Front‑end Performance Optimization: Testing, Analysis, and Sustainable Improvement Loop

Front‑end performance directly affects user experience and business conversion. Studies show that a page load time beyond 3 seconds can cause a 53 % user drop‑off, and each additional second can increase bounce rate by 32 %–90 %.

To proactively improve performance, the team adopts a three‑step “discover‑analyze‑validate” workflow driven by testing. First, a performance blacklist (black‑list) identifies pages with the worst metrics (e.g., first‑screen average time, second‑open rate) combined with traffic volume. Second, detailed analysis—splitting front‑end and back‑end aspects—pinpoints bottlenecks such as long API latency, heavy LBS calls, or slow resource loading. Third, targeted optimizations are applied, and the effect is verified through continuous trend monitoring.

Key metrics are defined from the user perspective: interactive time (average time to full load and user interaction), second‑open rate (percentage of pages opened within 1 s), and 5‑second‑plus rate (percentage of pages taking longer than 5 s). Data are collected via ARMS front‑end monitoring, stored in SLS logs, aggregated in MaxCompute, and visualized on the FBI platform.

The sustainable loop consists of:

Embedding a universal SDK for automatic data reporting.

Using the blacklist to prioritize pages.

Collaborating with front‑end engineers to analyze and fix bottlenecks.

Monitoring performance trends to confirm improvement.

Repeating the cycle when new regressions appear.

Analysis examples include terminal‑specific breakdowns (e.g., Alipay vs. other platforms), stage‑wise timing (t1‑t9), and high‑/low‑end device segmentation, revealing that API latency and LBS retrieval dominate the delay on the A‑channel page. Back‑end trace shows request‑response times far exceeding the ideal threshold.

Verification is performed by comparing before‑and‑after trend charts, ensuring quantifiable gains and early detection of regressions. The approach has been applied since October 2021, uncovering eight major performance issue categories and delivering measurable improvements.