Case Study
We’re not a regular consulting firm. More than Data & AI suppliers, we are reliable partners. Both transparency and independence are at the core of how we work and guide the development of strong and reliable partnerships with our clients.
We’ve built a technical call relapses prediction system for one of the major telecommunication companies in Iberia. Currently, 30% of customer calls generated a relapse (a new call that normally escalated the complaint) and hit rate was low. These customers were at a higher risk of churning as they were already having a lot of issues with their service and the company’s proactivity was the only measure that could improve customer satisfaction.
The project aimed to predict call center relapses by giving priority to customers that were at a higher risk of churning due to lower satisfaction with the company. We’ve included granular data from past interactions with the customer, while also taking into account service level and product information.
Our gradient boost model was able to predict call center relapses until a satisfactory Precision @ K metric. From our classification model, our customer could now select the top customers with higher likelihood of relapsing and contact them proactively, decreasing churn.
01
Feature Engineering
02
Baseline model
03
Deployment
04
Improvements
