A Comparative Application of Machine Learning Approaches to Win-back Lost Customers

Today’s consumer is more knowledgeable and conscious than in the past. For this reason, it is quite possible for consumers to leave their service/product providers and start receiving service from another service/product provider. Without a recovery strategy, companies often do not target their lost disloyal customer portfolio correctly and encounter the problem of lost customers. Lost customers can cause loss both in economic terms and in terms of business potential. At the same time, lost customers can also be considered as profits given to rival companies. What if the companies could foresee lost customers who would not want to receive service from them again? Could companies win back their customers? At this point, the article proposes using machine learning methods to recover lost customers for service providers. The customers that are likely to be lost in the future are estimated using the article’s past stories of an automotive company’s lost customers. The data used is completely real. LGBM, XGBoost, and Random Forest methods were used to estimate lost customers. Finally, the authors select the machine learning with the highest predictive success for customer recovery and discuss why this method might have worked well.