smartwebutah.com is my own site, and when I timed it, the page that explains what I do took 12 seconds to show its main picture on a phone. So I rebuilt it four ways, timed every one, and wrote down every step.

The four versions

I left the real site alone and built each version next to it on its own address.

  1. WordPress as it was. 73 pages and posts, a shop, memberships and 12 online courses.
  2. A frozen copy. A tool visits every page like a visitor and saves it as a plain file. Nothing runs when someone visits.
  3. WordPress as the editor. I still write in WordPress. A small separate site reads my pages from it and shows them to visitors. People call this headless.
  4. A full rewrite. New design, new pictures, the old writing cut down to what was worth keeping, and the shop and memberships gone.

What broke

A frozen page cannot remember anything, so anything that needs to remember has to go somewhere else.

  • The shop. 13 products and a cart. I dropped it. If your shop matters, keep WordPress or use a shop service run by someone else.
  • Memberships. I looked at the real data before dropping them: five "active" subscriptions, all set to be paid by hand, with $1 paid in 90 days. Nothing was charging anyone.
  • The courses. 12 courses and 528 lessons. They moved to a free academy on this site, with every old address redirecting.
  • Comments, forms and search. There were 15 comments, no form and no search box, so there was little to lose. If you have them, services run by someone else do the job.

What the stopwatch said

I tested the same five pages on every version that has them: the home page, the list of posts, a short post, the About page and the heaviest page. (The frozen copy never had the list of posts, because nothing on the old site links to it.) Each page got five runs on a phone and five on a computer, from the same Mac on the same network, and I show the middle result. The table is the phone test, as score out of 100, then seconds until the main picture shows.

On a phone, swipe the table sideways.

PageWordPress as it wasFrozen copyWordPress as editorThe rewrite
Home93 · 2.5 s63 · 6.7 s100 · 0.8 s99 · 1.6 s
Writing list90 · 2.9 snot on this version100 · 0.8 s96 · 2.2 s
A short post91 · 2.6 s91 · 2.6 s100 · 0.8 s99 · 1.7 s
About86 · 3.0 s70 · 4.4 s84 · 4.5 s99 · 1.6 s
The heaviest page71 · 12.2 s72 · 7.6 s100 · 0.8 s99 · 1.6 s
Seconds until the main picture shows on a phone, on the heaviest page (middle of 5 runs, shorter is better)
WordPress as it was12.2 sFrozen copy7.6 sWordPress as editor0.8 sThe rewrite1.6 s

On the heaviest page the main picture went from 12.2 seconds to 1.6. The page itself went from 2,117 kB to 94 kB.

  • The frozen copy was not the win I expected. It still carried the same big pictures, and on a phone its home page scored 63, below the original's 93. Freezing a heavy page gives you a heavy frozen page. It did answer in about 100 milliseconds on a short post, against about 1,160 for WordPress.
  • The editor-stays version looks best, and I do not trust it. It scores 100 on four of five pages, but its pages are much plainer, and several carry almost no pictures. Its home page is the opposite: 4,850 kB, the heaviest of any version, because I left one big picture in. It is not the same page, so the comparison is not fair.
  • The rewrite is the real result. It scored between 96 and 99 on every page, with the same ideas, cut down, and a design I like better. It is faster partly because it does less: no shop, no memberships, no courses.

How I tested

Lighthouse 12.8.2, Google's free speed test (the engine behind PageSpeed), on its phone setting (simulated phone, slow connection) and its computer setting. Five runs per page per setting, back to back, the middle result shown. Run on 30 October 2026. The sites were measured one after another, not at once. This is lab data from one machine, not what real visitors saw, and WordPress is measured from its own server while the other three come from Cloudflare, a company that serves sites from many places around the world. A few points either way is noise. Every one of the runs, the script and the written method are in the repository.

Which one is right for you

Keep WordPress if you sell things, people log in or you change the site every day. Freeze it if the site is pages people read and it already looks right, but shrink the pictures first. Keep WordPress as the editor if you like writing in it and have someone to build the other half. Rewrite if the design is tired and much of the writing is padding. The rewrite was the most work, and most of that was deciding what to cut.

The full step-by-step version, with a picture of every step, is in the guide.