FORMULATING CUSTOMER RETENTION STRATEGY BASED ON CHURN PREDICTION MODEL (CASE STUDY: S-FITNESS)

• Main Author: Fasya Ghasani - Nim: 14414028 , Nadhira
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/29470
• Institution: Institut Teknologi Bandung

<p align="justify">S-Fitness is a moslem women only sports club that was built on 2016. It is located in Cihampelas, Bandung. On their third year in 2018, S-Fitness have a very low retention rate, only 15%. The other 85% do not repurchase their membership period and become churners. Designing strategy to retain customer is initiated by building a churn prediction model. This research compares two models, logistic regression and decision tree. Data collected from 352 members from January – September 2017 and Januari – Febryary 2018 are used which involves six input variables. Decision tree models were found not effective to build a classification model using unbalanced sample because it predicts all sample into one group. Meanwhile, logistic regression model can do better to classify sample into two different groups. The logistic regression model shows four out of six input variables significant. They are price, attendance frequency, tenure (membership duration), sales promotion usage. The other two, age and marriage, which are socio-demografic variable, found to be insignificant. Model is used to calculate churn probability of S-Fitness member from March to May 2018 and to target them into sales promotion. Based on the significant variables, there are two other strategies that are proposed, attendant reminder message and community event. These two are designed to e increase the member attendance frequency. <p align="justify">