A performance review is most useful when it identifies the cause of a slow task and verifies the change. Treat speed as part of a dependable website rather than a guaranteed route to rankings.
Choose pages and conditions
Test the homepage and representative service, product and article templates. Include the route customers use to enquire or buy. Record the URL, date, test profile and connection conditions so before-and-after results mean something.
Separate lab results from real-user field data. Lab tests help investigate individual pages under controlled conditions. Field data, when available, describes actual visits over a period. A high Lighthouse score on a homepage does not certify every page or establish real-user responsiveness.
Investigate the largest content and early requests
Identify what users are waiting to see and which resources hold it up. Image dimensions, format, responsive variants and loading priority can matter. Fonts and styles may also influence when readable content appears. Check the actual request path rather than reducing quality indiscriminately.
An image loaded below the opening view normally has different priority from essential hero content. Use measured page behaviour to decide what should load immediately and what can be deferred. Avoid adding a preload simply because a resource exists.
Reduce unnecessary work
Review scripts, widgets and animations against the job they perform. A navigation effect that deliberately waits before loading the next page can make a fast site feel slower. Test whether removing a dependency changes the important journey and whether anything useful breaks.
Keep supported dependencies and maintenance ownership clear. Installing several overlapping optimisation tools can make diagnosis harder. Make a bounded change and retest before adding another layer of configuration.
Check server and caching behaviour
Where response time is a bottleneck, investigate hosting, application work, caching and external dependencies. Compression and asset caching can help, but settings need to suit the content. Do not cache private or personalised responses as though they were public static files.
Keep a rollback route for configuration changes and recheck forms, editing and login behaviour. A site that scores well but serves stale or incorrect information has not improved in the way the business needs.
Review the result and its limits
Compare the relevant metrics and repeat the customer task. Track useful enquiries or purchases over a suitable period where measurement exists, accounting for changes in traffic and promotions. Do not infer a universal percentage of revenue from a second of loading time.
See Google’s measurement guidance and our guide to speed and conversions. A technical SEO review can connect the findings to the wider website priorities.
