How to make your website faster, and why it helps SEO
Most speed advice optimises the number that was already fine. Here is what Google actually measures, what to fix in order, and how to tell which one is your problem.

Ama Mundum LLC Team5 min read
We measured this site last week expecting to find bloat. What we found was that the number everyone optimises was already fine, and the number nobody talks about was the problem.
The homepage sends 34 KB over the wire. It unpacks into 609 KB that the browser has to read. Those are both true, they are both about the same page, and only one of them was hurting us.
That gap is where most speed advice goes wrong. So before the tips, the distinction that makes them make sense.
Two different costs, often confused
Bytes on the wire is what gets downloaded. Your server compresses text before sending it, usually with gzip or brotli, and HTML, CSS and JavaScript all shrink dramatically because they repeat themselves. This is the number a file size tells you, and it is usually not your problem.
Bytes to process is what the browser does after the download: unpack it, parse it, build the page, run the scripts. Compression does not help here at all. A phone with a modest processor can finish downloading quickly and then sit there for seconds doing work.
Images are the opposite case. A JPEG is already compressed, so what you download is close to what you decode, and the file size on disk really is the cost.
So the honest rule: images are a download problem, and code is a processing problem. They need different fixes, and most articles only teach one.
What Google actually measures
Three things, under the name Core Web Vitals:
| What it means | Good | |
|---|---|---|
| LCP | How long until the biggest thing on screen appears | under 2.5s |
| INP | How long the page takes to respond when you tap | under 200ms |
| CLS | How much the layout jumps while loading | under 0.1 |
Notice what is not on that list: total page weight. Google does not measure how big your page is. It measures what the visitor experiences, and weight is only one of several things that affect it.
Fix these, in this order
1. Resize before you compress. This is the largest win and the one most often skipped. A 4000 pixel photograph displayed in an 800 pixel column is carrying twenty five times more data than it can possibly show. No compression setting recovers that, because the pixels are genuinely there. Getting the dimensions right first does more than any amount of quality tuning afterwards.
2. Serve WebP. At the same visual quality it is typically 25 to 30 percent smaller than JPEG, and every current browser reads it. For a page whose weight is mostly photographs, this is close to free.
3. Give every image a width and a height. Without them the browser does not know how much room to reserve, so the text moves when the picture arrives. That is CLS, and it is the cheapest of the three to fix: two attributes.
4. Do not lazy-load the main image. Lazy loading is for images below the fold. Applied to the hero, it delays exactly the thing LCP is timing, which makes the score worse while looking like an optimisation.
5. Compress last. Once the dimensions and the format are right, compression is a tidy up rather than a rescue.
The part that is not about images
If your images are already sensible and the page is still slow on a phone, the cost is in code: the HTML, the JavaScript and the work done before the page becomes usable.
The tell is the shape of the numbers. If the page downloads quickly and still takes seconds to respond, nothing you do to the images will help. That is what we found here: 34 KB arriving fast, then a large document to work through afterwards. Compressing an image would not have moved it by a millisecond.
Measure before you change anything
Run PageSpeed Insights against the real URL, and read the mobile tab rather than the desktop one. Desktop almost always looks good and almost never reflects your visitors.
Two cautions, both of which cost us time:
Lab numbers are not field numbers. The score comes from one simulated run on a simulated phone. Useful for comparing before and after, not a measurement of what your visitors feel.
Measure the thing you actually changed. It is easy to rebuild, measure the old version still running, and conclude something false. Check you are looking at the new one.
The tools
All of these run in your browser, so nothing is uploaded and there is no queue.
- Resize imageOpen tool first, because the pixel count is the biggest lever
- PNG to WebPOpen tool and JPG to WebPOpen tool for the format
- Compress imageOpen tool last, when the dimensions are already right
- Compress PDFOpen tool if the heavy thing on the page is a document rather than a picture
If you want the reasoning behind picking one image format over another, PNG vs JPG vs WebP works through that decision.




