
The Scenario: A Site That Looks Better and Ranks Worse
A Johannesburg homeware store ships nationwide. In Q3 the team launches a bold redesign: new hero, filterable catalogue, new blog template. Organic sessions look flat for six weeks, then slide. The site is 'better' visually, but organic revenue began leaking the day the templates went live, because design decisions quietly removed signals Google used to rank and value the site.
This is a diagnostic sequence for that situation. It treats design as build decisions that either preserve crawl equity, entity clarity and conversion value - or destroy them. Core Web Vitals tooling and mobile-first indexing are covered elsewhere; the real leaks sit deeper.
Why 'Page Experience' Is Not a Ranking System
Google states there is no single 'page experience' system and that meeting experience signals does not guarantee rankings. They are tie-breakers within a relevance system. When a redesign drops rankings, suspect relevance and crawlability before aesthetics.
Core Web Vitals thresholds (field data, not lab data)
LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1, assessed at the 75th percentile of field data. INP replaced FID in March 2024: a redesign adding heavy interactivity can pass in a lab and fail on mid-range Android on 4G - the profile of many South African shoppers, where data cost shapes abandonment.
Lighthouse lab scores are a design review tool, not a ranking verdict. Pull field data before concluding a redesign 'failed Core Web Vitals'.
Where design ends and content begins
Design controls structure: what is in the HTML, how deep it sits, how templates render it. Content controls words and entities. Leaks happen at the seam - hiding text, flattening headings or burying a page changes content signals whether intended or not.
Leak 1 - Information Architecture That Hides Value
Navigation depth, orphan pages, internal links
Redesigns often move pages from a flat menu into nested categories; pages once in the global nav sit four clicks deep or only in a footer accordion. Orphans lose link equity and are crawled less. Fix: a link-depth spec in handover - every revenue page within three clicks, descriptive anchors agreed before build.
Tabs, accordions, infinite scroll
Accordion content is usually indexable if in the DOM, but may be devalued as secondary and truncated in mobile-first rendering. Infinite scroll lazy-loading products behind JavaScript is riskier. Prefer paginated or progressively enhanced lists with real URLs.
Leak 2 - Template Churn and URL Restructuring
Redirect chains, canonical drift, the 90-day watchlist
Every 301 hop costs latency and dilutes signal; chains of two or three are worse. Canonicals drift when new templates hardcode a canonical pointing at a listing rather than the detail page. Treat the first 90 days as an active watchlist: crawl errors, canonical mismatches, redirect chains and index coverage, reviewed weekly.
Change-of-address hygiene
If paths genuinely change, implement 301s server-side, update internal links and sitemaps, and use Search Console's change-of-address tool for domain moves. Keep old URLs responding with 301s for at least a year.
Leak 4 - Semantic HTML and Structured Data as a Design Deliverable
Semantic markup is a design deliverable, not a developer afterthought. When handover specifies one H1, ordered H2/H3 nesting, landmark regions and descriptive alt text, the build inherits them. When it does not, agencies default to divs and image buttons, and Google reads a flatter document.
Google's structured data policy is blunt: markup must reflect visible content. A template that renders an FAQ accordion visually but leaves stale FAQPage JSON-LD creates a mismatch. Keep schema in version control and review it in the same pull request as the visual change.
Add an acceptance line to every design ticket: 'Deliverable includes heading hierarchy, landmarks, alt-text requirements and any schema.org changes.'
Leak 5 - Intrusive Interstitials and the Conversion Team's Footprint
Designers rarely own pop-ups, but they design the slot. Google's intrusive interstitial guidance applies to mobile: a full-screen overlay immediately after leaving search, or content pushed below the fold by a pop-up, is the problem. Cookie banners blocking main content count; a small dismissible strip does not. For South African sites, POPIA consent scripts are a common source of third-party load and layout shift, so they belong in the design spec.
This is a user-experience and layout point, not a legal compliance claim. Confirm consent wording with your legal team.
Leak 6 - Entity Clarity and Trust Signals
E-E-A-T is a rater concept, not a dial you turn on. Raters and users look for authorship, business information and visible trust pages. Redesigns often bury About, Contact and policy pages in footer accordions or turn author bios into tooltips. Surface company details, registered address, author names and dates at template level. An entity that cannot be identified is harder to trust.
Which Design Fixes Move Rankings and Which Are Myths
| Fix | Realistic effect | Myth to drop |
|---|---|---|
| Semantic HTML + canonical hygiene | Preserves crawl equity | 'New fonts lifted rankings' |
| SSR for revenue templates | Faster discovery | 'JS is invisible to Google' |
| Interstitial rework | Removes a mobile demotion risk | 'Pop-ups always boost CVR' |
Connecting Design to Commercial Outcomes (CVR, CPA, MER)
A design fix that improves indexation but not conversion is half a job. Track CVR, CPA and blended MER before and after launch using GA4 and, where practical, server-side tracking so consent and script blocking do not distort attribution. Prebo Digital's custom reporting approach treats attribution accuracy as a design-adjacent problem, and our conversion rate optimisation work ties template decisions to funnel metrics rather than vanity traffic.




