Efficient Optimization of JD App Review API: Reducing Peak Requests by 85% without Impacting User Experience

The JD app’s review section originally fetched data via a dedicated API that mirrored product‑detail traffic, causing a 1:1 request ratio and exposing the system to sudden traffic spikes during promotions.

Analysis of user behavior and exposure data revealed that the review floor rarely appears on the first screen, making it feasible to decouple its requests from the main product‑detail flow.

The engineering team devised a solution: trigger the review API only when the preceding floor becomes visible or when the user scrolls a configurable distance (default 10 px). A fallback “bottom‑floor” trigger ensures that reviews are still loaded if they appear on the first screen.

Implementation included asynchronous loading, experiment‑driven roll‑out, and graceful degradation to maintain functionality if the new feature is disabled. Monitoring confirmed that the approach did not affect perceived latency (TP99 ≈ 50 ms).

After two version releases and extensive testing, the optimization reduced peak QPS of the review API by up to 85% during major sales (e.g., Double 11), while preserving a smooth user experience.

The case study demonstrates the effectiveness of behavior‑driven request gating, configurable triggers, and robust fallback mechanisms for large‑scale mobile performance engineering.