01
Step one: pick the pages that earn money
Do not tune everything. List the three pages that bring in orders or inquiries, usually the home page, a top product or service page, and checkout or contact. Example: a bakery site picks its home page, its custom-cake page and its order form. Record how each loads on a mid-range phone over a typical mobile connection. Write down the date and the numbers in a simple sheet. Starting narrow keeps each session short, and it ties every later fix to a page that matters to revenue.
02
Step two: measure, then change one thing
Run the same public speed test on each page and note the largest visible element, layout shifts and total page weight. Pick a single cause to fix per session. The bakery finds a four-megabyte cake photo on its home page, so session one is only resizing and compressing images. Changing one variable lets you see what helped. If you stack five fixes at once, you cannot tell which worked or which broke something, and you lose the learning that makes the routine valuable.
03
Step three: retest, log and schedule the next session
After the change, retest on the same device and settings, then add a row to your log: date, fix, before and after. The bakery sees the home page load noticeably quicker and moves on to a delayed chat widget in session two, then font loading in session three. Set a recurring calendar slot, perhaps every two weeks, and re-check after any new plugin or banner. A short log also gives a developer or agency clear context if you later hand the work over.