The performance score comes from a controlled Lighthouse run. The Core Web Vitals assessment uses real-user experiences collected over the preceding 28 days. A successful test today can therefore sit beside a failing historical assessment without either result being wrong. 12
Read the assessment above the score
Check whether field data describes this URL or the whole origin, and compare the same device type. Identify the failing LCP, INP or CLS metric. Origin data can include other pages. 1
If you are checking Search Console as well, remember that its Core Web Vitals report groups similar URLs. A reported group needs investigation across representative pages, not just another homepage test. 3
Test what happens after the page loads
A normal Lighthouse page-load run does not reproduce every interaction. A menu can respond slowly after loading, or content further down the page can shift when it appears. These experiences can affect INP or CLS without being obvious in the initial lab result. 2
For a practical check, open the same page on a phone, use the menu, open a dialog and scroll through the main content. Record the interaction that feels delayed or the element that moves. Then investigate that specific behavior with a performance recording. This is a diagnostic procedure, not a substitute for field measurements.
Allow the evidence to update
Record the date of a fix so that you can compare later field data with the period it covers. Older visits remain in the rolling window for a while; a rerun does not erase them. 1
Search Console’s validation starts a 28-day monitoring period; it does not trigger re-indexing. 3 Keep the lab test as a quick regression check, and use the failing real-user metric to decide what needs work next.