← BACK TO PROJECTS
Creative Galileo

Predicting Drop-Off and Fixing Slow Spots in a 10M-Download App

An early-stage EdTech (Series A) with 10M+ downloads was losing users at the payment page.

10% faster load · 12% less lost-payment rate
HOW THE SYSTEM WORKS
DATA
AWS S3 telemetry
behavior from a 10M-download app
STEP
Churn model
scores every user's risk of leaving
STEP
Bottleneck ranking
finds the slowest funnel steps
OUT
Fixes that stick
−10% load time · −12% payment churn
App telemetry reveals who is about to leave and which slow screens push them out.
Built
Churn prediction models + user-behavior diagnostics surfacing slow spots in the app.
Approach
scikit-learn + AWS SageMaker for risk scoring; S3 user-behavior data analysis for funnel drop-off ranking.
Impact
10% faster load · 12% less lost-payment rate
Skills
scikit-learnSageMakerAWS S3Telemetry