Skip to main content
< Back to all work

Improving Customer Retention Through Machine Learning Insights

For a high-touch insurance company with a highly localized customer base, policy renewals are essential to continued stability and success. Indiana Farm Bureau Insurance has over 200,000 active polices but struggled to identify at-risk policyholders and reached out to us to find a solution. They turned to Fusion Alliance for assistance in predicting attrition.
JUMP TO OUTCOMES
Indy Farm Case Study Hero Image

Challenges

Unbalanced data
Retention data is very unbalanced. An overwhelming majority of policyholders do renew their policies and finding the specific data that would predict such a small volume of attrition is complicated.
Uncontrollable variables
Some policyholder attrition is predictable based on factors like customer complaints and demographic data. However, a large percentage of attrition is caused by random, uncontrollable issues such as a policyholder dying or moving out of state where coverage isn’t available.
Model selection
Our team had to decide the right model selection to base a machine learning algorithm. We looked at both the household and policy levels and chose the household level to get the most comprehensive dataset.

Solutions 

Indiana Farm Bureau Insurance worked with Fusion Alliance to build a machine learning model that would capture the likelihood a given household will have an active policy or still be a customer within 30, 60, and 90 days. 

Project goal

Fast setup of machine learning  

We wanted to provide Indiana Farm Bureau Insurance with accurate, actionable insight quickly.  
Our solution
  • Chose practical use cases with small but significant results 
  • Vetted use cases to drive maximum predictive value with minimal risk 
  • Automated machine learning technology instead of bespoke algorithm development 
Project Goal

Capture accurate, actionable results

Indiana Farm Bureau Insurance wanted to provide the customer retention team with insight on who to contact and how to position retention strategies. We needed to quickly capture accurate, actionable results.
Our solution
  • Built two machine learning modules including deep neural network and traditional statistical analytics 
  • Selected concrete questions that would leverage over 3 million policy snapshots and target 35,000 at-risk policyholders 
  • Drove returns through incremental machine learning investment 

Results

Implementing machine learning directly led to  improvements in identifying and connecting with policyholders most at risk of leaving Indiana Farm Bureau. It also builds a foundation to dig deeper into customer behavior in the future. 
2
machine learning models built, including a deep neural network
180
data points available for a given household per year and per month
25%
improvement in at-risk policy holder determination

Impact

By integrating machine learning models into Indiana Farm Bureau Insurance’s existing platform, the team received customer prediction returns in milliseconds. This information was leveraged by the customer retention team to build more effective communications to reduce attrition.  Based on this success, we can continue to optimize the model efficacy to suit extended business needs.

 

Improved Return on Investment 

 

Using an incremental approach to machine learning while focusing on practical use cases ensures a higher return on investment and faster results. This method lays the groundwork for a second phase in which we will enrich the existing modules to improve actionability and provide greater insight and empowerment to the customer retention team.  

In spite of highly unbalanced data and uncontrolled variables, Fusion created a solution that will improve policy retention and relationships with long-term customers, creating a framework that can continue to scale and improve over time.   

Recent work

Ready to talk?

Let us know how we can help you out, and one of our experts will be in touch right away.