A managed service that uses interaction records and catalog data to produce item recommendations, rankings, and user segments for applications.
Last reviewed: 2026-10-03
Amazon Personalize uses your interaction data to generate recommendations and user segments. It offers resources for real-time personalization and batch workflows. Item and user metadata can supplement interaction history, depending on the selected configuration.
Start with a defined recommendation surface and a time-based evaluation split. Compare the output against a simple popularity baseline, including new users and infrequent items. Check unavailable items and business exclusions before any experiment, then measure the chosen user outcome rather than clicks alone.
No. You need to provide suitable datasets and interaction records, then configure the resources for your use case.
Compare with a popularity baseline on a held-out time period. Include coverage, new-user behavior, and your chosen business outcome in the evaluation.