INDEX · 01 · OVERVIEW
SaVida Health · WordPress rebuild
Healthcare · 50+ clinic locations · MA · VA · OH · VT · ME
ENGAGEMENTTechnical SEO · Consulting scopeWINDOW2024 rebuild · in production since

A heavy WordPressstack, rebuilt for speed.

Before the content model could compound, the site had to load. A 5.8-second page, a bloated database, failing Core Web Vitals, and a 62% bounce rate were the pre-existing conditions the SaVida engagement inherited.This is the technical foundation that made the organic growth possible. Rebuilding the WordPress stack, pruning the database, tightening the template, and resetting the baseline. Crawlers need to reach the pages that rank. Patients in crisis don't wait six seconds for a site to load.
01 · Page load · LCP
1.9s
Down from 5.8s. Desktop, top-ten landings.
SOURCE · PageSpeed Insights
02 · Database weight
−68%
Rows + table size, post-cleanup
SOURCE · wp-admin DB tooling
03 · Core Web Vitals
Good
Failed → Good across LCP, INP, CLS
SOURCE · Search Console CWV
04 · Bounce rate
41%
Down from 62% post-rebuild
SOURCE · GA4 · same template window
INDEX · 02 · CHALLENGE
What the stack looked like on day one

A WordPress site that had been there too long, without a cleanup cycle.

Most clinical-network sites accumulate weight over time. Plugins stack. Themes fork. Post revisions never get pruned. Page builders leave deprecated widgets in the DOM. That's what we inherited here, and it was quietly throttling the content strategy above it.

  • 01
    A 5.8-second load time on the top ten landing pages, which is exactly where the highest-intent search traffic was landing.
    Patients typing “suboxone clinic near me” don’t wait six seconds. Every additional second on LCP above 2.5s compounds bounce rate exponentially in a crisis-intent category.
  • 02
    A database bloated with orphaned revisions, autosaves, and expired transients, slowing every uncached page query.
    68% of the database turned out to be deletable rows that hadn’t done useful work in months. The cleanup alone cut query time before a single server change.
  • 03
    Core Web Vitals in the Failed tier across LCP, INP, and CLS on the clinic-location-page template.
    The template that carries 70% of non-brand organic traffic was the worst-performing one. Every location page’s CWV score dragged the domain’s crawl priority down with it.
  • 04
    A 62% bounce rate on organic landings, masking the real underlying content engagement.
    When the page takes six seconds, you can’t tell whether the user bounced because the content was wrong or because they gave up waiting. Fixing the infrastructure was the only way to see if the content was doing its job.
  • 05
    A page-builder stack with deprecated widgets on 14 location pages, shipping unused JavaScript on every request.
    Dead widgets rendered invisible DOM, forced layout shifts, and bloated the main thread. We audited and cut every widget that didn’t participate in a real user interaction.
INDEX · 03 · THE REBUILD
What we changed, in order

Prune first. Rebuild the template second. Measure everything.

Performance rebuilds fail when they start with new infrastructure before the old infrastructure gets cleaned. We pruned the database and cut dead widgets before touching the template. Every change was isolated, measured, and reverted if it regressed CWV anywhere on the site.
STEP 01
Database cleanup.

Dropped orphaned post revisions, autosaves older than 30 days, expired transients, and residue from three uninstalled plugins. 68% row-count reduction, same content.

wp-admin DB · one-time sweep
STEP 02
Template rewrite.

Rebuilt the clinic-location-page template in native WordPress blocks + a thin custom loop. Killed the page-builder dependency that accounted for 14 deprecated widgets and ~40% of the main-thread work.

Core blocks + custom loop
STEP 03
Image + script discipline.

Images converted to WebP at responsive sizes via srcset. Render-blocking scripts deferred. Unused plugin scripts dequeued per-template so they only load where they run.

WebP · conditional enqueue
STEP 04
Caching + CDN tightening.

WP Rocket page cache rules reset to respect the updated template hierarchy. Cloudflare rules tightened to cache aggressively and bypass wp-admin only. HTML/CSS/JS hit the edge.

WP Rocket + Cloudflare
STEP 05
Core Web Vitals watchdog.

Search Console CWV report wired to Orbit. Every deploy that touches the site gets a post-deploy crawl and fails loudly if any page leaves the Good tier.

GSC CWV · Orbit alert
STEP 06
Post-rebuild burn-in.

60-day window before measuring the bounce-rate and content-engagement impact, so the lift was attributable to the infrastructure change, not seasonal traffic or content rotation.

60 days · GA4 attribution
INDEX · 04 · BEFORE / AFTER
The delta, measured

Four metrics. Same site. Different infrastructure.

Each row is the same measurement taken before and after the rebuild on the same URLs, same template, same content. No redesigns, no new pages. The content strategy was already working; it just needed the stack to stop getting in its way.
Page load time
5.8s
1.9s

Desktop LCP, measured in PageSpeed Insights across the top ten landing pages.

Database weight
−68%

Row count + table size after cleanup of orphaned post revisions, autosaves, expired transients, and legacy plugin residue.

Core Web Vitals
Failed
Good

All three metrics (LCP, INP, CLS) moved into Good across the clinic-location-page template, which carries most of the non-brand traffic.

Bounce rate
62%
41%

GA4, 30-day rolling window pre-rebuild vs. 30-day window 60 days post-rebuild. Same template, same content, faster delivery.

INDEX · 05 · WHY IT MATTERS
The technical story is half the story

A fast site did not cause the lead growth. It removed the reason the lead growth wasn't happening yet.

Most WordPress-performance case studies stop at Core Web Vitals scores. That misses the point. A clean stack doesn't earn leads by itself. It earns leads the content was already trying to produce but couldn't, because the infrastructure was eating half the intent before it reached the page.
What the rebuild unlocked

Four compounding effects.

  • Crawlers reached more pages. Faster server response lifted crawl budget; Search Console indexed coverage grew without adding content.
  • CWV feeds ranking signals. Good-tier Core Web Vitals on the clinic-location template removed a negative ranking factor that had been quietly capping the pages that carry the most non-brand intent.
  • Real engagement became visible. Dropping bounce from 62% to 41% wasn't the rebuild convincing people to read; it was the rebuild getting out of the way so the people who wanted to read could.
  • Attribution stayed honest. A fast site means the first-session CRM pixel fires on the actual first session, not after a reload. The +9% YoY organic-lead story is only believable because the measurement stack was clean.
The content story it supported

Same engagement, different case.

The content model + attribution infrastructure built on top of this rebuild is what drove SaVida's +9% organic-lead YoY growth through Google's AI Overviews rollout, while the industry lost 32% of its #1 clicks.

This page covers the foundation. The full engagement covers what was built on top of it.

Read the full SaVida case study
The claim

Your WordPress siteis probably slower than this was.

If your site predates 2022, there's a good chance it's carrying a database that hasn't been pruned, a template loaded with deprecated page-builder widgets, and Core Web Vitals that are quietly capping the pages you depend on to rank. A 5.8-second page doesn't just lose bounce; it loses crawlers, rankings, and attribution accuracy before any of that shows up in a report.