Customer Churn prediction for a Fortune 100 financial corporation
Client Profile
The client is a Fortune 100 financial services company headquartered in the USA, serving 70 million plus customers globally. Employing over 200,000 people, it has its operations in 35 countries, covering its 8000 branches across the globe.
Business Problem
The client was losing its customers to fintech startups. Startups pursue analytics initiatives to perfectly deal with customer dynamics acquisition, service and retention. The client, therefore, was in a dire need of a strong analytics and business intelligence (BI) framework that could allow it to proactively reach out to customers who are at a risk of leaving. This would allow it to maintain a stable customer base, while enhancing its customer service strategies.
Key Challenges
- Data existed in silos and needed to be converted into an integrated repository.
- Different types of missing values and outliers demanded application of contextual imputation techniques.
- Analytics workflows required to be mapped with operational procedures.
- Data analytics had to be executed within the framework of regulatory norms
- Business Intelligence and analytics capabilities had to be built from scratch
Technologies Used
Python, MS Power BI, MS Azure Cloud
Solution
- A robust classification model was built using data fields such as Customer Data, Call Center Records, Server Logs, Web Click Stream etc. to form customer segments of risky clients.
- The Python model was integrated with Microsoft Power BI visual dashboard to let the decision makers get real-time insights into current status of customer churn, sales, service efficiency and campaign effectiveness.
- Using MS Azure Cloud, the dashboards were made accessible over mobile devices for stakeholders, so that they can tap in the trends at any point of time.
Business Impact
- Likelihood of customer churn was predicted with 99% accuracy.
- Realizing that customer age, number of complaints, existing account balance and number of products had highest influence on customer churn rate, the client tweaked its existing strategies to better address these factors.
- Analytics offered key insights to reengineer campaigns for younger customers and those with low account balance.
- Regular BI insights paved a way for conceptualization of separate products and strategies for female customers in the 25-35 age band, who were more likely to quit.
- Over the period of 6 months/12 months, the customer churn rate was reduced by 28%.