CarCare24

CarCare24 is a Magento 2 webshop for car maintenance with ~3,400 products and ~13,600 generated pages, running in four languages (NL, EN, DE, FR). The shop runs on Hypernode with Cloudflare Free on top, Infortis Ultimo theme, PHP 8.1.34-fpm, and as of May 2026, Magento 2.4.6-p15.

The challenge

Mobile load time was critical. Lab measurements via PageSpeed Insights (Slow 4G, 4× CPU) showed an LCP of 14 to 16 seconds. Not a single outlier, but nine repeated measurements with comparable results. For a shop in the automotive parts segment, that means lost conversions on mobile, the largest traffic segment.

Beyond LCP there were:

• A11y score at 90 due to missing alt attributes and contrast issues
• Best Practices at 88 due to CSP violations and unsafe-inline scripts
• SEO issues from missing canonicals on CMS pages

What I worked on

9 performance waves

Each wave targeted a specific bottleneck. In summary:

• Hero image — WebP conversion, loading="eager", fetchpriority="high". The LCP element was waiting on a 2× oversized JPEG.
• Third-party scripts — Defer gate added so tracking and chat only load after user interaction. Significantly reduced blocking time on slow 4G.
• Redis and FPC — Magento Full Page Cache and session storage tuned. TTFB dropped from ~974ms to ~390ms in CrUX over 28 days.
• PHP-FPM workers — Configuration adjusted to the Hypernode worker model. Less queue wait time under load.
• Checkout INP — Reduce cart abandonment via faster interaction in checkout. INP improved for selected flows.
• SRI hashes — Race condition found where SRI hashes were out of sync with served assets post-deploy (P1 incident). Resolved via Tieman_Csp module.
• CMS canonicals — CanonicalForCms component added to prevent duplicate content on flat pages.

Custom Tieman_Csp module

Four components in one module:

• CanonicalForCms — Canonical tag on CMS pages where Magento does not set it automatically
• DeferThirdPartyScripts — Interaction gate for all non-essential scripts
• InteractionGate — Generic defer wrapper for lazy-loaded widgets
• SriRepositoryLock — SRI hash synchronization, fixes the race condition at deploy time

Multi-tenant context

CarCare24 runs as stores 1 through 4 plus website 1 in a multi-store database. Every change is built multi-store aware so all four languages benefit consistently.

Magento p14 to p15

On 15 May 2026, the upgrade from 2.4.6-p14 to 2.4.6-p15 was completed. I supported the deploy process and handled post-deploy verification.

Results

Lab vs CrUX — what do these numbers mean?

Lab scores (PSI) are worst-case: Slow 4G simulation, 4× CPU throttling, no CDN benefit, no warm cache. The range 38 to 65 mobile and 71 to 92 desktop reflects PSI variance of ±15 points on the same page. A single run at 8.6 seconds LCP says less than the average across nine runs.

CrUX is real user data from Chrome users over 28 days. Those 2.14-second LCP and 111ms INP are what visitors actually experience. All Core Web Vitals passing.

Stack

• Magento 2.4.6-p15, PHP 8.1.34-fpm
• 328 enabled modules (incl. Tieman_Csp custom)
• Infortis Ultimo theme
• Hypernode hosting, Cloudflare Free
• ~3,400 products, ~13,600 pages
• 4 store views: NL, EN, DE, FR

What this difference means

An LCP that goes from 15 seconds to 2.14 seconds for real users is not a benchmarking exercise. Visitors on mobile see the hero within two seconds. Checkout runs faster. SRI issues that caused a production incident during a deploy are structurally resolved, not temporarily patched.

Lighthouse Before and After

Two measurements via Google PageSpeed Insights (Slow 4G, Emulated Moto G Power, Lighthouse 13.0.1). Left is the starting point, right is after the performance waves.

Lighthouse measurement before and after the performance waves. Lab data via Google PSI.

---

Is your webshop also too slow on mobile? Get in touch for a no-obligation analysis.