Customer Response Prediction for a Specific Campaign in Penti via Data Classification

Abstract

Penti was established in 1950 in Turkey. It is growing in production and retail activities, expanding its offerings of mainly women's and girls’ socks, home wear and beachwear. Different campaigns are organized periodically in order to increase sales and customers are informed about the campaigns mostly by SMS regardless of their previous purchasing behavior. When informing the customers, there is no analytical tool used to find the right group of customers who will receive SMS about the campaigns. Therefore, the response rate to campaigns is not always as it is expected and cost of sending SMS is high. Number of SMSs that are bought from an outsource company and responses to the campaign directly affect the profit. With all these reasons, a study is made on creating an analytical tool just for a specific campaign in Penti. In this study, an analytical tool which will predict the possible responses of customers to a specific campaign is developed by using WEKA software. While developing this tool, different classification algorithms are used. Each method is compared in terms of their accuracy rates and the best method which has the highest performance is selected as the tool. At the end of the study, the analytical tool is shared with the company. By the help of this tool, SMSs can be sent to the customers considering important attributes and the company can reach the right group of people who will do shopping at stores after those informative SMSs.