Field data vs lab data

Why PageSpeed Insights, Search Console, Lighthouse, and GTmetrix disagree.

These tools answer different questions. Some report delayed real Chrome-user field data. Others run immediate lab diagnostics. The fix starts by labeling the source before debating the score.

>

No signup needed. Takes about 30 seconds.

Decision preview

Search Console says poor. Lighthouse says good. Treat Search Console as field evidence, then use the lab run to inspect the likely cause.

Source: CrUX field data

Diagnostic: Lighthouse lab data

Next check: same URL, same device, same source label

Source labels

The same page can be good, poor, and unmeasured at the same time.

That does not mean one tool is broken. It usually means the report is using a different source, window, device profile, or URL group.

LabelWhere you see itWhat it meansHow to use it

CrUX field data

PageSpeed Insights, Search Console

Real Chrome-user experience data when the page or origin has enough public samples.

Trust it first for user impact, but remember it is delayed.

Lighthouse lab data

PageSpeed Insights, DevTools, CLI, GTmetrix-style tests

A point-in-time diagnostic run in a controlled device, network, and browser setup.

Use it to debug a likely cause and verify same-day changes.

Search Console URL group

Search Console Core Web Vitals report

A site-level grouping of similar URLs with the same metric and device issue.

Use it to prioritize groups, not to prove one exact URL was fixed today.

Synthetic report

GTmetrix and other lab tools

Controlled testing that can add waterfall, filmstrip, location, and monitoring context.

Useful debugging evidence, but compare settings before arguing about scores.

Common contradictions

Start with the question each tool is answering.

Field data answers what a public Chrome sample experienced. Lighthouse and synthetic tools answer what happened in a controlled test. Search Console helps prioritize groups of affected pages.

Search Console fails, Lighthouse passes

Field data is showing a real-user problem.

Inspect the failing metric and affected URL group, then use Lighthouse to find the most likely cause.

Lighthouse fails, Search Console passes

The lab run found a reproducible risk.

Fix cheap, obvious bottlenecks, but do not overrule passing field data from one noisy lab run.

GTmetrix looks good, PSI field data fails

You may be comparing lab evidence with CrUX field evidence.

Check source labels first, then rerun a same-device lab test after each fix.

No field data is available

The page may not have enough public Chrome samples.

Use a consistent lab baseline now and keep monitoring until CrUX data appears.

Use nimo for the handoff

Keep the sources visible until the fix is verified.

nimo turns mixed evidence into a narrower decision: one likely cause, one owner, one first safe fix, and one rerun instruction.

Verification sequence

1

Run a baseline lab check before the change.

2

Ship the smallest safe fix for the failing metric.

3

Rerun the same lab check after deploy.

4

Watch CrUX field data before calling the field issue closed.

Sources checked

Claims stay tied to official source pages.

Checked on May 20, 2026. Recheck the official docs before adding new tool capability, quota, ranking, or measurement-window claims.

FAQ

Quick answers for conflicting reports.

Which source should I trust first?

Use CrUX field data first when you need user-impact evidence. Use Lighthouse or GTmetrix-style lab data when you need to debug the cause or rerun right after a deploy.

Can a lab test prove a field-data fix worked?

It can prove the tested URL changed in that lab setup. It cannot prove the delayed CrUX field window has recovered yet.

Where does nimo fit?

nimo keeps the source labels visible, explains the contradiction, and turns the evidence into one owner, one first safe fix, and one rerun check.

Need the machine-readable version? Use /field-vs-lab-core-web-vitals.md.