Hyper-personalisation with an end-to-end machine learning platform
“The Melio team helped us to identify not only solution-specific recommendations to our architecture but also broader bank-wide improvements.”
Head of Data Science - Large Bank
Case Study
Industry
Banking
Company Size
3000+ employees
Problem Space
Marketing analytics
Client value management
Data-driven recommendations
Machine learning
Generative AI
Technology Used
AWS SageMaker
AWS Bedrock
AWS S3
Gitlab
Python
Business Context
A large bank within South Africa was looking to deliver hyper-personalised product and service recommendations for their retail clients. The existing methods to engage with clients were outdated and manual, requiring extensive time to plan and deliver marketing campaigns for clients across basic segmentation profiles. The bank is on a high-growth trajectory with substantial increases in client growth and product offerings being planned for release into the financial services market in the near term.
Challenge
The technical complexity in delivering hyper-personalised recommendation
The bank's data science team was tasked with helping the bank improve revenue using more advanced analytical methods. This included the need for personalised product and service recommendations using machine learning and generative AI technologies.The data science team faced several challenges in meeting the needs of the business including:
- Increasing machine learning and MLOps skills and capabilities to deliver the solution
- Developing a custom sequential-aware next best product recommendation engine
- Deploying the ML recommendation solution on bank’s AWS cloud platform
- Integrating a generative AI (Large Language Model) into the recommendation engine to create personalised, channel-specific client messaging.
Solution
An Engine of Growth with ML and Gen AI
Melio was engaged by the head of data science to assist them in resolving their challenges and developing the initial product recommendation engine. Melio was responsible for:
- Architecting and implementing the SageMaker ML platform that would be used to build and deploy the recommendation engine
- Designing and implementing the multi-layered client recommendation engine
- Integrating a generative AI LLM using AWS Bedrock into the recommendation engine architecture
- Designing and applying an ROI framework to measure the expected and actual value delivered by the recommendation engine
- Set up technical playbook and engineering template as best practice guides for the data science team
The fast feedback loop improves efficiency & ROI
Our collaboration with the bank resulted in a transformative machine learning solution that significantly enhanced cross and up-sell engagements:
🌟
Improved Revenue: Achieved ~30% revenue improvement through targeted up-sell and cross-sell engagements, powered by ML recommendations.
💼
Reduced Client Churn: Witnessed a substantial drop in client churn due to more personalised, relevant client communication enabled by enhanced data quality.
📈
Proof-of-Value: Demonstrated the tangible benefits of AWS services, particularly SageMaker ML and Bedrock for generative AI content, paving the way for impactful AI adoption.
💪
Skill Enhancement: Accelerated ML development and deployment skills within the bank's data science team, fostering a more proficient and capable workforce.