Website performance
Slow website on mobile? What to check before paying for fixes.
Start by recording what feels slow on the page a customer actually uses. Waiting for a photograph, tapping an unresponsive menu and losing your place when content moves are different problems. A single PageSpeed score cannot tell you which repair your business needs. For service businesses in Stoke-on-Trent and beyond, the useful outcome is a page customers can read and use comfortably, with a clear way to ask for a quote.
Check a customer journey, not just your homepage
Choose a main service page, a work example and the enquiry page. On your phone, follow the links a prospective customer would use. Note where you wait, where a tap seems ignored and whether anything moves while you are reading or reaching for a button. Stop before submitting an enquiry unless a delivery test has been agreed.
Record the exact page address, phone, browser, connection and cookie choice. Try a normal mobile connection as well as office Wi-Fi where practical. A page already stored by your browser can feel different from a first visit, so include a fresh private session. This is an observation exercise, not a controlled performance benchmark.
Read the evidence behind a PageSpeed score
PageSpeed Insights combines two kinds of evidence. Its real-user section summarises eligible Chrome visits over a rolling 28-day period. Check whether it describes the exact page or the wider website origin: a whole-site summary cannot identify a fault on one service page. If there is insufficient data, that means the report cannot assess that real-user experience; it is not a pass or a fail.
The Lighthouse section is a simulated test that helps investigate loading. Its score can vary between runs. Keep the same page and Mobile setting when comparing changes, save the reports and repeat the test to see whether a result is consistent. Do not compare one desktop run before a repair with one mobile run afterwards.
Keep three measures separate: LCP concerns when the largest visible piece of content appears; INP concerns how promptly the page visibly responds to interactions; CLS concerns unexpected movement. A standard Lighthouse page-load run does not replace real interaction evidence. Use the report to ask a better question, then check the actual menu, page and form.
Match the symptom to a specific investigation
The main photograph arrives late: ask which element the report identifies as the largest content and what delays it. A large image file, a late image request, a slow initial server response and delayed rendering need different fixes. A smaller file will not resolve every delay. If that image is visible at the top of the page, ask the developer to check that it is not being held back by lazy loading, which postpones loading until an image is near the visible area.
The page is visible but the menu or form feels stuck: ask for a recording of that interaction and an inspection of the work the browser is doing. Scripts, third-party widgets and expensive page updates can delay feedback. Removing a useful booking tool without checking its role may create a different customer problem. Agree the affected control and a safe change to test.
Text or buttons move as the page loads: record which section moves and what appears immediately before it. Images, embedded panels and font changes are candidates to inspect. Reserving the right space for content can help; simply compressing photographs does not address every layout shift.
A worked example: the quote button that keeps moving
Imagine a hypothetical landscaping business whose patio page loads its heading quickly, then inserts a large gallery above the quote button. On a phone the button moves just as the visitor tries to tap it. The owner describes the website as slow, but reducing the heading font size or changing the host would not directly address the observed movement.
A useful repair brief would identify the gallery, request space for its initial image before it arrives, and check the behaviour with both a fresh visit and a returning visit. The developer would still need to inspect the real cause. The acceptance check is that this gallery no longer unexpectedly displaces the button in the tested layouts, while the images, navigation and quote route still work.
Keep the evidence honest: a successful phone check establishes that tested journey, not every device or a sales increase. If the page also has a late main image or a sticky menu, list those separately rather than assuming one repair resolves all three.
Copy this brief when asking for a repair
Page and task: “On [exact URL], a customer needs to [read the service / open the menu / request a quote]. On [phone and browser] using [connection and cookie choice], I see [specific symptom]. It happens [on first visit / after scrolling / after tapping]. Here is a recording and the dated report, if available.”
Evidence and scope: “Please identify the cause, explain the smallest change that addresses it, and separate that from optional design work. Tell me which pages, plugins or integrations the change affects, how it can be reversed, and what you will charge before starting.”
Acceptance record: “Compare the same page and mobile test settings before and after, record the actual symptom again, check narrow and wider screens, and confirm that the service content, menu and enquiry route remain usable. Note anything you could not test.” Arrange any live form delivery test with the person receiving it rather than sending repeated test messages.
Use one short line per issue: page → symptom → evidence → agreed change → check result → remaining limits. Keep the old report as well as the new one. A report that says only “performance improved” is less useful than a record showing what changed and which customer task now works better.
Know when to stop optimising and ask for a wider review
A focused repair is a sensible first option when the cause is identifiable and the current platform can support the change. A wider review becomes useful when several essential pages or integrations share the problem, the platform prevents a maintainable fix, or the enquiry journey is confusing even after loading improves. Ask for those constraints to be explained before agreeing to a rebuild.
Google says there is no single page-experience signal, and good Core Web Vitals do not guarantee high rankings. A faster page also does not establish more enquiries. Keep relevant search visits, completed requests and qualified business conversations separate when assessing results.
Zero Labs is based in Stoke-on-Trent and works with businesses across the UK. If you want help deciding what deserves attention, send the public page address and describe the delay. We can discuss a review of the visible journey and agree whether deeper performance investigation needs access to your platform. You do not need to promise yourself a new website before understanding the fault.
Take the next step.
Explore the related services and guides to plan your next step.