Enabled customer retention by understanding buying patterns for a technology giant

Case Studies:Mu Sigma
Published On: 17 December 2015
Views: 1794

What We Did: With Mu Sigma’s decision sciences workbench- muRxTM, a robust framework was developed which enabled the client to compare buying pattern of accounts that were at risk of lapse.

 

The Impact We Made: The client was able to retain 3% more customers than before by proactively addressing a possible deviation from an account's buying pattern.

 

Summary - Understanding customer buying patterns

Client’s commercial business division caters to a diverse set of customers (across various dimensions such as the line of business, type of industry, etc.) with varying buying patterns. The existing methodology used to understand customers' buying patterns could not derive actionable insights. The client suffered a loss of $3.4 BN due to customers lost in first half of 2014. Hence, the business approached Mu Sigma to solve this problem.

 

About The Client - Leading technology company

The client is one of the largest technology giant and is a Fortune 500 organization. It has more than 80,000 employees worldwide and a revenue of more than $120 Bn. It is present in multiple countries and has many stores.

 

The Challenge - Legacy framework

The Business’ existing framework of tracking customer purchase behavior was unable to capture insights for the diverse set of customers. As a result, the variation in customers' buying pattern across business units, type of industry, etc. was not observed for anomalies, if any. Lack of this information led to customers lapsing without any prior information, making it challenging for the business to retain them.

 

The Approach – Application of a decision science workbench

Mu Sigma used its decision science workbench - muRxTM, to solve this business problem.

  • The team built a Linear Regression model on the active customer account base to predict next purchase

  • The team also built a Logistic Regression model on the inactive customer account base to predict the probability of next purchase

  • These models were built on 4.5 years of data and validated on 1 year of data

  • As a result, 9 significant drivers were identified for calculating actual order gaps like recency, average revenue in one year, number of transactions, etc. 

 

The Outcome – Improved customer retention

  • Post model processing using muRxTM helped in knowing when an account is likely to make a purchase. For inactive customers, based on the probability of purchase, business was able to plan promotions for reactivation.

  • Customer who deviated from their buying pattern and those who were at a risk of lapsing were highlighted

  • The client understood which are the factors responsible for customer reactivation to plan promotions accordingly

  • Interestingly, the client was now able to retain 3% more customers as compared to what they could retain with their existing framework

Back to Top