Investigate performance and server rendering
Core Web Vitals: How to Investigate a Failing Metric with Field and Lab Data
Investigate a Core Web Vitals problem by separating Search Console field data from lab tests and checking whether a targeted change helps.

A failing Core Web Vitals report tells you what visitors experienced, not which line of code to change. Record the affected metric and device segment first. Then use a repeatable lab test to investigate a possible mechanism. After a change, the lab test can tell you whether that mechanism improved under the test conditions; later field data tells you whether the affected visitors' experience improved.
What the three metrics measure
Largest Contentful Paint (LCP) measures loading performance. Google's good-experience target is for it to occur within 2.5 seconds of the page starting to load. Interaction to Next Paint (INP) measures responsiveness, with a target under 200 milliseconds. Cumulative Layout Shift (CLS) measures visual stability, with a target under 0.1. These are different symptoms: a page can load promptly yet respond slowly to a tap, or respond promptly while its content moves unexpectedly. Google defines the metrics and targets in its Core Web Vitals guidance.
A passing assessment requires all three metrics to meet their targets at the 75th percentile of page loads, evaluated separately for mobile and desktop. One excellent lab run is not that assessment. The Web Vitals measurement guidance explains the percentile and all-three-metrics requirement.
Start with the field-data finding
Open the Core Web Vitals report in Google Search Console. Record the affected metric, whether the finding is for mobile or desktop, and the URLs you need to investigate. The report is based on real-world usage data, rather than a single test of a page you happen to have open. That makes it a useful starting observation, but it does not identify the cause.
Choose a representative affected URL. For example, suppose the URL you select is https://example.com/products/blue-chair. This is a teaching URL, not a page with a known score or fault. Note whether the other affected URLs use the same product-page template. If they do, a shared component is worth examining; if they do not, avoid assuming that one template change will address every URL. Keep the original metric and device segment in your notes so that a convenient desktop test does not quietly replace a mobile finding.
Use the lab to investigate a mechanism
Run the representative URL in a tool that distinguishes field results from lab results. The Core Web Vitals analyzer documentation describes field data as observations from real Chrome users and Lighthouse lab data as a controlled simulation useful for debugging. Record which result you are reading. A lab value and a field value answer different questions even when they appear on the same screen.
Match the lab scenario to the question as closely as you can, then keep it comparable for the retest. Record the device choice, simulated network condition, page version, and the action you took, such as reloading or interacting with the page. Google's Chrome DevTools walkthrough uses preserved console logs and a simulated Slow 3G connection to observe metrics. That connection is a test condition, not a description of every visitor's network.
Investigate the reported metric, rather than changing everything a performance tool mentions:
- LCP: Check whether the loading delay appears connected to the initial response or to content appearing later. Time to First Byte and First Contentful Paint can help investigate LCP, but neither alone proves the cause of a poor LCP result.
- INP: Reproduce a relevant interaction and inspect what happens during it. Lab-only Total Blocking Time can flag potential interactivity issues, but it is not INP and cannot substitute for the real-user result.
- CLS: Watch for content moving unexpectedly as the page loads. If an image or ad slot appears to be involved, inspect whether its space is reserved before it loads. Do not label that element the cause merely because it is on the page.
These checks turn a broad report label into a testable hypothesis. For the example product URL, the next step depends on what you actually observe, not on the fact that it is a product page.
Change one suspected mechanism, then verify twice
Make one targeted change that addresses an observed problem. For instance, if an image appearing without reserved space is followed by visible movement in your CLS test, reserving its dimensions is a reasonable change to test. The image is a candidate mechanism, not a diagnosed cause until the comparison supports it.
Repeat the same lab scenario after the change: same URL, device choice, simulated network, and relevant action. Compare the observed movement or timing with the earlier run. If it improves, you have evidence that the change helped under those conditions. If it does not, revisit the hypothesis before stacking on unrelated fixes.
Finally, check later field data for the affected device segment and pages when sufficient data is available. Reassess all three metrics at the 75th percentile, not just the one you changed. A better lab run does not establish that visitors now pass the assessment. Passing it establishes performance against the Core Web Vitals targets; it does not, by itself, establish a search-ranking outcome.




