A high number of pop-ups, emails, text messages and calls are constantly rocketing the screens of gadgets with customers devoting less time on average to each notification. Thus, telco operators should use time and frequency with caution and use the extent amount of collected information, from previous interactions, to increase the effectiveness of market campaigns without jeopardizing customer experience.
The current case study was part of a project together with a major telco operator to develop a prediction-prescription engine capable of increasing the sales volume performed by a call center operation without eroding customer loyalty.
The used methodology consisted of a three-step approach:
Predictive models – Development of a predictive model to leverage information into predictions of ideal contact timing using a tree base boosting machine learning algorithm considering a large set of variables grouped in four main segments: customer profiling variables, former contacts made, gadget, and tv-box usage.
Optimize the timing of contact – Creation of a prescriptive engine capable of prioritizing clients’ contacts, which were organized in ordered lists per day/hour according to contacting a client outside its preferred contacting window.
Methodology validation – A one-week pilot test was devised, running on an operational dialer adjusted to apply the newly design contacting strategy. Customers were randomly divided into control and test groups, assuring that both groups had similar behavior in every metric.
The pilot results showed a 23% increase in commercial activity, as well as higher customer satisfaction, as the sales conversion was achieved with fewer contact attempts. In the aftermath of the project, the methodology was internally productized, sustaining the promising results achieved during the pilot test.
