Why Core Web Vitals Matter: Optimizing Global Front‑End Performance

Introduction

ICBU, a long‑standing overseas digital commerce unit, focuses on global user experience and page performance. In increasingly complex network environments, measuring front‑end performance globally and providing actionable insights for developers is essential.

Performance Metric Collection

What is CWV?

Core Web Vitals (CWV) are a subset of web metrics that every site should measure; they appear in all Google tools and represent distinct aspects of user experience.

CWV includes LCP (Largest Contentful Paint), CLS (Cumulative Layout Shift) and FID (First Input Delay).

Why use FCP + CWV as core metrics?

FCP + CWV provides a unified, industry‑standard measurement, clear grading intervals, and objective guidance for optimization.

Why abandon custom first‑paint metric?

Easy to cheat: manual reporting leads to low data trust.

Lack of consensus: different teams define it differently, making cross‑team comparison impossible.

Optimization difficulty: many factors affect first‑paint time, making root‑cause analysis hard.

Metric too narrow: only measures load time, ignoring interaction latency, layout shift, and element load timing.

Why add FCP on top of CWV?

LCP reflects most loading scenarios, but for short‑lived pages or first‑hop scenarios, the page may leave before LCP fires. Adding FCP captures the moment the first content is rendered, enabling finer‑grained problem diagnosis.

Differences between our thresholds and industry standards

Thresholds are set to balance consensus (most pages should meet them) and feasibility (at least 10% of domains must satisfy the “good” range). Adjustments are made for international scenarios, e.g., FCP thresholds derived from custom first‑paint standards.

Threshold Logic

LCP

Based on the “one‑second principle” and research by Card et al. and Miller, a “good” LCP range is 0.3‑3 s; considering FCP’s good threshold of 1.8 s, the LCP good candidate becomes 1.8‑3 s.

CLS

Internal testing shows values ≥0.15 are disruptive, while ≤0.1 are acceptable; therefore 0.1 is chosen as the “good” CLS threshold to balance experience quality and feasibility.

FID

Studies indicate ~100 ms is perceived as immediate response; 100 ms is set as the “good” FID threshold, with 300 ms as the “poor” threshold.

Performance Data Analysis

Multi‑channel data flow

Data is collected via Web (multiple points) and App (UT channel) and mirrored to SLS for comparison and correction.

Multi‑dimensional aggregation

Metrics are aggregated across 75th percentile, median, and 90th percentile, and filtered by scenario, URL, spm_id, country, and device type.

Why use the 75th percentile?

The 75th percentile balances coverage (most pages meet the target) and robustness (less affected by outliers) compared to higher percentiles.

Performance Report Presentation

Custom dashboards display core metric cards with grading, trends, and optimization suggestions, as well as network load chain diagrams, real‑time trend charts, global traffic distribution, and user performance distribution based on Google CrUX data.

Conclusion

ICBU’s ongoing exploration of global front‑end performance provides a foundation for continuous user‑experience improvement, while recognizing that good metrics do not automatically equal good experience.

References

Defining Core Web Vitals thresholds – web.dev

Cognitive Unified Theory – ACM

Response Times: 3 Important Limits – Nielsen Norman Group

Perceived Causality – APA

Perfect Virtual Button – ACM

Web Vitals Changelog – Chromium

CWV in Google Search Ranking – Google Developers

CWV FAQ – Google Support