As India moves towards mobile portability, it is one reality mobile operators will have to live with in the coming years. While some in the industry fear that mobile portability would further aggravate churning which is already affecting them, others feel it would prompt operators to improve their services, offering the customers the comfort of staying with them while they actually have the choice to leave.

Indeed, now operators have an advanced technology to achieve this: predictive analytics - mathematical modelling done with advanced computing technology. Advanced analytical tools emerging today are capable of building the profile of individual customers and help operators identify those who show signs of leaving them.

“Such tools can analyse a large number of data,” says Mr Krishna Mohan, Founder and CEO, Fifth C Solutions, a Bangalore-based company that offers mathematical modelling-based predictive analytics solutions to telecom industry.

“If we deal with a few thousands of bits of data, we may perhaps handle it with analysts. But, as you can imagine, mobile phone operators handle millions of data bits and so they need reliable tools to track the profile of customers showing signs of leaving them for other operators,” Mohan said.

Math language

Theoretically, mathematical modelling is nothing but an abstract that describes a system using the language of mathematics. It is extensively used in natural sciences and in engineering disciplines like physics and electrical engineering.

“Statistical techniques have been used by the medical fraternity for a long time to predict the nature and impact of a disease,” explains Dr Hans Mathews, Fifth C’s Chief Scientist.

“Now that we have very advanced technology especially things like unique Artificial Intelligence (AI) that allows robots to perform precise tasks, the mathematical modelling techniques can be converted to algorithms that can produce accurate predictions of behaviour based on data,” adds Mathews.

Scoring over risks

Description of a system — people in this case — is done by a set of variables that represent its qualities. In telecom, for instance, details such as billing and payment history, call usage patterns (such as type, duration, volume and call destinations), customer complaint calls to the service provider, etc are taken for churn prediction. The fully automated process then assesses the risk factor of each customer and gives out a “risk score”.

“The risk score is an indication of a customer’s propensity to leave the service of the mobile operator,” Dr Mathews says.

“In addition, they are also assigned the precise risk factor that would explain how a customer is given a high or low score. The factors could be ‘sticker shock’ on a recent high bill or a recent rental plan change,” he adds.

This, he says, will help the mobile operator’s retention programme to be focused on the most vulnerable customers. The challenge for scientists, according to Dr Mathews, is identifying the variables which are vital for the predictions to be accurate.

Mr Jagdish Kini, a director at Fifth C, says analytics technology offers mobile phone companies the ability to “slice and dice” data fast and discern vital information. “They can do this with precision,” he adds. 

He says the technology can also be used in predicting behaviours of credit card users and shoppers, which would enable service providers to fine-tune their marketing efforts.

Once the ‘wavered’ customer is identified using the predictive analytics, experts say that promo activities and other specialised strategies can be focussed on those customers. This is especially possible with the availability of mobile phones and e-mails, which serve as a personalised platform for marketers.