Choose one page and one task
If buyers complain about the catalogue, ask for the page where they noticed a delay. The homepage might load quickly and use a different template. Find out what the buyer was waiting for: a product photo, a filter or a response to a tap.
Open the page on a phone, then repeat the check in a tool with a mobile profile. Loading and script delays show up more clearly on a slower device or connection. Record the conditions of each check separately.
Separate lab results from real visitor data
PageSpeed Insights can show field data from the Chrome UX Report and a Lighthouse lab test. Check whether the field data is for the URL you selected or the entire site. If there isn't enough field data, mark the result as unknown.
Use the lab test to investigate what might be causing the problem.
| Metric | What a visitor notices | What to examine |
|---|---|---|
| LCP | When the largest visible element that qualifies for the metric appears | Which element was selected and how much time the server, load and render took |
| INP | How long it takes to see a response after tapping or typing | Field data and a recording of a slow interaction |
| CLS | How much content moves unexpectedly | When text, images or buttons shifted and what caused the move |
| TBT | Long tasks occupy the browser in the test | Which scripts keep the main thread busy during loading |
Find what's delaying the main content
If visitors wait for the main image, record a page load and find the request for that file. Did the request start late or take a long time to download? The browser might have fetched it quickly but waited for a script before showing it. Each case needs a different fix.
If blocks jump, see what appeared just before the shift. Reserve space for images and choose a place for late notices. Test the fix with the same template and screen width where you saw the shift.
Record a slow tap and find what held it up
Open the filter or menu during a performance recording. Note the delay between the tap and the response, and find the task keeping the browser busy. If a third-party script is in the way, have a developer test it on a staging copy while preserving normal page behaviour.
Lighthouse TBT helps you spot long tasks in a lab page load. INP measures how quickly a page responds to interactions during a visit. A change in TBT can point to where you should look, but it can't stand in for field INP.
Repeat the test under comparable conditions
Record your baseline and agree how you'll check the fix. Run the test several times before and after, with the same settings. Keep the range of results. Check field data separately once there have been enough new visits.
| What to record | Example test record |
|---|---|
| URL and task | Category page, first load and filter selection |
| Device and conditions | Phone model or emulation profile, browser, connection and cache state |
| Date and time | Test time and site version |
| Source | Lighthouse for lab data, CrUX for field data marked “page” or “site” |
| Number of runs | Three runs per version, median and range |
| Limitation | There isn't enough field data for this URL yet |
Choose the audit that matches your question
Choose the free audit when you need to understand your own website first. Choose the competitive audit when your next decision depends on comparing your website with relevant alternatives. Both turn supported evidence into practical actions.