Your best argument is a photo of the finished kitchen. Let them see it.

Dan builds extensions. He does not win work by explaining joinery — he wins it by showing a homeowner a kitchen that used to be a garage, in a house that looks like theirs. So the gallery is the site, and the site is the gallery.

Which is why it was 38MB. Every project page loaded every full-resolution photo straight off the camera, on a page builder that shipped its whole widget library alongside them. On a phone, the first image landed after nine seconds — and homeowners research three builders in one evening.

The photos did not get smaller to look at. They got smaller to download: converted, lazy-loaded, and served from a cached page that never boots PHP. Same portfolio, same impact, arriving while they are still interested.

Using xSpeed, my site is a builder’s portfolio and it is far too heavy. Benchmark it on mobile, then convert every image to WebP/AVIF, lazy-load the project galleries, remove unused CSS, and turn on page caching. Do not change how the photos look at full size. Re-test on mobile and show me the before/after page weight and first-paint time.

What happens: Your AI reports the new page weight and first-paint time, usually cutting a portfolio page to a fraction of its size without touching the design.