Start with the experience, not a score
A website can feel slow in different ways: the main content may appear late, a menu may not respond promptly or the page may move just as someone tries to tap a link. Ask which task is difficult and on which page or device. A single headline score is a useful signal, but it does not describe every visitor’s experience.
Choose pages connected to real business goals, such as a key service, booking or contact page. Open them on a phone and a larger screen and note what happens. If a visitor cannot complete an enquiry, investigate that problem alongside loading speed rather than treating the two as unrelated technical projects.
Compare field information and a test run
Some performance reports combine field data from eligible real-world browser experiences with laboratory tests performed under controlled or simulated conditions. The two can differ because devices, locations, connections and page visits vary. A page may not have enough field data for a separate report; that absence is not evidence that the page is fast or slow.
Use the exact page address, record when you tested and read the report’s explanations. Laboratory tests help reproduce a problem and examine possible causes; field information can show a wider range of real experiences where available. Repeat a test after a meaningful change, but do not treat small differences between individual runs as proof of success.
Understand the main user-facing measures
Core Web Vitals are intended to represent aspects of loading, responsiveness and visual stability. Largest Contentful Paint relates to when a prominent part of the page appears. Interaction to Next Paint concerns responsiveness to interactions. Cumulative Layout Shift describes unexpected visual movement. Their definitions and reporting can evolve, so use current documentation when assessing them.
A practical example of layout movement is a button shifting after an image or banner loads. A delayed response might make someone tap a control twice. These measures do not cover every aspect of usability, accessibility or business success. Good reports are helpful, but they do not guarantee a search position or a sale.
Investigate speed as a visitor journey
- ObserveChoose a real task and page
- MeasureCompare available field and test data
- ImproveChange the most relevant issue
- RetestCheck the task and report again
Look for expensive page elements
Large images, multiple custom fonts, video, tracking scripts, embedded widgets and complex page-builder components can all contribute to loading work. The host and server response also matter. Before removing anything, identify what it does: a booking widget may be central to the customer journey even if it adds weight.
Start with changes that have a clear reason. Prepare appropriately sized images, remove genuinely unused scripts and review whether an embed must load immediately. Keep important visual content and accessibility intact. Ask your developer to explain the expected benefit and any trade-off; optimising a metric by making a page harder to use is not an improvement.
Be careful with plugins and quick fixes
Caching, compression and delivery services can help in a suitable configuration, but installing several tools with overlapping functions may introduce conflicts. Test changes in an appropriate environment and keep a usable backup. If a booking form, consent choice or image gallery stops working, a faster report is not a successful outcome.
Avoid disabling necessary security or accessibility features merely to improve a score. Check that content is current after caching and that updates appear as expected. Decide who owns performance work after launch; an improvement can be undone when a large image or additional third-party widget is added later.
An illustrative investigation
Imagine a Hertfordshire consultancy whose service page appears slowly on a phone. The owner and developer test the same page, note that the large banner image loads late and check whether the content system has generated a suitable smaller version. They also inspect a third-party video that starts loading before the visitor asks to play it.
In this hypothetical example the team prepares an appropriately sized image and changes how the video is introduced, then checks the page and its important links again. They compare reports under similar conditions and ask whether the actual page feels clearer. The illustration describes a method; it does not predict a particular improvement.
Agree what matters with your supplier
Ask which pages were tested, which issues affect visitors most, what the proposed changes involve and how success will be checked. Confirm whether work includes image preparation, plugin review, hosting changes and testing on real devices. A supplier should be able to discuss limitations and avoid promising a perfect score for every visitor.
If field data is available, review the distribution and page context rather than only the overall label. Keep a short record of the change, date and observed result. Revisit the page after major content or platform changes, especially when a new campaign introduces larger media or additional third-party services.
A practical first review
Pick one important page, note the task a customer wants to complete and run a current performance report. Check whether field data is available, identify the largest relevant opportunity and discuss it with the person who maintains the site. Keep a backup or recovery plan before changing plugins, server settings or the theme.
After the change, open the same page on a phone, follow the menu and test the contact route with the recipient’s agreement. Compare the report and the experience, and document any side effects. Continue with the next issue only when the first improvement works for visitors as well as the measurement tool.
Turn a report into a short work order
Choose one issue from the report and describe the affected page and visitor task. Ask the developer what causes the issue, which change is proposed, whether a backup or staging test is needed and what might stop working. Record the baseline test conditions and the date. If the report shows no field data, note that limitation rather than presenting the laboratory result as a measurement of every visitor.
After the change, repeat the same task and test under reasonably comparable conditions. Check that a booking widget, consent control or image still works before calling the task complete. If the report changes only slightly, consider natural variation and investigate the user experience. Keep the work order with the website-maintenance record so the team can revisit the decision when content or integrations change.
