Learning from Machines About Attrition

Spending money to acquire customers who leave the bank is not only frustrating it's also expensive! Once effective onboarding and cross-selling systems are installed, financial institutions need a way to predict which customers are at risk of leaving so they can take decisive actions to keep attractive—i.e., profitable—customers. It's a three-step process:

  • Predict who is at risk of leaving  
  • Trigger retention activities
  • Measure the effectiveness of retention activities 

In this post we'll talk about a few things that machine-learning can tell us about who is likely to leave a bank. 

Figure 1 plots a few characteristics that differentiate customers who are at high and low risk of attrition.

  Figure 1 - Machine learning on >50,000 accounts, values listed are the average of the top and bottom 35 accounts as ranked by attrition risk.

 Figure 1 - Machine learning on >50,000 accounts, values listed are the average of the top and bottom 35 accounts as ranked by attrition risk.

The plot shows the following common characteristics of customers at high risk of attrition:

  1. Few accounts
  2. Few services
  3. Rarely have direct deposit
  4. Low or negative checking balance
  5. Low tenure 

The findings from machine learning square with what a banker would intuitively expect, so we are confident we have a good machine here that will systematically process customer data and identify customers at risk of leaving so notifications can be sent to the front lines to execute retention activities.    

In summary,  it's imperative to install effective onboarding and cross-selling systems to reduce attrition risk. Simultaneously, banks need to leverage machine learning to identify attractive customers at risk of attrition and take immediate and aggressive steps to retain them, which is only possible with machine learning. If you'd like to learn more about how to retain more customers, get in touch with us—we'll show you what we've learned.