Machine Learning Based Underwriting Models
Prepare for Nextgen models
Overview
- A recent Economist Intelligence Unit survey of 208 bank risk management executives yielded a surprising result: Over half of these senior retail, commercial and investment bankers say they lack sufficient data to support robust risk management.
- Till now traditional and counterparty transaction made up majority of the data now, unstructured data – text, images, audio etc. – represents a source of additional insights
- Big Data Analytics can help Banks in general and Risk management to gain more accurate and real-time intelligence, drawn from a variety of sources
- Big Data Analytics can allow banks to detect potential risks faster, react faster and make robust informed decisions based on thousands of risk factors
- Several startups are using social network data to score customers based on credit quality (ex. Zest Finance, Kreditech etc.) and some (ex. LendUp and Lendo) provide loan services based on social networking data
- Big Data Analytics helps banks in targeting specific customer segments by combining data from sources such as past buying behavior, demographics, real tie location, sentiment analysis from social media etc. This also helps marketing professionals for making informed and more accurate decisions
Case Studies
Case Study I
A mid-sized European bank used data sets of over 2 million customers with over 200 variables to create a model that predicts the probability of churn for each customer. An automated scorecard with multiple logistic regression models and decision trees calculated the probability of churn for each customer. Through early identification of churn risks, an outflow of nearly 30 million per year was avoided
Case Study II
Impact of using advanced, predictive analytics on marketing effectiveness for a leading European bank.
The bank shifted from a model where it relied solely on internal customer data in building marketing campaigns, to one where it merged internal and external data sets and applied advanced analytics techniques to this combined data set. As a result of this shift, the bank was able to identify and qualify its target customers better. In fact, conversion rates of prospects increased by as much as seven times.
Case Study III
A European bank built a ‘propensity to save’ model that predicts the probability of its customer base to invest in savings products, which in turn leads to increased cross-selling. The input to this model included data sets of 1.5 million customers with over 40 variables. The analytics team tested over 50 hypotheses through logistic regression propensity models to calculate the probability of savings for each customer. The pilot branches where this model was implemented witnessed a 10x increase in sales and a 200% growth in conversion rate over a two-month period compared to a reference group 17.
