ML-powered lead scoring for financial marketing campaigns

The challenge

A US-based provider of innovative marketing services to banks and financial institutions was using ML-powered direct mail lists for lead scoring — ranking prospective customers most likely to convert. However, the ML capabilities were outsourced to an external vendor whose process was a complete "black box" — no transparency into how models scored leads, no ability to iterate on algorithms, and questionable accuracy that couldn't be independently verified. The client wanted to bring ML capabilities fully in-house with complete ownership, understanding, and the ability to continuously improve their models.

Our solution

We began with a deep dive into the client's business context, identifying key priorities, risks, and constraints. After evaluating the existing vendor's models and mapping all data sources, we designed and built an end-to-end AWS-native ML pipeline: raw customer data flows from S3 through AWS Glue for ETL processing, then into SageMaker for model training and inference. Our models outperformed the previous vendor's accuracy by 15%. Critically, we completed a full knowledge transfer — comprehensive documentation, team training sessions, and hands-on workshops — so the client's engineering team could independently operate, maintain, and improve the ML pipeline going forward.

Key highlights

Tech stack