Case studiesInsurance
Australian Speciality Insurance Broker
Client name withheld: the rebuild is built and measured, but not yet signed off.
One broker's website, built twice. To a visitor the two look the same; measured, they are nothing alike. The WordPress site the broker runs today scores 29 for performance on a phone in Google PageSpeed Insights and loads 12.9 MB. Our rebuild scores 97 and loads 0.5 MB.
Outcome
Mobile performance in Google PageSpeed Insights, current site to rebuild
Client
Industry
Region
Services touched
- Web
01The challenge
A website that looked finished, and measured badly.
The broker places specialist insurance for Australian businesses, and its website runs on WordPress. To a visitor nothing looks wrong: strong photography, a clear headline, a quote button and a contact button.
Measured, it tells a different story. In the mobile test in Google PageSpeed Insights, the homepage scores 29 for performance. Its largest element takes 22.7 seconds to appear, well past the four seconds Google counts as poor, and the page downloads 12.9 MB across 176 requests. Accessibility scores 76, Best Practices 73 and SEO 83.
The structure underneath has the same problems. Across the 192 pages we could crawl, not one of 1,397 images carries descriptive alt text, 82 pages do not have exactly one main heading, only 171 of the 192 titles are unique, and the sitemap lists eight category pages by a relative address that search engines can't use. None of it shows in a screenshot, which is exactly the problem.
02The approach
Rebuilt to look the same, so only the engineering changed.
We rebuilt it to look the same on purpose: the same photography, the same headline and the same calls to action. With the design held still, any difference in the numbers can only come from how the site is built.
The rebuild is written from scratch in Next.js, with no theme and no page builder, and served from Vercel's edge cache, so its server answers in milliseconds. Images come in AVIF at the size the screen needs, where the current site sends the same full-size JPEGs to every device. Every one of its 180 pages has a unique title, a canonical address and exactly one main heading. Every image either describes itself in alt text or is marked as decorative. Structured data describes the business as an insurance agency on every page, with Service, Article and FAQ markup where each applies.
03The result
The same site to look at. A different site to use.
Measured side by side on the same day, the rebuild scores 97 for performance on a phone against the current site's 29, and 100 on desktop against 39. Its largest element appears in 2.6 seconds on a phone rather than 22.7. It downloads 0.5 MB in 29 requests, where the current site downloads 12.9 MB in 176. Accessibility, Best Practices and SEO all score 100, against 76, 73 and 83, and it passes the Agentic Browsing check the current site fails.
Nothing about the design changed to get there. That is the point of the comparison: two websites can look identical and perform nothing alike, and the difference is in how they are structured and built.
We have withheld the broker's name because the rebuild hasn't been signed off. The numbers are as measured.
Performance on a phone for the rebuild, against 29 for the current site
The current site, mobile
29The rebuild, mobile
97Performance · mobile · rebuild
97Performance · mobile · current site
29Performance · desktop · rebuild
100Performance · desktop · current site
39Accessibility · rebuild
100Accessibility · current site
76SEO · rebuild
100SEO · current site
83Google PageSpeed Insights, Lighthouse 13.5.0, both homepages measured side by side on 27 September 2026: medians of five runs per device. Current site: mobile 28–39, desktop 39–49. Rebuild: mobile 95–99, desktop 100 in all five. Accessibility and SEO shown for mobile; both sites were crawled from their own sitemaps the same day.
The deeper outcomes
Behind the headline metric.
- Performance on a phone: 97 for the rebuild, 29 for the current site, medians of five runs
- Performance on desktop: 100 in every run, against a median of 39
- Largest element on a phone: 2.6 seconds, against 22.7
- The homepage on a phone: 0.5 MB in 29 requests, against 12.9 MB in 176
- Accessibility, Best Practices and SEO: 100, against 76, 73 and 83
- Every image described or marked decorative, where none of 1,397 carried descriptive alt text
- A unique title, a canonical address and one main heading on all 180 pages
In the work
What's next
The rebuild is finished and waiting on sign-off, and the broker stays unnamed until then. When it goes live, the measurement moves from Google's lab test to the field: real-visitor Core Web Vitals, Search Console and the enquiry count, before and after, on the same pages.
Related
Other engagements that ship the same kind of outcome.
Start
Want this kind of outcome for your business?
The strategist below picks up the same lens we used for Australian Speciality Insurance Broker. Tell them what you're trying to grow. By the end of the conversation you'll have a plan, a quote, and a senior strategist on the line.