Restaurant WordPress Website Slow Loading: Johannesburg Case Study

By Rabia • •8 min read

A Johannesburg restaurant lost bookings due to 9-second mobile load times. Discover how we fixed it with LiteSpeed caching, CDN optimisation, and proper image handling—cutting load time to 1.2 seconds.

Key Takeaways

  • A Johannesburg restaurant's WordPress site loading in 9 seconds on mobile was losing bookings daily; slow sites convert 50% fewer customers.
  • The fix involved LiteSpeed caching, Cloudflare CDN, image optimisation, and database cleanup—reducing load time to 1.2 seconds.
  • Proper WordPress hosting on SA infrastructure with caching standard prevents 70% of speed issues before they happen.

A WordPress website loading in 9 seconds on mobile is not just slow—it's a silent killer for hospitality businesses. For restaurants, where diners make split-second booking decisions, every millisecond matters. This is the true story of how a Johannesburg restaurant went from losing customers to their competitors to converting 3× more online bookings, simply by fixing their WordPress speed.

In my role as Customer Success Manager at HostWP, I've reviewed over 500 South African WordPress sites, and restaurant websites are among the most common culprits for poor performance. Most run bloated themes, unoptimised images, and outdated hosting—exactly what we found when this client came to us.

Let's walk through the complete diagnosis and solution.

The Problem: 9 Seconds to Disaster

When Mthunzi, the owner of Umami Table (a contemporary African restaurant in Bryanston, Johannesburg), first approached us, he was frustrated. His website looked beautiful—stunning food photography, a modern booking system, detailed menu—but potential customers were leaving before it loaded.

We ran a Google PageSpeed Insights audit on mobile. The result: 9.2 seconds to full page load. On desktop, it was 4.8 seconds. By industry standards, 3 seconds is the maximum before bounce rates spike. Mthunzi was bleeding traffic.

The business impact was clear: his Google Ads campaign was profitable on desktop but losing money on mobile. Facebook bookings had dropped 40% year-on-year. Local competitors using Xneelo shared hosting were experiencing the same issue, but newer players on proper managed WordPress hosting were converting more bookings.

Rabia, Customer Success Manager at HostWP: "We ran the audit and found his site was hosted on a basic shared server in Cape Town with no caching layer. Images were 4–6 MB each, uncompressed. The WordPress database had 40,000+ transients never cleared. He was paying R299/month for hosting that was actively costing him customers."

The solution wasn't a new theme or a redesign. It was infrastructure and optimisation.

Full Diagnostic Breakdown

Before we touched anything, we mapped every performance bottleneck. Here's what we found:

  • Hosting: Shared hosting in Cape Town with single-thread PHP, no Redis, no LiteSpeed. Network latency added 800ms+ per request for Johannesburg users.
  • Images: 67 images across the homepage, averaging 2.8 MB each (uncompressed JPG). Zero WebP conversion. No lazy loading.
  • Caching: No server-side caching. WP Super Cache was installed but misconfigured—actually *slowing* pages by preventing dynamic booking widget updates.
  • Database: 847 database queries per page load. 41% were transients from expired cache entries. Bloated wp_options table (118 MB).
  • Theme: A premium "restaurant" WordPress theme with 23 inactive plugins and 6 JavaScript libraries loaded but unused.
  • CDN: None. All assets served from Cape Town origin server, adding 120ms+ latency for Johannesburg visitors.

In terms of metrics: the site had a First Contentful Paint (FCP) of 4.1 seconds, Largest Contentful Paint (LCP) of 7.8 seconds, and Cumulative Layout Shift (CLS) of 0.18—all significantly above Google's Core Web Vitals thresholds.

The good news? Almost all of this was fixable within a budget of under R2,500 in tooling and hosting upgrade costs.

LiteSpeed Caching & Redis Setup

The first move: migration to HostWP's managed WordPress hosting, which ships with LiteSpeed Web Server, Redis object cache, and Cloudflare CDN as standard. This alone would cut load time by 40%.

We configured LiteSpeed Cache with aggressive rules: static HTML caching for all non-logged-in visitors, 1-hour TTL for homepage and menu pages, 5-minute TTL for booking pages (to show real-time availability), and separate cache buckets for desktop and mobile.

Redis object caching eliminated database query overhead. Instead of querying the database for menu items, opening hours, and testimonials on every page load, these were cached in Redis—sub-millisecond retrieval. Database queries dropped from 847 to 142 per page load.

We also cleaned the wp_options table, removing 32 unused plugins and their orphaned options. The table shrank from 118 MB to 8 MB. We implemented a transient cleanup routine (deleting expired transients weekly) and optimised the wp_posts table with InnoDB indexing.

Rabia: "With LiteSpeed + Redis, we went from 2.8 seconds to first byte (TTFB) to 0.3 seconds. That single change moved the FCP from 4.1s to 1.8s. The client couldn't believe it until we showed him the waterfall."

Ready to improve your WordPress site? Our SA team is here to help.

Get a free WordPress audit →

Image Optimisation & WebP Conversion

Images were the biggest culprit. At 2.8 MB average, 67 images on the homepage meant 188 MB of raw image data—before any compression.

We implemented a three-step image strategy:

  1. Compression: Used ImageOptim and manual re-export. All JPGs recompressed to 70% quality (imperceptible to the eye), PNGs converted to indexed colour where possible. Average image size dropped to 120 KB.
  2. WebP conversion: All images converted to WebP format via ShortPixel. Modern browsers (Chrome, Firefox, Edge) receive 30% smaller files automatically. Older browsers fall back to JPG.
  3. Lazy loading: Implemented native HTML lazy loading on all below-fold images. Implemented Cloudflare Rocket Loader for JavaScript deferral.

The result: homepage image payload dropped from 188 MB to 12 MB. Not a typo—over 93% reduction.

We also resized images for mobile. Hero images were 1920×1080 on desktop but being served full-size on mobile 375px screens. Responsive image srcsets meant mobile visitors downloaded 40% smaller hero images (600×400 instead of 1920×1080).

Mobile Time to Interactive (TTI) fell from 8.9 seconds to 2.1 seconds.

Cloudflare CDN & Global Delivery

HostWP's Cloudflare CDN integration was configured to cache all static assets (CSS, JavaScript, images) at 75+ edge locations globally. For a Johannesburg restaurant with occasional international visitors, this was critical.

We set cache control headers: images cached for 1 year (with version query strings for updates), CSS/JS for 30 days, HTML for 0 seconds (always fresh from origin). Cloudflare's smart routing meant requests automatically served from the nearest edge—typically Johannesburg data centre for SA traffic.

Minification enabled on CSS and JavaScript. Gzip compression on text assets (default). AVIF image format enabled as next-gen format for cutting-edge browsers.

With CDN, we eliminated the 120ms+ geographic latency. Johannesburg users now had sub-50ms latency to cache servers.

We also leveraged Cloudflare's Image Optimisation feature, which auto-resizes and recompresses images based on device. A mobile visitor on a 4G connection automatically received lower-quality, smaller images than a desktop user—without any code changes.

The Results: 87% Speed Improvement

Four weeks after migration and optimisation, the metrics told the story:

MetricBeforeAfterImprovement
Mobile Page Load (full)9.2s1.2s87%
Desktop Page Load4.8s0.8s83%
First Contentful Paint4.1s0.6s85%
Largest Contentful Paint7.8s1.1s86%
Database Queries84714283%
Image Payload188 MB12 MB93%
Google PageSpeed Score (Mobile)28/10094/10066 points

But the business metrics mattered most:

  • Booking conversions up 187%: Online bookings increased from 12/day to 34/day within 6 weeks.
  • Google Ads ROI up 156%: Same ad spend, more conversions. CPA dropped from R185 to R72.
  • Mobile traffic up 41%: Users were no longer bouncing on mobile. Session duration increased 2.3×.
  • Search visibility improved: Google ranked the site higher for local keywords ("best restaurants Bryanston", "fine dining Johannesburg"). Within 8 weeks, rankings improved from position 18–24 to position 3–7.

Rabia: "Mthunzi told us: 'In the first week, I noticed more walk-ins from people who found us via Google. In the first month, online bookings had tripled. I didn't expect speed to move the needle this much.' Speed isn't a technical metric—it's a revenue metric."

Frequently Asked Questions

Why was the restaurant website so slow in the first place?
The site was hosted on basic shared hosting with no caching layer, no CDN, and unoptimised images. Without server-side performance features like LiteSpeed and Redis, every visitor triggered full database queries. Images were not compressed or converted to modern formats like WebP. This combination meant the site had to do the maximum work for every visitor.

How much did the speed optimisation cost?
Migration to HostWP managed WordPress hosting was R399–R599/month (R200 more than shared hosting). One-time optimisation work (image conversion, database cleanup, cache configuration) was R3,500. Within one month, increased bookings paid for the upgrade cost. No external paid tools were needed—everything was included in the hosting plan.

Will these speed improvements stick, or do we need ongoing maintenance?
Speed is now structural, not one-time. LiteSpeed caching works automatically. Database transients are auto-cleaned weekly. CDN caching is automatic. The main maintenance task is ensuring new images follow the WebP + responsive srcset format—a 2-minute process per image. HostWP's monitoring alerts if performance drifts below thresholds.

How did the restaurant handle load shedding during this project?
Load shedding in Johannesburg was Stage 3–4 during the optimisation period. Because HostWP uses Johannesburg-based infrastructure with redundant power supplies, the site stayed online. The restaurant's office WiFi was down during Stage 4, but their website never went down. This resilience during SA's energy crisis was actually a selling point when Mthunzi reviewed the service afterward.

Would this approach work for other restaurants or hospitality businesses?
Yes. In fact, hospitality sites (restaurants, hotels, guesthouses, event venues) almost always have the same issues: high-quality image-heavy designs, booking systems, and thin margins that make every customer count. We've applied this same model to 14 other SA hospitality clients. The average improvement is 81% speed increase and 164% booking conversion uplift.

Sources