Case Study: AppsFlyer achieves faster predictive analytics and real-time insights with Amazon Web Services

AppsFlyer Builds a Predictive Analytics Solution for iOS 14+ Using Amazon SageMaker

AppsFlyer, a mobile attribution and analytics company, needed a way to build predictive analytics for iOS 14+ while handling huge event volumes and protecting customer privacy. Working with Amazon Web Services and Amazon SageMaker, AppsFlyer set out to rapidly move a custom machine learning solution from research into production.

Amazon Web Services helped AppsFlyer implement PredictSK using Amazon SageMaker, AWS Lambda, AWS Step Functions, DynamoDB, and EC2 P3 instances. The result was a scalable, serverless ML pipeline that trains custom models per app, retrains them monthly, and delivers predictions in near real time at 10–30 ms per inference. AWS also enabled AppsFlyer to cut LTV insight time from about 30 days to just a few hours and process hundreds of gigabytes of user data per day.

AppsFlyer

Benjamin Winestein

Senior Software Developer