Overview
A leading Non-Banking Financial Company (NBFC) in Sri Lanka was facing rising customer churn and lacked the tools to effectively identify and act on early signs of risk. NeenOpal partnered with the client to develop a machine learning–driven churn prediction model that accurately identified at-risk customers, enabling timely and personalized retention strategies.
- 90%
- Accuracy in Predicting Customer Churn
- 20–30%
- Churn Reduction via Targeted Campaigns
- 3x
- ROI Boost from High-Risk Segments
Customer Challenges
As part of their retention efforts, the client faced several challenges that made it hard to manage and reduce customer churn. These issues led to gaps in their strategy and execution, resulting in less effective outcomes.
Lack of a Predictive Framework to Identify and Prioritize At-Risk Customers
Without a system to proactively identify potential churners, the client was unable to detect them early, reducing the impact of re-engagement efforts. Additionally, the absence of risk segmentation made it difficult to prioritize high-risk customer segments, leading to inefficient allocation of retention resources.
Limited Visibility into Churn Drivers
Despite a wealth of historical data, there was no clear understanding of what behaviors or patterns preceded churn, making it hard to design focused strategies.
Generic Engagement Across All Customers
Communications were not tailored based on individual customer risk or behavior, leading to poor impact and missed opportunities to build loyalty.
Solutions
NeenOpal implemented a machine learning model to learn from the historical trends of Customer churn and predict the churn probabilities of the active customers. These Churn probabilities helped the client target the at-risk customers and optimize their retention strategies.
Behavioral Feature Engineering
Customer behavior was analyzed and aggregated monthly to capture evolving trends — such as loan repayment consistency, transaction frequency, and engagement signals — giving the model a dynamic understanding of churn risk.
Balancing the Dataset
As churned customers represented a minority, data imbalance was addressed using suitable techniques to ensure the model could learn effectively from both churned and active customer classes.
Model Transparency with Feature Importance
To make the model explainable to business users, we extracted feature importance scores to reveal which behaviors or patterns were most predictive of churn. This not only improved trust in the model but also provided direct inputs for strategic decisions.
Services
Python
Jupyter Notebook
SQL Server Management Studio (SSMS)
Tableau
Benefits
Early Detection of At-Risk Customers
The churn model flagged early behavioral shifts, such as reduced repayments or activity. These signals enabled timely interventions, boosting the client’s ability to effectively retain vulnerable customer segments.
Targeted Retention Campaigns
With churn insights, the client created personalized messages and loyalty offers for specific groups. This enhanced engagement improved campaign relevance and helped reduce customer churn more efficiently.
Reduced Budget for Retention Activities
By targeting only high-risk segments, the client optimized retention spending. This reduced wasted effort on low-risk customers while maintaining strong outcomes with fewer resources and better focus.
Conclusion
The churn prediction model significantly strengthened the client’s customer retention strategy. By using data to anticipate churn and personalize interventions, the NBFC saw measurable improvements in customer satisfaction, campaign efficiency, and revenue continuity. More importantly, this initiative laid the groundwork for a customer lifecycle management system powered by data-driven intelligence.
FAQ
Get quick answers about how we built a churn prediction solution to improve customer retention.
What challenge was the gold loan provider facing?
The client lacked accurate churn prediction, making it difficult to identify at-risk customers and reduce attrition.
What solution did NeenOpal implement?
NeenOpal developed a machine learning churn prediction model that analyzes customer behavior to forecast churn risk.
What benefits did the solution deliver?
The model enabled proactive retention actions, improved customer engagement, and supported data-driven marketing strategies.
Authors
Case studies
Related case studies
Tableau Dashboard Development for a US Insurance Organization: 4 Data Sources Unified
Our client, an insurance organization with field teams spread across every US state, partnered with NeenOpal to replace a manual, spreadsheet-driven reporting process with a consolidated Tableau dashboard environment governed by row-level security. Operational data lived in four separate systems covering Airtable, Salesforce, Google Analytics 4, and Google Search Console, and bringing them together meant exporting and stitching the data by hand, which left leadership without a single reliable view of performance. NeenOpal consolidated all four sources into an Azure SQL Server layer, reviewed and corrected the data quality issues that surfaced along the way, and delivered three Tableau dashboards across seven screens that refresh automatically every day. Row-level security was implemented across all fifty states and the independent districts, driven by an entitlement model the client maintains through a single Google Sheet, so every representative sees the states they are responsible for and nothing more.
Read case study
Microsoft Fabric Migration with Row-Level Security for Multi-Client Analytics
Our client, a B2B financial wellness and HR technology company, where we have worked on their data that delivers reporting to multiple business customers, partnered with NeenOpal to consolidate a fragmented reporting stack into a single, secure analytics platform. The client's finance and user-engagement data lived across MongoDB, AWS RDS, Google Sheets, and HubSpot, and was surfaced through standalone Power BI dashboards built on AWS. Serving each downstream client meant securely embedding a per-user filtered dashboard for external users logging into the client's portal, and app-owns-data embedding at that scale requires dedicated capacity. On shared, standalone Power BI the team had worked around this by building and maintaining a separate dashboard and semantic model for every client, which increased infrastructure cost and slowed delivery. NeenOpal designed and delivered a four-week Microsoft Fabric migration that unified every source into one governed semantic model with dynamic row-level security. By rebuilding the pipeline on a medallion architecture inside a Fabric Lakehouse and enforcing access with a single Power BI security role, the client replaced many duplicate dashboards with one unified report that automatically shows each user only the data they are permitted to see.
Read case study
How a Leading Gold Loan Provider Eliminated Retention Guesswork with ML-Powered Churn Prediction
Our client is a leading financial institution based in Sri Lanka, providing specialised gold loan products across a retail network of more than 50 branches and serving an active customer base estimated in the hundreds of thousands. The business model is built around short-term, collateral-backed lending, where customer retention at loan maturity directly determines portfolio revenue. Despite years of accumulated transactional data across five core SQL Server tables, the organisation had no mechanism to convert that data into customer-level insight. NeenOpal partnered with the client to build and deploy a machine learning churn prediction system, alongside a customer scoring model and Tableau dashboards, enabling the organisation to shift from reactive, instinct-driven outreach to precision-targeted retention.
Read case study
Enterprise Data Platform on Fabric for Secure, Scalable Analytics
Our client, a large enterprise organization, used Dynamics 365 Finance & Operations (F&O) as its primary ERP system for managing financial and operational processes. Although F&O handled transactional workloads effectively, reporting and analytics were constrained by limited flexibility, performance concerns, and a lack of centralized data access. To enable scalable analytics, governed reporting, and self-service BI, the client adopted Microsoft Fabric as a unified data and analytics platform.
Read case study
Transforming Tax & Wealth Management Onboarding with Microsoft Copilot Studio and Power Platform
Tax and wealth management firms struggle with client onboarding due to manual processes that cause delays and errors. This case study illustrates how we utilized Microsoft Copilot Studio and Power Platform to automate and streamline onboarding. The solution reduces processing time, minimizes errors, and enhances client satisfaction through intelligent document processing and AI-driven investment recommendations.
Read case study
Developing a Credit Scoring Model for a Leading NBFC in Sri Lanka
Our client, a leading NBFC in Sri Lanka, partnered with NeenOpal to enhance its gold loan disbursement process by building a machine learning-based customer scoring model. Instead of traditional credit checks, the model enabled smarter loan-to-value (LTV) decisions, allowing the client to confidently offer higher LTVs to trustworthy customers based on historical gold loan and transaction behavior.
Read case study