We deliver measurable results
Get to know the client
The client was owned by one of Vietnam’s biggest private conglomerates. Through strategic partnerships with top-tier brands, the platform offers users access to exclusive deals and benefits from thousands of popular businesses.
Leveraging rich, multidimensional customer data, the client helps its partners better understand and engage their target audiences by delivering highly personalized, in-app experiences tailored to individual user behavior and preferences.
Challenges
They faced challenges with
The client was facing challenges in retaining its user base. Despite having millions of customers, the company struggled with high churn rates and ineffective retention strategies that limited customer lifetime value (CLV) and long-term growth. The organization needed a predictive, data-driven approach to improve engagement and optimize retention campaigns to fight against:
High Customer Churn
Many users became inactive after their initial interactions, significantly impacting overall revenue and long-term loyalty.
Reactive Retention Campaigns
Marketing efforts were launched too late, often after users had already disengaged due to the lack of predictive analytics.
Data Scalability Limitations
The existing data infrastructure couldn’t process large volumes of behavioral data in real time, hampering personalization and insight generation.
Inefficient Budget Allocation
Retention campaigns targeted broad audiences rather than high-risk segments, leading to poor ROI and missed opportunities.
SOLUTIONS
Our tools for success
Advanced Data Processing & Feature Engineering
Using Cloudera Data Platform, we integrated and processed large-scale user data, including:
– Purchase frequency and recency
– Clickstream and app engagement
– Reward redemption history
– Customer service interactions
Machine Learning-Based Churn Prediction
We built a robust machine learning pipeline on Cloudera Machine Learning (CML), enabling:
– Data Preprocessing: Structured and cleaned behavioral and transactional data
– Model Training & Evaluation: Benchmarked Random Forest, XGBoost, and LSTM models to identify the most accurate predictors of churn
– Real-Time Churn Scoring: Deployed models using Apache Spark to assign churn risk scores dynamically
– Ongoing Monitoring: Continuously retrained models with fresh data to improve predictive accuracy over time
