‘Banks need to implement analytical modeling systems’

How are social media inputs being employed by the banking sector for decision-making? Also, to what extent are they influencing the bank's decision?

Social media is a platform to interact and understand customer behavior, which presently is being experimented by large retailers and FMCG manufacturers. However today, financial organizations are seeing the potentiality of social media as an interactive medium to connect with their customers, to market their products, and improve customer loyalty. The banking sector is now opening to the new platform where the generation is entrusting advices in this networking platform rather than the marketing professionals.

The marketing and customer support functions of a bank nowadays are greatly influenced by discussions and comments made on public domain/forums. However the use of social media inputs in the banks is at an initial stage, and the responses are mostly reactive with customer complaints and negative feedback on products and services. The banks need to take a step to be more proactive and implement an approach where predictive capabilities are developed.

What kind of analytical models have you employed? It has been observed that implementation of these models and ongoing sustenance of the models to keep on improving their predictor power is a big challenge for the banks. How are banks dealing with this challenge?

Our analytical models can be grouped under various categories such as acquisition (market analysis, offer testing, response modeling, channel analysis), customer development (activation, cross sell, profitability, campaign effectiveness, risk based pricing, customer segmentation, life time value), retention (churn prediction, reactivation, customer loyalty), and risk analysis (fraud prediction, behavior scoring for PD, collection strategy, roll rate analysis, sensitivity analysis).

The primary challenge faced by the banks while implementing analytical model is to derive sufficient data from authentic sources, to ensure predictive ability of the model. For example, banks may not store rejected applications, but these data can become a key data while developing an application score, for understanding the historical credit policies and rules associated with approval.

The second main shortcoming is that the banks do not consider to monitor the performance of the model at regular intervals. It has been observed, once a model is developed and deployed, most banks do not monitor its performance, as the concept of keeping a check on profitability and portfolio with up-to-date data, is bit uncommon. In order to perform model monitoring on a regular basis, the banks need to frequently extract data from back-end systems which is a cumbersome task in absence of enterprise data warehouses, as it consumes time and resources.

In order to overcome the challenges, the banks need to implement analytical systems that are capable of retaining data and have an integrated rating engine that can periodically assess the performance of the model and provide necessary recommendations based on performance assessment. n