Research Guides
A Core Web Vitals Worksheet for Field and Lab Results
A performance review can contain two apparently conflicting findings: a page loads quickly in a test, while visitors experience slow loading. Before choosing a fix, record what each measurement actually represents. Otherwise, a useful diagnostic can turn into an unsupported claim about a whole website.
This guide provides a reusable evidence sheet for website owners, editors and researchers. It uses public performance documentation and fictional examples. It does not report measurements of Colquhoun.shop or claim that its dashboard collects Core Web Vitals.
1. Keep the three metrics and their units visible
Core Web Vitals cover loading, responsiveness and visual stability. Google's recommended good thresholds are:
- Largest Contentful Paint (LCP): 2.5 seconds or less. This concerns loading performance.
- Interaction to Next Paint (INP): 200 milliseconds or less. This concerns responsiveness to interaction.
- Cumulative Layout Shift (CLS): 0.1 or less. This is a unitless measure of visual instability, not a duration.
The assessment uses the 75th percentile, with mobile and desktop considered separately. For a complete assessment, all three metrics must meet their targets. A single fast load cannot establish that result. See Google's Web Vitals definitions and thresholds.
Copy the metric name, value and unit together. Writing “performance improved from 3.2 to 2.4” leaves the reader guessing whether you measured seconds, a score or something else.
2. Identify the measurement before interpreting it
PageSpeed Insights presents two kinds of evidence. Its field section uses Chrome User Experience Report (CrUX) data from a previous 28-day collection period. Its lab section uses Lighthouse to test a page under controlled conditions.
When the requested page lacks sufficient field data, PageSpeed Insights may show origin-level data instead. An origin groups pages under the same scheme, hostname and port. It is broader than the particular URL you entered. If neither level has enough data, field results may be unavailable. Missing data is not a zero score or proof of poor performance. Check how PageSpeed Insights reports field and lab data.
Start each note with a plain description, such as “mobile, URL-level field data, collection period shown in the report.” If the view changes to origin data, create a separate entry rather than silently replacing the page result.
3. Copy this evidence sheet
Use one record for each distinct measurement. The following is an editorial worksheet, not an official Google scoring system.
- Question: Which visitor problem or decision prompted this review?
- Target: Exact URL requested and final URL after any redirect.
- Scope: URL, origin or page group, exactly as the source reports it.
- Evidence type: Field observation or lab test.
- Time: Collection-period start and end, or test timestamp and time zone.
- Context: Mobile or desktop; for lab tests, tool version, device and network settings, cache state and login state when known.
- Result: Metric, value, unit, percentile where supplied, and any missing metric.
- Record: Saved report, screenshot or trace, with a clear filename.
- Interpretation: What this observation supports, plus the main uncertainty.
- Next check: A named owner, a proposed action and the evidence needed to close it.
Keep the interpretation separate from the result. “The lab trace identifies a slow image request” is an observation. “That image explains the field problem” is a hypothesis until the contexts and evidence support the connection.
4. Use disagreement to choose the next test
Lab and field results can differ because visitors have different devices, networks, cache histories and page states. They also interact with pages in ways an initial page-load test may miss. Total Blocking Time can help investigate responsiveness, but it is not interchangeable with INP. Google's guide to lab and field differences explains these limits.
Use this decision checklist:
- Field and lab both show slow loading: Inspect the lab trace for a plausible cause, then check whether it applies to the affected visitors.
- Field is slow but the lab is fast: Look for a context the test omitted, such as a different viewport, returning state or interaction. Record the candidate explanation before testing it.
- Lab is slow but field is good: Investigate the tested conditions without describing all visitors as slow. The diagnostic may still reveal a useful improvement.
- Only origin field data is available: Keep the origin finding distinct from the tested page. Do not assign the aggregate result to every URL.
- Field data is absent: Report that limit. Use lab results for a bounded investigation, without declaring a real-user pass or failure.
For interaction or post-load problems, a developer can record the relevant actions in Chrome DevTools. Google's Core Web Vitals workflow describes using field evidence to locate problems and traces to investigate them. Choose a representative page and action sequence, not whichever example produces the most dramatic score.
5. Report repeated tests without selecting the best run
Lighthouse's performance score combines weighted metric scores. It can vary with underlying conditions, including content experiments, network routing and the device running the test. The scoring model can also change over time. Preserve the tool version alongside the result. See Lighthouse performance scoring.
As a practical review convention, run the same lab test three times under the same stated settings, preserve every report and summarize the median and range for the metric you are investigating. Three runs are not a statistical guarantee; they simply make a single unusually favorable result harder to hide.
Fictional example: Before an edit, lab LCP values are 3.2, 3.4 and 3.6 seconds. Afterward, they are 2.6, 2.7 and 2.9 seconds. The median falls from 3.4 to 2.7 seconds, about 21%. A defensible note says “median lab LCP improved under these test conditions.” It does not claim that visitor experience improved by 21%, or that Core Web Vitals now pass.
These values are invented to demonstrate reporting, not observed customer results. If the test configuration changed, disclose that before making a comparison.
6. Close the review with a measured claim
Record the change, its release time and the result you expect to move. Recheck the relevant lab scenario, then review suitable field evidence as it becomes available. A rolling collection window can contain experiences from both before and after a release, so do not describe its next update as a clean experiment.
Keep search outcomes separate. Google says Core Web Vitals are used by its ranking systems, but good scores do not guarantee top rankings. Chasing a perfect score solely for SEO can be a poor use of effort. See Google's page-experience guidance.
The most useful handoff ends with three short statements: what was measured, what changed and what remains unverified. That gives the next reader a finding they can inspect and a next step they can repeat.
Sources checked November 11, 2026. The worksheet and reporting examples are original editorial guidance.