Project Description



The Client, a multinational roadside assistance provider, was confronted with falling responses to their Direct Mailing activities, which had been a significant source of leads in many years. Break even point continued to fall and the Client had to find a new method to identify leads.


The Client was applying known business rules to select leads to DM campaigns, and wanted to challenge this method. By using One Prediction, the Client could predict which leads are more likely to respond positively to a direct mailing, per category and which leads should receive each of the products offered. Also lowering the need to send out unnecessary direct mailing to leads not likely to respond positively to the offerings.


Using existing data from the Clients purchase history, car attributes, customer preference data (geo-social data) and responses to previous marketing efforts. One Prediction could produce a list of leads, ranked by their likelihood to positively respond, from the most likely to the least likely, on a monthly basis. In addition, a list of leads, ranked by their likelihood to positively respond to one of two product offers. And last but not least, identifying a threshold at which ROI is still positive.


The Client could reduce the number of letters sent by 30% and at the same time increase ROI with 25%.