Problem:
The growing number of promotional activities supported by mailings led to overwhelming the customers with marketing communication. One of the retail chains in Poland wanted to reduce the number of emails sent per customer while increasing their relevance in terms of the product range promoted.
Solution:
A scoring model (machine learning) was built, predicting the probability of a given customer’s interest in the product range promoted. A flexible tool for automated model building was developed supported by communication specialists. The success of the model resulted in a project to expand it with further data sources to increase forecast accuracy
Effects:
- Fourfold increase in conversion in target groups using the model
- Reduction by about 60% of the costs of preparing dedicated models for each campaign
- More flexibility and shorter waiting times for the model. A dedicated scoring model for the campaign can be created in less than one working day
Data sources:
- Transaction systems
- Loyalty Program
- Calendar of promotional campaigns
- Website traffic
