Performance
WebPageTest
WebPageTest is a free lab testing service that loads a page in real browsers from chosen locations and devices, and returns waterfalls, filmstrips, and metrics. It gives more control over network, location, and repeat views than a quick score.
How it is measured
Choose a location, browser, connection profile such as a Moto G on 4G, and number of runs. Each run returns a waterfall, a filmstrip, Start Render, LCP, CLS, TBT, and a content breakdown. Use 3 to 9 runs and read the median.
Test first view and repeat view to see cache effects. Use scripting for logged-in flows, blocking of specific hosts to estimate a third party's cost, and side-by-side comparison of two URLs. It is lab data, so pair it with field data.
Worked example
A bookshop in Dublin suspects its review widget. It runs WebPageTest from a London 4G Moto G, 5 runs each, once normally and once with the widget's domain blocked. Median LCP is 4.1 s against 2.9 s, and TBT is 920 ms against 260 ms. The filmstrip shows the product image arriving 1.2 s earlier.
The owner moves the widget behind a click.
How it differs
WebPageTest produces waterfalls from a real browser at a location you choose, while a waterfall is only the chart, and local DevTools draws one too. WebPageTest adds filmstrips, remote locations, and repeat views that a local waterfall cannot. Lighthouse is lighter and returns a score with audits.
Common errors
Running one test. Using the default fast connection. Picking a location far from your audience. Comparing runs from different locations. Ignoring repeat view. Treating the grade letters as targets.
In practice
Run your top page five times on a mobile profile from the region of most visitors and save the median result URL. Look at the waterfall and filmstrip for the first stall, make one change, run again with the same settings, and compare in the side-by-side view.