NOTE / PERFORMANCE
Measure your own site speed before you pay anyone.
You can get the three numbers that matter in about ten minutes, without installing anything. This note shows how to take the measurement, how to read it, and when the honest answer is that nobody should be paid to fix it.
1. Take the measurement
- Open your site in Chrome, in a normal window (not incognito — extensions change the result).
- Open DevTools (F12), then the Lighthouse tab.
- Choose Mobile, not Desktop. Check Performance only, so the run stays fast.
- Run it three times and write down the middle value, not the best one.
- Do the same for one product or service page, and one page you know is heavy.
Write the numbers and the date somewhere you will find them again. Without a recorded baseline, a later "it feels faster" cannot be checked by anyone, including you.
2. Read the three numbers
| Metric | What it actually asks | Good |
|---|---|---|
| LCP | How long until the biggest thing on screen is visible — usually the hero image or the headline. | ≤ 2.5 s |
| CLS | How much the layout jumps while loading: text that moves, buttons that slide away from the finger. | ≤ 0.1 |
| INP (lab proxy: TBT) | How long the page takes to respond when someone taps or types. | ≤ 200 ms |
A Lighthouse score is a summary of many things, including things that do not affect your visitor. Read the three metrics above, not the big number at the top.
3. Find the cause
In small business sites the same five causes repeat. Each one has a signature in the report:
| Cause | How to spot it |
|---|---|
| An oversized hero image | LCP element is an image, and the "Properly size images" or "Serve images in modern formats" item has real savings. |
| Render-blocking script in the head | "Eliminate render-blocking resources" lists a script or stylesheet, and First Contentful Paint is slow while LCP is close behind it. |
| A font from a third party | A network request to another domain in the waterfall, and text that appears late or swaps shape. |
| Widgets you forgot about | Chat, pop-up, heatmap or tracker scripts from other domains, often larger than your own page code. |
| Layout that loads late | CLS above 0.1, and the report names the elements that shifted. Usually an image with no width and height, or a banner injected after load. |
4. Decide what to do
| What you see | What it usually means | Who can do it |
|---|---|---|
| One image is the whole problem | Compress and resize it, add width and height, use a modern format. | You, in an afternoon |
| A plugin or widget dominates | Remove it, delay it, or accept the cost knowingly. | You, with a decision to make |
| Theme or template code blocks rendering | Needs code changes in the template, and a way back if it breaks. | A developer |
| The server is the bottleneck | TTFB is slow on every page. No page-level fix will help; hosting or caching is the answer. | Hosting, not design |
| Numbers are fine | Nothing to fix. Spending money here buys nothing. | Nobody |
5. When it is not worth paying for
- The site gets a handful of visitors a week and does not take enquiries. Speed will not change a number that is already near zero.
- The bottleneck is the host or a platform you cannot change. Fixing the page will not move TTFB.
- The site is about to be replaced. Optimise the new one, not the one being retired.
- You need something else entirely — more enquiries, better copy, a working form. Speed is not the constraint.
What a paid audit adds
Not the numbers — you have those now. What it adds is the boring part that decides whether the fix works: field data instead of one lab run, three templates measured the same way so one page does not hide another, a ranked list where each item carries its cost and expected effect, and a re-measurement under identical conditions so before and after mean something.
If your numbers are already good, the audit will say so and there is nothing to sell you. That is the point of publishing this note.
This note is free and complete. If you can fix it yourself with it, do that — it is the better outcome for both of us.