Website speed test

Lab figures and field figures tell separate stories

A Lighthouse measurement proceeds in controlled conditions: one emulated device, a throttled network, a repeatable fetch of the address. Identical addresses score identically every single attempt; lab scores therefore excel at confirming whether an edit helped. Field figures work differently: Google's CrUX statistics aggregate CWV readings that actual visitors generated on the address across a rolling 28-day window, judged via the 75th percentile so sluggish experiences cannot hide. A domain can score near-perfect within lab conditions yet fail its CWV verdicts when actual devices, weak networks, third-party tags drag typical visits downward. Lab figures pinpoint the failing stage; field verdicts summarize the visitor experience. Wherever Chrome lacks sufficient real-user information, the panel says so outright instead of inventing figures.

Thresholds come straight from Google

Verdicts consult Google's published CWV thresholds — the same figures its ranking signals consume: Largest Contentful Paint within 2.5s, Interaction to Next Paint at or below 200ms, CLS below 0.1 — all judged at the 75th percentile of genuine visits. FCP and TBT accompany them as lab context, useful when diagnosing where fetching stalls; neither carries a fail verdict, since Google defines no field threshold there. INP lives exclusively as field information — no emulated fetch can guess which interaction feels sluggish to a visitor. The opportunities list preserves the same honesty: entries are Lighthouse observations, annotated with the milliseconds and kilobytes Google estimates recoverable.

Zero scripts injected; zero snippets shipped

Monitoring products typically request a JavaScript tag so the vendor can observe visitors — adding third-party heft to the pages awaiting improvement. This tool adopts the opposite posture: the lab measurement unfolds wholly within Google's PageSpeed Insights infrastructure, and field figures come from Chrome usage statistics Google maintains. The measured domain serves zero extras, embeds zero beacons, and acquires zero visitor-side RUM tags to maintain and later remove. Retention is minimal: the measured address, its metrics, a timestamp — preserved so a share link outlives the browser session; the person behind the measurement is not logged at all.

Website speed test FAQ

How do score and CWV verdicts differ?

Scores condense Lighthouse's lab observations from emulation. CWV verdicts measure field figures against Google's thresholds on the 75th percentile — the two disagree in practice, and the field verdict is what Google's page-experience signals actually consume.

What does an absent field row indicate?

CrUX publishes metrics solely for addresses and origins with sufficient Chrome visitor volume within the 28-day window. Quiet addresses lack a field row — lab scores stand, and the readout presents an unmistakable no-data state instead of guessing. When the precise address lacks a row but its origin carries one, origin-wide figures render, labeled.

Does measuring a page alter the target?

The target itself is untouched: measurement is carried out via Google's infrastructure over statistics Chrome gathered previously. Zero visitor-side scripts and snippets get added, and solely the measured address, its metrics and a timestamp are preserved.

What does the JoeFlow handoff create?

It carries the measured address into a free JoeFlow workspace where speed monitoring covering your domain is configurable — scheduled re-measurements and trends arrive in the Joe Sites area as that capability ships. Registration is complimentary, and the tool that referred you is preserved as attribution.