Using Data to Identify Service Requests that Turn Critical
Our client, a leading technology manufacturer wanted was spending millions of dollars trying to retain the customer once they become critical, so they wanted to build an early warning system which could help them in identifying the service requests which could potentially turn critical.
Analyzed previous critical requests and identified patterns of behavior.
Included unconventional but critical data variables like email structure (number of people copied on an email, the use of bold fonts, sentiment), number of webex meetings, customer service logs, product manuals etc.
Model included 120 variables and analysed over 2 million records
The model was able to predict critical service requests with over 95% accuracy
The early warning solution is estimated to have saved the client $20 mn every quarter
Time taken to effectively address a critical service request reduced significantly. This resulted in an overall increase in customer satisfaction