CS-08 — Automation & AI · Cloud architecture
Customer analytics and churn prediction on AWS
Analysis of customer usage behaviour, segmentation, cross-selling opportunities and churn prediction models, with analytics infrastructure on AWS.
- My role
- Data Engineer and researcher (IKERLAN).
- Period
- Not documented
- Disclosure
- Anonymised details
- Technologies
- PythonJupyterAmazon SageMakerAWS GlueAmazon S3AWS LambdaAmazon SNSApache Kafka
- Last reviewed
- 13 November 2026
Context
A client with a large volume of usage and behaviour data wanted to understand how its services and products were used, segment its customers, find cross-selling opportunities and anticipate churn.
Challenges
- 01Extracting useful consumption patterns from telemetry and usage data.
- 02Selecting and building predictive features.
- 03Evaluating the models by their usefulness to the business.
Solution & architecture
Exploratory and behavioural analysis, segmentation through clustering, cross-selling analysis, feature engineering and selection, and classification models to predict churn, supporting the design of an analytics infrastructure on AWS (SageMaker, Glue, S3, Lambda, SNS) with data integration via Kafka.
Key decisions
Not documented yet
Trade-offs
Not documented yet
Results
Results will be published once verified
Lessons learned
Not documented yet
The exact contribution of each AWS service is still to be validated, so no detailed architecture diagram is presented. This project connected engineering with Ander's interest in economics and commercial analytics.
Ask about this case study
Questions here are answered with the case study “Customer analytics and churn prediction on AWS” as context.