Core Web Vitals audit: turn measurements into fixes

Audit LCP, INP and CLS using field evidence, reproducible lab diagnosis and template-level priorities rather than a single performance score.

A Core Web Vitals audit should connect user experience measurements to specific engineering changes. A single test score cannot explain which visitors are affected or why. Begin with representative page templates, devices and journeys. Separate real-user field data from controlled lab tests: field data describes observed experience, while lab tools help reproduce and investigate a cause under known conditions.

Read the metrics in context

The current Core Web Vitals are Largest Contentful Paint, Interaction to Next Paint and Cumulative Layout Shift. Google’s guidance uses the 75th percentile to assess field performance, with good thresholds of LCP at or below 2.5 seconds, INP at or below 200 milliseconds and CLS at or below 0.1. Record the source and collection period. A page without sufficient field data should be treated as an evidence gap, not declared fast by default.

Trace the dominant cause

For LCP, identify the actual largest element and inspect the path to displaying it: server response, resource discovery, transfer and rendering. For INP, reproduce the affected interaction and inspect long tasks, event handlers and rendering work. For CLS, locate elements that move after initial layout, such as images without reserved space or late content insertion. Diagnose the relevant template rather than applying generic changes to unrelated pages.

Prioritise a small set of repairs

• Choose templates with meaningful traffic and a demonstrated user-experience problem.
• Write one hypothesis per change, such as earlier discovery of the hero image or less work during a filter interaction.
• Validate functional behaviour, accessibility and visual layout alongside the performance measurement.
• Compare repeatable lab runs and allow the field collection period to reflect the released change.

Report impact and remaining uncertainty

Record the baseline, environment, affected pages and evidence behind each recommendation. Distinguish a lab improvement from a confirmed field improvement. Third-party scripts, device mix and network conditions can affect results, so avoid promising a universal score. Assign an owner to performance budgets for images, scripts and future page templates. The audit is useful when it produces reproducible fixes and a way to detect regressions, rather than a report that becomes obsolete after the next content update.

• Performance and modernization →
• Next.js Performance: Diagnosing Slow Pages and Large Bundles →
• Website Migration SEO Checklist for Developers →