AI Case Study
T-mobile reduces churn by up to 50% by identifying and retaining highly-influential 'tribe leader' customers with advanced predictive modelling
T-mobile USA analysed the social relations between subscribers to understand its impact on churn. They identified 'tribe leaders' who have strong influence in a large connected group. These leaders have outside impacts if they churn taking more of their network with them. Using this insight and the analysis T-Mobile was able to reduce churn by up to 50% by focusing on retaining these influencers.
Industry
Telecommunications
Mobile Telecommunications Services
Project Overview
"To better predict customer behavior, T-Mobile USA has started to include social relations between subscribers in its churn management model. The organization uses a multi-graph technique, similar to the methods used in social network analysis, to identify so-called 'tribe leaders'. These are people who have a strong influence in larger, connected groups. If a tribe leader switches to a competitor’s service, it is likely that a number of their friends and family members will also switch; it is like a domino effect. With this change in the way it calculates customer value, T-Mobile has enhanced its measurement to include not just a customer’s lifetime subscription pending on mobile services but also the size of his or her social network or 'tribe.'"
Reported Results
"After just the first quarter of using its new churn management model, the organization’s churn rates plunged by 50% compared with the same quarter in the previous year."
Technology
Function
Marketing
Customer Management
Background
"A problem that challenges telecommunications providers is that of customer churn (the loss of customers over a period of time). To help reduce churn, organizations typically analyze usage patterns of individual subscribers and their own service quality. They also offer specific rewards to keep some customers loyal, based on parameters such as customer spending, usage, and subscription length. In the past these retention efforts based on individual customer value have achieved some improvement in loyalty, but customer churn remains an issue for providers"
To better predict customer behavior, T-Mobile USA has started to include social relations between subscribers in its churn management model. The organization usesa multi-graph technique, similar to the methods used in social network analysis, to identify so-called “tribeleaders”. These are people who have a strong influence in larger, connected groups. If a tribe leader switches to a competitor’s service, it is likely that a number of their friends and family members will also switch; it is like a domino effect. With this change in the way it calculates customer value, T-Mobile has enhanced its measurement to include not just a customer’s lifetime subscription pending on mobile services but also the size of his or her social network or “tribe” (see Figure 8).
Benefits
Data
"Creating this completely new perspective of its customers required T-Mobile to enrich its legacy analysis of data (historically taken from billing systems and communications
network elements). In addition, roughly one petabyte of raw data – including information from web clickstreams and social networks – is now ingested to help track down the sophisticated mechanisms behind customer churn."