
- Ready to spend: Bargain vs. Premium
- Purchase frequency: frequent vs. rare
- Occasions: gift vs. personal use
Important: The segmentation characteristics of the representative survey must be compatible with the characteristics of your own CRM data. If a segmentation of buying behaviour is desired, CRM data consisting exclusively of age and gender will hardly lead to a targeted transfer. The CRM data is, so to speak, the “model” for the segmentation survey. A review of the CRM data base available in the company is absolutely necessary. An overview of the type and completeness of the data collected must be determined and any data that may be available separately or in multiple copies must be merged.
2. Exercise data for CRM
This is where the first merger takes place. The determined target group segmentation is linked to the company’s CRM data by first collecting the segmentation characteristics from step 1 for only a small representative sample of customers in the CRM database. This allows the surveyed customers to be assigned to the target group segments. This process step is important for first developing training data records for your own CRM.
3. Segmentation of the CRM data
In the final step of the linkage, classification rules are then trained on this initially small customer data record, which then enables an assignment (classification) of all customers based on your CRM data and independent of the variables used for the original segmentation. The aim is to check the determined segmentation to see whether rules can be developed that allow you to assign the customers to the corresponding segments in your own CRM database. Depending on the data situation, these rules are developed by using various modern classification methods such as discriminant analysis, neural networks, support vector machines or decision trees (random forests).