Social Network Analysis for Telecom Business Intelligence and Customer Relationship Management.
SNAzzy - Social Network Analysis for Business Intelligence - is primarily a telecom social network analysis project. SNAzzy analyzes the social network formed by customers of telecom operators and derives non-trivial social network intelligence about the call network. The intelligence is often in terms of gobal structure of the overall call network, automatically discovering communities, and analysing customer churn.
SNAzzy is also capable of analyzing any kind of social network or graph, not just Telecom networks.
Present Contributors
Kuntal Dey
Sougata Mukherjea
Seema Nagar
Amit A. Nanavati
IBM India Research Lab, Delhi
Past Contributors
Natwar Modani
Dipanjan Chakraborty
Koustuv Dasgupta
Rahul Singh
Siva Gurumurthy
Gautam Das
Sambuddha Roy
Vinayaka Pandit
Anupam Joshi
Ruta Mehta
Usha Nandini Raghavan
Project Description
Telecom Business Intelligence today grossly relies on bills generated by individual customers to measure the importance of the customer to the operator's business. Conventionally the more the amount of the direct revenue generated, the more valuable is the customer perceived to be. While it is indeed true that direct revenue generation contributes a significant proportion to the operator's bottom line, the conventional profitability measurement methodology fails to identify the indirect revenue generators and the impact these customers have in the overall network of voice and text (SMS) calls. As an obvious consequence the importance assigned to an indirect revenue generating customer does not necessarily reflect the true value of the customer to the operator's network and revenue stream. Furthermore, while targeting to rope in customers from other networks, the marketing teams today do not leverage the social relationships of targeted foreign-network customers with the telecom operator's own network's customers. In addition, customer churn remains a problem with many of the telecom oeprators. This is particularly true to a high degree in mature and saturated markets where almost every customer has a mobile phone connection, and is also becoming a rising concern for emerging markets, calling for attention to arrest the problem.
Analysis of Telecom Call social network graphs often provide insigts for telecom operators to measure the true value of a customer. The telecom call graph analyses are computationally expensive, and yet, as mentioned earlier, plays a key role in finding parameters beyond bill generated to mark the importance of each customer to the telecom player. SNAzzy analyzes the social networks of the customers of a telecom using the Call Detail Records (CDR) of the telecom operator. Using the CDR analyses, SNAzzy comes up with three kinds of core analyses for further consumption by the Business Intelligence team using the business rules. The three kinds of core analyses capabilities of SNAzzy are:
Global Structure Analyses
Community Discovery
Customer Churn Prediction
In the Global Structure Analysis (GSA) phase, the telecom call graph is viewed from a telescopic level. Global parameters such as indegree distribution, outdegree distribution, strongly connected components, people's importance rank distribution, graph diameter, approximate number of bipartite cores etc. are found from the call graph.
In the community discovery phase, the different structural communities formed by the voice and text (SMS) call graph are figured out. Among these structures are cliques (where every member of a community communicates with every other member), stars (with one person in the center - the hub - communicating with a number of people on the fringe - the spokes - where the people at the fringe almost never communicate between each other) and different kinds of dense subgraphs within the graph. Members of structural call communities belonging to foreign/competetor networks are identified and SNAzzy provides a way to rank thus found foreign network members in context of the telecom service provider's acquisiton strategies.
In the churn prediction phase, SNAzzy uses the recent churner list as well as the CDR for the recent past month to predict the churners for the upcoming month along with a score of churn propensity. The higher the churn propensity score, the higher is the chance of churn of the customer rooted at social causes. The list of customers such obtained, along with the churn propensity, can be catered to the telecom business intelligence team to arrest/minimize churn from the potential churner.
Last updated 16 Apr 2009