Predicting Mobile Value-Add-Services Likelihood

Project Details


2012 (Nov)


Sagar Gupta, Raghuraman Chandrasekhar, Yash Chandwani, Sudhanshu Dharmadhikari, Saurabh Choudhary





The Indian mobile telephony market is a classic example of a volume based strategy. Average revenue per user (ARPU) is one of the lowest, while the subscriber figures are second only to China. The market itself is highly fragmented with more than 5 players holding less than 90% of the market. In this scenario, identifying new sources of revenues is critical for survival.

This study focuses on identifying potential customers for the mobile value added services (MVAS) – an increasingly significant revenue driver for Indian operators. As most customers have already adopted MVAS in some form, identifying pockets of opportunity while maintaining accuracy of prediction is essential to reduce promotion costs.

Post data preparation and standardization, two data mining approaches were adopted, Logistic Regression and K Nearest Neighbor (KNN). Lift and Decile charts were used to compare the performance of the models. Benchmarking of the model was done based on the Naïve Majority class for comparison with the output of our selected model The study shows that smartphone ownership and monthly mobile expenditure are major
influencers for MVAS adoption and the fact should be central to the promotion strategy.

Application Area: