NA DESIGNSMeasure a representative visit
Open the homepage and an important service page on a normal phone connection. Note when the main message becomes readable and when you can use the first button. A fast office connection alone is not a good representation of every customer.
Use a performance report to identify likely causes, then repeat under comparable conditions. Record the URL, date, device setting and test result. One unusually good or bad run should not decide the whole redesign.
Start with the largest visible content
A large hero image or video can dominate the first load. Check its file size and whether the same visual could use a smaller file at the displayed dimensions. Keep a useful poster image for videos so the first screen remains understandable while media loads.
Google’s web performance guidance recommends making the main content resource discoverable early and avoiding lazy loading for the main above-the-fold image. See the Largest Contentful Paint guide for technical detail.
Treat lower-page images differently
Images far below the first screen can generally load as the visitor approaches them. Reserve image dimensions so content does not jump when pictures arrive. Use descriptive alternative text for informative images and empty alternative text for purely decorative images.
Check that compression has not made menu text or portfolio screenshots unreadable. The smallest file is not helpful if customers cannot inspect the work.
Audit fonts, scripts and third-party widgets
- Load only font families and weights the design actually uses.
- Review whether a chat widget, animation library or embed serves a current business need.
- Keep important copy and links available without waiting for optional scripts.
- Check for missing files or repeated downloads before adding another optimisation tool.
Test forms and menus after changing scripts. An improved score is not a success if the enquiry button stops working.
Record a before-and-after comparison
Change one meaningful group of resources at a time. Repeat the same page test and compare loading, readability, layout stability and functionality. Retain the earlier files so you can undo a regression.
Report only what was measured: for example, a reduced image file size or a repeatable test improvement. Do not turn a lab score into a claim that sales increased. For ongoing checks, agree a maintenance scope and review schedule.