Renue Hawaii: Every Page Cleared 90 on Mobile. No Rebuild, No Redesign.
Renue Hawaii is a carpet, upholstery, and stone care cleaning company in Honolulu serving hotels and commercial properties. Their WordPress site worked. It just loaded like it did not want to. PeakSpire ran a Speed Sprint against the existing site, on the existing host, with the existing design. When the final gate closed, all 33 pages on the sitemap passed a 90 or better mobile PageSpeed gate. The design never moved.
The About page was the last holdout: hero paint went from 6.4s to 2.23s. Site-wide, the share of test runs scoring under 90 dropped from 40% to 6%.
The 90-second summary
| Metric | Before | After |
|---|---|---|
| Pages passing the 90+ mobile gate | 2 pages still failing entering the final week (About 75, Blog 83) | 33 of 33 (N=5 cache-busted runs per page) |
| About page, mobile PSI (n=10 median) | 75 | 96 |
| About page, mobile LCP | 6.4s | 2.23s |
| Blog, mobile PSI (n=10 median) | 83 | 92 |
| Blog, mobile LCP | 4.0s | 2.90s |
| Share of test runs under 90 (unique, mobile) | 40.0% | 6.2% |
| Mean mobile score across those runs | 91.0 | 96.3 |
| Worst single run | 82 | 88 |
| Best single run | 96 | 99 |
| Median mobile LCP | 2.18s | 1.75s |
| Mean mobile LCP | 2.45s | 1.98s |
| Theme rebuild / platform migration / design change | none | none |
Engagement: Speed Sprint, a fixed-scope mobile performance engagement for WordPress sites. Wrapped 2026-06-03 with the final plugin version deployed to both the production site and the staging mirror. Original Elementor design preserved. Original WordPress stack preserved.
The challenge
The site had two problems you can read straight off a PageSpeed report.
Pages were stuck below the 90 line. Entering the final week of the sprint, two pages still failed the gate: the About page at 75 and the Blog at 83. The About page was the worse one. Its hero background image was doing everything wrong at once: marked lazy-loaded when it was the first thing visitors should see, tagged low priority when it needed high, and shipped with no preload at all. On a phone, that page took 6.4 seconds to paint its main content.
Invisible text and late paint everywhere else. Two icon fonts, one weighing in at 785 milliseconds and the other at 640 milliseconds, were loading without a font-display rule, so text sat invisible until the fonts arrived. Elementor’s default entrance animations held the main content hidden for an extra paint cycle on every load. And script evaluation was blocking the largest contentful paint on enough runs that the worst-case score sat at 82.
None of this required a rebuild. It required someone to go page by page and fix what the page builder and the cache plugin could not fix themselves.
What we did: the Speed Sprint
Six changes to the existing WordPress and Elementor site running on LiteSpeed shared hosting. No theme rebuild, no migration, no design changes. Every change was staged and verified on a mirror before it shipped, and every change was reversible.
1. Critical CSS inlined in the head
The site’s combined CSS, about 496KB on production, inlined directly into the HTML head under a 560KB safety cap. The browser stops waiting on a render-blocking stylesheet round trip.
2. Hero images preloaded properly
Page by page, the true hero element got an explicit preload with the right mobile WebP variant, flipped from lazy to eager and from low to high priority. The About page fix alone moved it from 75 to 96.
3. Hero AVIF inlined
A 14.7KB mobile hero image in AVIF, embedded base64 directly in the HTML. First paint no longer waits on the network for the hero at all.
4. Icon fonts set to swap
The 785ms and 640ms icon fonts got font-display rules so text renders immediately in the fallback and swaps when the font lands. No more invisible text window.
5. Entrance animations stripped server-side
Elementor’s animation classes removed at the PHP layer, before the browser ever sees them. Content paints on the first frame it is able to.
6. Cache settings locked, verified on the wire
LiteSpeed Cache settings locked and every change verified by fetching the rendered HTML and checking the actual bytes, because server-side logs proved unreliable on this stack. Trust the wire, not the log.
The results
Two tables. The first is the final closure gate from the last week of the sprint: n=10 medians per page, mobile, after the About and Blog fixes landed. The second is the site-wide before and after at the earlier fleet-wide ship, which is where the bulk of the movement happened.
Final gate, n=10 medians per page, mobile:
| Page | PSI | LCP | CLS | TBT |
|---|---|---|---|---|
| Home | 97 | 1.84s | 0.037 | 0ms |
| About | 96 | 2.23s | 0.037 | 0ms |
| Blog | 92 | 2.90s | 0.037 | 0ms |
| Hotel Services | 92 | 2.24s | 0.107 | 0ms |
| Area Rug Cleaning | 94 | 2.45s | 0.037 | 0ms |
| Bleach Stain Removal | 95 | 2.70s | 0.037 | 0ms |
| Commercial Carpet Cleaning | 95 | 2.53s | 0.037 | 0ms |
Every page in that table cleared 90 with total blocking time of zero milliseconds.
Full-sitemap gate, all 33 pages, N=5 cache-busted mobile runs per page: 33 of 33 passed. Lowest page median was 90, the most common score was 96, and the highest was a perfect 100. Across all 165 individual runs, 92% scored 90 or better, 79% scored 95 or better, and six runs came back a perfect 100. Desktop passed 31 of 33, and both desktop misses traced to Google’s own testing workers choking, not to the site.
Site-wide before and after at the fleet-wide ship (mobile, deduplicated runs):
| Metric | Before | After | Change |
|---|---|---|---|
| Unique sub-90 rate | 40.0% (2 of 5) | 6.2% (1 of 16) | -84% relative |
| Mean score | 91.0 | 96.3 | +5.3 |
| Worst run | 82 | 88 | +6 |
| Best run | 96 | 99 | +3 |
| Median LCP | 2.18s | 1.75s | -20% |
| Mean LCP | 2.45s | 1.98s | -19% |
The floor moving from 82 to 88 is the part that matters most. The script-evaluation failures that were blocking the largest paint were engineered out entirely, so the worst experience a real visitor gets now looks like the old average one.
And it held. A post-wrap sweep re-verified all 60 user-facing URLs at 90 or better on mobile. A controlled re-measure of the homepage (19 runs, 60-second spacing) put the median at 98 and the mean at 96.1, with the share of runs below 95 cut from 47.2% before the hero fix to 10.5% after, roughly a 4.5x reduction in run-to-run variance. The 95th-percentile hero paint sits at 1.93 seconds, close to the practical ceiling for this hosting tier.
How these numbers were measured
- Medians, never single runs. PageSpeed lab results swing run to run on a real WordPress site. Two of the 165 final-gate runs came back under 70, and both traced to contention on Google’s testing workers, about 1.2% of runs. We disclose that instead of hiding it, because a single PSI run is noise and a median is signal.
- Every run cache-busted. Identical URL plus strategy makes PSI deduplicate and hand back stale results. Every run carried a unique cache-bust parameter so it measured a fresh load.
- Every page verified individually. A global change can lift one page while quietly sinking another. All 33 sitemap pages were gated, not just the homepage.
- Design verified pixel-by-pixel. The hero change shipped with a screenshot diff against the pre-change baseline: the mean pixel difference was 0.035 out of 255, a change no human eye can register. Speed work that moves a pixel is not speed work, it is a redesign nobody ordered.
What this means for a service business
Google uses page experience signals, including Core Web Vitals, in how it ranks mobile results, and a faster largest contentful paint measurably reduces the share of visitors who leave before the page finishes painting. What we will not claim is a ranking or lead-count jump for Renue: this engagement measured and fixed speed, and those outcomes were not instrumented. Anyone selling you a guaranteed ranking jump off a speed fix alone is selling you something they did not measure.
Renue Hawaii Speed Sprint, at a glance
- Client: Renue Hawaii, carpet, upholstery, and stone care cleaning, Honolulu. Site: renuehawaii.com.
- Engagement: Speed Sprint, fixed scope, mobile performance on every page.
- Final state: 33 of 33 sitemap pages at 90+ mobile, wrapped 2026-06-03, final plugin version live on production and its staging mirror.
- Design: unchanged, verified by pixel diff.
- Rebuild required: none. That is the entire point.
Or call directly at (343) 842-4444, or email info@peakspire.ca. Founder picks up.
About PeakSpire
PeakSpire is an Ottawa web design and local SEO company for service businesses, run by Siraj Ahmadzai. Focused services: Lead-Gen Websites, Speed Sprints, Local SEO, and Care Plans. Fixed pricing, founder-led, no agency fluff. More at peakspire.ca and About Us.