Running production experiments with AWS AppConfig experimentation
A detailed walkthrough of AWS AppConfig experimentation, which layers A/B testing on top of AWS AppConfig feature flags to let teams test changes against real production traffic without standing up a separate experimentation platform. Two full examples are built: a frontend test of a redesigned 'Add to cart' button (via console) and a backend cache TTL comparison in an Amazon ECS service (via CLI and Java/Spring Boot). The post covers architecture (control plane, delivery via the AppConfig Agent, measurement, and safety), capturing treatment assignment logs through CloudWatch Logs and Data Firehose into S3, setting up CloudWatch alarm guardrails, gradual exposure ramping, analyzing results with post-exposure attribution using Athena SQL queries, promoting a winning treatment without a gap in service, and cost/cleanup considerations for experiment-run billing.


