Sales Leads Conversion - Predictive Modeling

Summary

DOMO is a technology company that provides a cloud-based business intelligence software. Being a growing company, DOMO needs to move quickly to acquire new customers. One of their strategy is to obtain thousands of leads every single week. However, the majority of the leads will not convert and only a small portion will move to the next step where account executives will take over and contact each of those leads.

By understanding the characteristics of promising leads, DOMO will be able to reduce the cost of obtaining leads and help the pre-sales team to be more efficient. We train a machine learning algorithm to create a scoring system (between 0-1) where good leads will have value closer to 1 and bad leads will have value closer to 0 to separate the good from the bad. The machine learning model also gives us the variables (characteristics) that are most common among the good leads which will enable DOMO to obtain more leads with specified characteristics.

The model's performance turns out to be quite satisfactory with AUC score of 89%. We also learn that there are several factors that influence the likelihood of conversion negatively and there are others that have positive impacts. The figure on the left represents the variable importance of model output. Red-colored bars represent variables that have negative impact on the conversion and blue-colored bars represent variables with positive impacts. The size of the bar represents the magnitude of the impact, meaning that larger bars represent larger impacts on the likelihood of converting. The company needs to reject leads that have the characteristics that are marked red on the figure above. The company also needs to focus their sales effort on the leads that are marked blue on figure above.

The machine learning model is able to identify 25% of the lowest leads and is not affecting the final conversion. This means that the company have improved their sales leads operations by at least 25%.

R Codes on GitHub