TOWARD A SMART LEAD SCORING SYSTEM USING MACHINE LEARNING

Abstract

The segmentation of new commercial leads is a crucial task for modern and highly competitive businesses, to identify new profitable opportunities and enhance their Return On Investment (ROI). Business Lead scoring involves assigning a score (i.e., a buying probability) to each possible lead generated for the business. The interactions of these leads with the business marketing channels across the internet are converted into multiple attributes, including useful pieces of information (e.g., contact details, lead source, channel) and behavioral hints (e.g., reply speed, motion tracking). This process can help assess the quality of the opportunity and its position in the purchasing process. Furthermore, an accurate lead scoring process can help marketing and sales teams prioritize the selected leads and appropriately respond to them within an optimal time frame, increasing their propensity to become clients. The use of machine learning algorithms can help to automate this process. In this paper, the authors compared the performances of various ML algorithms to predict lead scores. The Random Forest and Decision Tree models have the highest accuracy scores of 93.02% and 91.47%, respectively, whereas the training time of the Decision Tree and Logistic Regression models was shorter, which can be a decisive factor when dealing with massive datasets.