Predictive Modeling: Customer Attrition and Retention Scenario for Financial Services

Author: Ram and Amitayu


It is well established now that retaining an existing customer is a lot easier than acquiring a new customer. Typically, acquiring a new credit card customer cost $50-$80. Surprising though, a lot of banks and financial services organizations still focus a lot more on customer acquisition programs. It could be due to multiple reasons but one of the most important reasons is that the KPI used for measuring business success is still customer counts and not the value of customers.

There are multiple ways of increasing share of wallets (SOW) from the existing customers and a few levers are
• Increasing balance and spend from the credit card and transaction accounts
• Cross-selling diverse products to for augmenting product holding of the customers
• Retaining and engaging customers for sustainable from value from the customers


A Scenario: Retaining Customers

One of the leading US banking and financial institution was looking to leverage its massive cardholder base to expand its business by offering an enhanced suite of new-age retail banking products (Cross Sell Scenario at that point in time) such as Certificate of Deposit (CD) and Online Savings.

Launched a few years ago, these products were designed primarily to cater to the deposit needs of the customer. The focus was centered on two products– Certificate of Deposit (CD) and Online Savings which were considered strategic from the bank’s perspective in the given context. While the latter was in high demand, the former was losing ground and popularity owing to the low interest rates offered for CDs across the US.

A high volume (both in count and $) of CDs was up for maturity and the bank management was looking for vital insights on the possible fate of the huge funds invested in CDs. Due to relatively unstable future around interest rate scenario, the CDs have higher risk of fund erosion. And the expected risk to fund movement was not from similar products offered by the competitors, but online savings products from competitors.

Online Saver offers almost comparable interest rate and additionally flexibility around terms. Thus, it was important to retain these funds into the banks own online savings product. Since all CD holders had an online savings account linked with the CD, hence it was not a problem of cross-Sell. The online saver could be used as a retention proposition to the potential attritors.

Online savings market in the US was booming, fairly competitive and all the financial services organizations were offering attractive rates to retain customers. As a result, customer retention was of paramount importance for both the products, which made it imperative to assess the likelihood of customer attrition for both products.  A statistical predictive model was built to identify prospective attrition customers and understand attributes of potential attrition customers; key for formulating right retention strategies.

Customer Attrition Modeling Approach Used

Having studied the business problem, we offered to build statistical models using regression methodology to predict the likelihood of customer attrition from these portfolios over a given period of time in the future. The model, would identify the most risky and vulnerable customers in advance, and allow the business to implement preventive strategies to retain them. The model would also provide vital insights into the underlying factors causing attrition, and help business strategize their retention plan.

For the CD a churn by definition was simply a non-renewal, however if the non-renewal was followed by a transfer of at least 70% of the maturity amount into the Savings Account then that particular case would not be deemed churn.

For the Online savings the definition was relatively simple, a churn could either be a hard attrition where the bank loses the account completely or even an 80% balance drop within the performance period would be considered a case of attrition.

The model was built on historical data of 2 years provided by the bank, using multiple statistical techniques.  All accounts considered for the model development sample had to satisfy a number of selection and exclusion criteria on the basis of their vintage with the bank, type of account ownership.  A lag period of 2 months was incorporated between the observation window and the performance window, to allow business to take necessary retention actions once they have implemented the model and prevent a prospective churn. Data used, consisted of transaction history, static customer information, customer product holding, customer demographics etc

Results and Outcome

Some of the key attributes from the Savings model which had significant contribution in causing a churn came from the transaction behavior of the customers, primarily their RFM attributes. For CD the key factors were mostly the customer behavior demonstrated in the last few renewals, primarily the number and amount of renewals, and the CD term.

Both the Models went on to generate a fairly high degree of lift with an ability to identify 70% of potential account churn by targeting only 20% of the customers. This eventually translates to a potential cost savings of around US$56 million in customer targeting strategies for the online savings product only. After the pilot the business implementation of the model resulted in Balance retention of more than US$340 million even at a 50% customer retention rate.

As further recommendations we went on to propose newer cross selling strategies to improve the retention rate, followed by an optimization of the response campaigns which will empower business with an end-to-end analytical solution on CRM

## numbers and scenario are tweaked as objective is to share learning not the client specifics

1 thought on “Predictive Modeling: Customer Attrition and Retention Scenario for Financial Services”

Leave a Comment