A performance score is useful when it helps identify something worth improving. It becomes less useful when every decision is driven by making the number turn green. A visitor needs the important content to appear, buttons to respond and the page to stay still while they use it.
For a business, those are practical concerns. A customer on a phone may be comparing suppliers between appointments or trying to book while using a poor connection. Improving that experience is a better reason to optimise a website than collecting a perfect screenshot from a testing tool.
Know which question the report is answering
PageSpeed Insights can show real-user information when sufficient data is available, alongside a simulated Lighthouse test. The first helps describe experiences recorded from actual visits. The second helps investigate a page under specified conditions. They are related, but a single simulated run does not describe every customer’s experience.
When a report says there is no real-user data, it means the tool does not have enough qualifying information to show that assessment. It does not prove that the site is fast or slow. Record the available evidence and use appropriate lab testing while collecting a more useful picture.
Understand the three Core Web Vitals
Google’s Core Web Vitals guidance identifies loading, responsiveness and visual stability as the three areas measured. The recommended good thresholds are:
- Largest Contentful Paint, or LCP: 2.5 seconds or less.
- Interaction to Next Paint, or INP: 200 milliseconds or less.
- Cumulative Layout Shift, or CLS: 0.1 or less.
The assessment uses the 75th percentile, with mobile and desktop considered separately. That gives a more useful target than selecting the fastest visit or the most flattering test run. These metrics describe particular parts of the experience rather than replacing a full review of the website.
Speed supports SEO, without guaranteeing rankings
Google says its ranking systems use Core Web Vitals, while also explaining that good results in these reports do not guarantee top positions. Its Search Central guidance places these signals within a broader page experience and search context.
That means an improvement from a very slow page can be worthwhile, while a move from an already strong score to a perfect one may have limited practical significance. Your content still needs to answer the search, the website needs to be accessible to crawlers and the business needs convincing reasons for customers to choose it.
Our SEO work considers those elements together. Performance testing should help the page do its job, alongside useful content, appropriate internal links and clear service information.
Find what delays the important content
A large image may be part of an LCP problem, but its file size is only one possible cause. A browser might discover it late, wait for other resources or be delayed before it can render the content. The official LCP optimisation guide breaks the delay into parts so the fix addresses the actual bottleneck.
A practical review might check the server response, how the main image is referenced, whether the right image size is delivered and whether unnecessary work blocks rendering. Compressing every image again will have little effect if the important resource is still being requested too late.
For website design, this is a reason to think about photography, video and visual effects early. A strong first impression can be planned around sensible loading priorities rather than repaired only after a heavy page has been built.
Check what happens when people interact
A page can appear quickly and still feel awkward when a visitor opens a menu, selects a filter or adds an item to a basket. Google’s INP guidance focuses on responsiveness to interactions. Heavy JavaScript and lengthy tasks on the browser’s main thread can contribute to delays.
Try the actions customers actually use. Open the mobile navigation, change a product option and submit an agreed test form. A homepage-only loading test can miss problems in a catalogue or checkout. Consider the effect of third-party chat, tracking, maps and review widgets on those journeys too.
Our development reviews look beyond the first screen. The purpose is a working experience, including the moments when the customer is trying to do something.
Keep the layout steady
If a page moves just as someone taps, it can turn a routine action into a mistake. The CLS optimisation guide discusses causes such as images without reserved space, dynamically inserted content and font changes.
Set suitable dimensions or aspect ratios for media, plan where banners will appear and test font loading. On a shop, a late-loading recommendation widget should not push the purchase controls around. On a service site, a contact button should remain usable while supporting content arrives.
Animations deserve their own review. Keep the page readable, honour reduced-motion preferences and avoid effects that repeatedly hide content or interrupt interaction. Motion can support a brand without making a customer wait to understand the offer.
Make changes you can compare and recover
Record the test URL, device mode and important conditions before changing anything. Run a small number of comparable tests, then investigate consistent findings. Scores vary with test conditions, so repeated runs until one looks excellent do not establish a reliable improvement.
Keep changes focused. Optimising a main image, reducing unnecessary scripts and changing caching rules at the same time can make an unexpected result harder to diagnose. Maintain a recovery copy and check forms, navigation and commerce features after the release.
Suitable website hosting provides part of the foundation, while ongoing care helps detect regressions when software or third-party features change.
Prioritise the improvement customers will notice
Begin with important pages and important tasks. Fix a slow product page before spending hours removing a harmless warning on a rarely visited page. Balance the likely benefit with the time, complexity and risk of the change.
A good performance review ends with a clear explanation: what was delaying or disrupting the experience, what changed and how it was checked. That gives you something useful to maintain, rather than a number that may look different tomorrow.