Marketing site rebuilt on a maintainable stack after a page-builder dead end

Every landing page had been assembled in a visual builder by a different person, so the same section existed in five slightly different versions. Page weight had grown past what mobile visitors would wait for, and editors could not change a headline without breaking a layout.

Engagement

Role: Fixed-scope web development build, founder-led.

Shape: Content inventory and redirect map first, then the component build.

Duration: Two weeks for the marketing site tier.

Stack: React 18 · TypeScript · Tailwind CSS · Vite · AVIF/WebP pipeline · Schema.org JSON-LD.

What was wrong

  • The same section existed in five near-identical builder variants, so a copy change had to be made five times.
  • Page weight was dominated by unoptimised hero images with no width or height attributes, causing layout shift.
  • Several landing pages competed for the same query with no canonical relationship between them.
  • There was no staging environment, so edits were made on the live site.

What was built

  1. Content inventory of every live URL, mapping each one to keep, merge or redirect before a line of code was written.
  2. React and TypeScript rebuild with a small set of reusable section components instead of per-page markup.
  3. Image pipeline emitting AVIF and WebP with explicit dimensions, and fonts preloaded with matched fallback metrics.
  4. 301 map for the merged URLs, regenerated sitemap and structured data for the organisation, breadcrumbs and FAQs.

Architecture after the work

Component inventory — Every page is assembled from a small named set of section components, so a change to a pattern applies everywhere it is used.

Deterministic media — Images are emitted as AVIF plus WebP with explicit dimensions and a preloaded LCP hero, and fonts ship with matched fallback metrics.

One page per intent — Near-duplicate landing pages were merged and the retired URLs 301-redirected to the surviving page, with self-referencing canonicals throughout.

Machine-readable pages — Organisation, breadcrumb and FAQ structured data is emitted from the same data that renders the page, so the markup and the schema cannot drift apart.

Trade-offs and what was left out

  • The blog stayed on the existing CMS for this phase — moving it was not required to fix the landing pages and would have doubled the scope.
  • No traffic or conversion figures are published here, because a reader cannot verify them.

Outcome

Marketing can publish a new page from existing components without a developer, and every old URL resolves to a single canonical page instead of a near-duplicate.

Frequently asked questions

Is it worth rebuilding a page-builder site rather than tidying it?
It depends on how many variants exist. When the same section has been rebuilt by hand on every page, tidying is per-page work that never ends. A component rebuild converts that into a one-time cost, after which editors reuse patterns instead of recreating them.
What happens to rankings when pages are merged?
Every retired URL is 301-redirected to the page that replaced it, and the surviving page carries a self-referencing canonical. The redirect map is written and reviewed before launch rather than discovered afterwards from crawl errors.

Related services