The Design Boutique · Rebuild
The Design Boutique
thedesignboutique.com moved off WordPress onto a content system with no plugins, no theme and no update treadmill. The design is unchanged by design. Everything underneath it is not: the migration was the visible half, and the editor it moved into is the other. Every figure below was measured, not estimated.
Speed
Measured with Google's own Lighthouse, on the same pages, on both sites. The phone numbers matter more than the desktop ones, because that is where most of the traffic is and where the old site struggled most.
The interesting part was the cause. The assumption going in was that unoptimised images were the problem. They were a problem, but the largest single cause was the hero's autoplaying video, which pulled roughly 840 KB of Google's JavaScript onto the critical path before the page could render. The video now loads once the browser is idle, behind a poster frame that holds its place, so nothing about the design changed and the page renders in a third of the time.
Search
Connected to the site's own Search Console data, so these describe the real site rather than a simulation. Two findings were worth more than the scores.
The SEO score reads 100 once indexing is on. That is arithmetic rather than a measurement: while the rebuild sat behind its staging noindex it scored 69, and the single failing audit was “page is blocked from indexing”, which carries 4.04 of the category's 13.04 total weight. That is exactly the missing 31 points. Nothing else in the category fails.
| What was found | Detail |
|---|---|
| 26 pages Google has never indexed | Live, reachable, returning 200, and absent from Google entirely. Several have had no impressions in sixteen months. |
| 92,068 impressions, 51 clicks | The home page over 28 days. Being shown constantly and passed over, which is a title and description problem rather than a ranking one. |
| Crawl delay of 10 seconds | Set in the old robots.txt. Google ignores it; Bing does not, and ten seconds a page is slow enough to matter on a site this size. |
| Sitemap split by content type | One index over five files instead of one flat list, so Search Console reports coverage per section: “none of the case studies are indexed” instead of a single number. |
Attack surface
The old site ran Elementor Pro, ACF and Rank Math Pro among others. Every plugin is code from someone else running with full access to the site and its database, and every one is a thing to update. A cookie-triggered backdoor was found on the main site during this engagement.
The rebuilt site has no plugins, no theme and no PHP. There is no admin login on the public site to brute force, no database reachable from the internet, and nothing to patch on a Tuesday because somebody published an advisory. The editor is a hosted service with its own authentication; the website itself is static files and a small number of server routes that we wrote.
What moved
30 pages, 53 blog posts, 14 client case studies and 18 event pages, with their addresses preserved so existing links and Google's index still resolve. The old site answered /blog/some-post with a redirect to /some-post; so does the new one, because without it every link ever shared to a post would have broken on the day of the switch.
122 documents arrived with no search title or description, because the original import brought page content but not the Rank Math metadata sitting beside it. Those were migrated separately from the live site.
What the editor gained
| Before | After |
|---|---|
| Edits went live the moment they were saved | Changes are private until published, and every version is restorable |
| SEO fields, with no indication of whether they were any good | A score with the reasons, a live preview of the Google result, and the page's real click and impression figures beside it |
| Broken links discovered by being told about them | Every dead address listed, with a one-click redirect |
| Site speed known only when somebody complained | Measured nightly, with the history charted |
| Nothing addressing AI assistants | A readiness check per page, and crawler rules that name the answer engines explicitly rather than leaving it to chance |
What was added to the editor
Moving the pages was the visible half. The other half was making the editor answer the questions that otherwise mean opening four Google products. Five dashboards, and four tabs beside the page you are writing.
None of it is a plugin, and all of it belongs to the client at the end of the engagement.
Beside every page
Dashboards of their own
Forms
Replacing the form plugin with “a contact form” would have been a downgrade dressed as a simplification. What went in is a builder, driven entirely from the editor.
Consent that blocks
Most cookie banners ask the tracking tools to behave. A tool that ignores the request, or that was added by a route the banner does not know about, carries on regardless.
This one sits in front of the browser's connection to the network and refuses the request itself. It governs Meta, Bing, Clarity and call tracking whether or not Tag Manager fired them, which is the gap the usual approach leaves open.
What is not claimed
Because a case study that only reports wins is not worth reading.
- Nothing here proves a ranking change. The rebuild is faster and better structured, and both correlate with ranking. Neither guarantees it, and anyone promising otherwise is guessing.
- AI visibility cannot be measured by anyone. Google publishes search figures; ChatGPT, Perplexity and Google's AI answers publish nothing equivalent. The site is built to be quotable, which is the part that can be controlled. Whether it gets quoted is not observable.
- The desktop layout is not pixel-identical. A page-by-page comparison at two screen sizes put mobile within one percent of the original and desktop about eleven percent more compact, from differences in how the two systems space and size content. No content is missing; it was checked element by element. It is a deliberate open item rather than an oversight.