A school website is never finished. Someone posts a vacancy on a Monday and a newsletter on a Thursday, the fee schedule changes every year, and the term dates go up before the last term has ended. The people doing that work are at the school, and they do it for years without a web developer anywhere near them.
Herschel Girls School is an independent Anglican school for girls in Claremont, Cape Town, founded in 1922. It teaches girls from Grade 000 to Grade 12 across a Pre-Preparatory, Preparatory and Senior School, and its website serves prospective parents, current families, staff, job applicants and a century of alumnae at the same time.
The school’s new website went live on 9 November 2026. On a phone, the homepage now shows its main content in 4,245 milliseconds, down from 9,538. Its mobile performance score went from 52 to 81, and the page no longer jumps about as it loads. The finished site is in our work section: Herschel Girls School Website Update 2026.
As with Rockfig Lodge, one thing up front: we built the previous site too. It ran on Elementor from 2021, and for five years the school’s own team kept it current, which is exactly what a school website is for. This rebuild moved it to Bricks Builder in a refreshed look. More importantly, it gave the people who run the site a structure that is much harder to get wrong.
The results in brief
- Mobile performance score: 52 to 81.
- Mobile time to appear: 9,538 ms to 4,245 ms.
- Mobile layout shift: 0.2875 to 0.0000253. Google counts anything under 0.1 as good.
- Desktop: the score held at 98, and time to appear went from 1,081 ms to 883 ms.
- Content carried over: 69 pages and 244 news posts, with every legacy address redirected.
- Builder: Elementor and JetEngine replaced with Bricks and ACF.
- Design: refreshed in the school’s current brand.
- Found during the launch: four redirect defects (one of them an infinite loop), 222 empty pages in the sitemap, an email key that had expired, and social sharing tag issues.
All performance figures are PageSpeed Insights lab tests of the homepage: the old site on 6 November 2026, the new site on 9 November 2026. What they do and do not prove is set out below.
Table of Contents
The stack that served the school from 2021: Elementor and JetEngine
The previous site was built on Elementor for layout, with JetEngine from the Crocoblock suite handling the custom content: vacancies, publications, Council members. For a school in 2021 that was a sound combination. It gave the school structured content and let its own team run the site without calling us to change a date, and that is how it ran for five years.
Three things changed underneath it.
The brand moved on. The school refreshed its identity, and the website was the most visible place still waiting to catch up.
The content outgrew the structure. Five years of news, newsletters, publications and vacancies adds up. By the time we moved it, the site held 69 pages and 244 news posts, all of which had to arrive intact.
In a page builder, editing content and editing layout are the same act. This is the one that matters most for a school. When the fee schedule lives inside a page layout, the person updating it is one careless drag away from moving a column or deleting a section. Nobody at a school should have to be careful with a layout to change a number.
None of that makes the old site a mistake. It did its job for five years. But the school needed a different kind of site for the next five.
The brief: a site the school’s team can keep current
Herschel wanted three things: a refreshed look in its current brand, nothing lost from the old site, and a structure that lets the people who run it update it with confidence.
The third is the real brief, and it is what a school website needs more than most business websites do. Many business sites change a few pages a year. A school site changes weekly, through many hands, for years. Vacancies open and close. Newsletters and the school magazine go up every term. Council members change. Admissions, fees and term dates roll over every year. Each of those is a recurring job, and each one should be a form to fill in, not a page to rebuild.
The second constraint was continuity. Every page, every news post and every old address had to arrive somewhere sensible, because a school’s website is linked to from years of letters, newsletters and other sites, and parents bookmark it.
What we replaced: Elementor and JetEngine out, Bricks and ACF in
- Elementor out, Bricks in. The layout now lives in shared templates. Apart from the homepage, every page and news post renders through them, filled from the content the school enters. The school edits words, images and fields. It does not edit layout.
- JetEngine out, ACF in. The school’s recurring content is now structured content with its own fields:
- Vacancies, each with a switch for an on-site application form. A vacancy can take applications on the site, or point applicants elsewhere, without anyone touching a page.
- Publications and newsletters, listed and downloadable.
- Council and committee, maintained as entries, not as text in a page.
- Page introductions, written in a field on each page, so the opening of a page can be updated without opening its layout.
- Forms that reach the right person. The Contact form routes to 11 departments. All six forms have spam protection. Notifications are sent through authenticated email from the school’s own address, and replying to one goes straight to the person who wrote in.
- A custom icon set. Icon fonts were replaced with a custom set of inline SVG icons. An icon font makes the browser download a whole font file for a handful of symbols, and until it arrives the icons show as blanks or stray letters. An inline icon is part of the page itself.
- An Alumnae Union section with a Digital Archive, so past pupils have a home on the site.
- Everything carried over. 69 pages and 244 news posts, with every legacy address redirected to its new home.
Because the old site was also ours, the migration was a conversion we understood from the inside, the same kind of Elementor to Bricks conversion we offer to any site. The difference here is that the design was refreshed at the same time, rather than held to 1:1 parity as it was on Rockfig.
The numbers: Herschel before and after
Measured with PageSpeed Insights lab tests of the homepage, on mobile and desktop. The old site was measured on 6 November 2026, three days before the switch. The new site was measured on 9 November 2026, launch day, after the layout-shift fix described below.
“Time to appear” is how long a visitor waits before the main image or heading shows on screen. Google calls it Largest Contentful Paint. “Layout shift” measures how much the page jumps about while it loads: the thing that makes you tap the wrong link because a logo arrived late and pushed everything down.
| Measure | Old site, 6 November 2026 | New site, 9 November 2026 |
|---|---|---|
| Mobile performance score | 52 | 81 |
| Mobile time to appear | 9,538 ms | 4,245 ms |
| Mobile layout shift | 0.2875 | 0.0000253 |
| Desktop performance score | 98 | 98 |
| Desktop time to appear | 1,081 ms | 883 ms |
On mobile, that is a 55.49% reduction in the time a visitor waits for the homepage to show its main content. On desktop the score was already high, and the time to appear still fell by 18.32%.
The layout shift, fixed on launch day
The mobile score did not arrive at 81 on its own. The new build scored 65 on its first test on the live site, and the reason was layout shift. We measured both causes before changing anything.
The logo had a width but no height. The browser reserved no space for it until the image arrived, and then every page jumped down by 97 pixels. The logo now has a height set at its exact proportion, so its space is held even before it loads. The jump went from 97 px to 0.
The body text had no fallback font. Until the school’s typeface, Montserrat, arrived, the text drew in the browser’s default font and then reflowed when the real one replaced it. The fix is a fallback that takes up the same space: Arial, scaled to match. We measured the scale on the site’s own copy rather than taking a published figure, and at weight 400 Montserrat is 112.03% the width of Arial. With the fallback in place, the reflow across the whole page fell from 270 px to 39 px.
That is why we test the live site on launch day and keep going until the number is right. The figure that matters is the one visitors get, not the one staging got.
What we are not claiming yet
- These are lab tests of the homepage only. They do not cover the news, vacancy or admissions templates, and a lab test simulates a phone rather than measuring real visitors.
- The new site was measured on a cold cache, in a run straight after the cache was purged. A run against a warm cache is likely to score a little higher, so the figures above are the conservative ones.
- The old and new sites ran on different servers. Some of the difference may come from the server rather than the build, and these tests cannot separate the two.
- Mobile time to appear is still above Google’s threshold. Google counts 2.5 seconds or less as good, and 4,245 ms is not there yet. The next gains are making the hero image discoverable earlier and lighter image delivery.
- There is no real-visitor data yet. Google’s field data from real Chrome users is collected over a rolling 28-day window, and it says nothing about admissions enquiries or applications. We will not claim an effect on either until there is a meaningful window to measure.
The part nobody could see
A launch is the moment a site’s history meets its new build. Everything that worked on staging has to work on the live address, with the real cache, the real email and the real search engines looking at it. Some of what follows came from five years of the old site’s service and some came from the move itself. None of it was visible to a parent looking at the homepage, and nothing reported any of it.
Four redirects that went the wrong way
Moving 313 pieces of content means every old address has to point to its new one. Before launch we followed each redirect hop by hop rather than trusting where the browser ended up, and found four defects.
One rule had been written backwards, with its source and destination swapped, which produced an infinite loop. A browser shows that as “too many redirects”. The list of redirect rules shows nothing wrong at all.
Another old address, the one for the Application Process page, landed on the Senior School Culture page. The cause was an image uploaded years earlier that happened to have the same short name as the page address. Images on the site redirect to the page they belong to, so the image claimed the address before the redirect rule could. Renaming the image fixed it without touching the file or any page that uses it.
All four were fixed before the switch. A redirect that goes to the wrong place does not show up as an error anywhere. It just quietly sends a parent from a bookmarked admissions page to something unrelated.
Every address checked on the live site
The site sits behind a caching layer: a network that keeps copies of pages close to visitors so they load faster. After a switch, that layer can go on serving its copies of the old pages until it is told to let them go, so clearing it is part of the launch, and so is proving that it worked.
We purged the cache and then checked all 545 addresses in the sitemap to confirm each one was serving the new build. Checking a launch from your own browser proves very little, because you are often not seeing what everyone else sees.
222 empty pages were in the sitemap
Publications, committee members and Councillors only ever appear inside listings. But by default, WordPress gives every entry of every content type its own page. So there were 196 Publication pages, 13 Committee pages and 13 Councillor pages, all with no content and no heading, all indexable, and all listed in the sitemap that tells search engines what the site contains.
It is the same shape of problem as the 13 blank pages we found on Rockfig. These pages are now kept out of search results and out of the sitemap, which now lists only real content. We left them resolving rather than turning them into errors, because their addresses may already be in Google’s index.
The site’s email had stopped going out
At launch we test every form by actually submitting it. The notifications were failing: the key the site used to send email through its previous email service had expired. Nothing on the site showed it. The form still thanked the visitor, and the email simply never left.
The site now sends through authenticated email from the school’s own address, and replies go to the person who filled in the form.
The social sharing tags needed tidying
When a link is shared on WhatsApp, Facebook or LinkedIn, the platform reads a set of tags in the page for the title, description and image to show. With a new builder and the SEO layer both able to print those tags, they had to be reconciled so that each page gives a platform one clear answer. Each page now has one set, and the default share image is a photograph of the girls outside the main building.
The same pass brought the school’s structured data, the information written for search engines and AI assistants to read, up to date: it now carries the school’s contact details and its current crest, and every page, vacancy and news post is credited to Herschel Girls School.
No page had a written search title
Google would have shown the homepage as “Home | Herschel Girls School”. We wrote search titles and descriptions for 34 key pages, from admissions and fees to the Alumnae Union, and listed them for the school to approve, because they are the school’s words in search results.
What this has to do with SEO and AI search
When a parent asks an AI assistant for an independent girls’ school in Cape Town that teaches from Grade 000 to Grade 12, the answer is assembled from pages the assistant can read and facts those pages state plainly. What gets a school cited is being readable, not being fast.
So look at what the site was presenting to a machine on launch morning, before the launch pass, with five years of settings carried across. 222 empty pages listed as content. Organisation details with no contact information and an out-of-date logo. Search titles that said “Home”.
The launch pass fixed that: written titles and descriptions on 34 key pages, school and page structured data that describes the school accurately, and a sitemap of real content only. It is the same work we do on any SEO and AI visibility engagement, and on a rebuild it belongs in the launch, not in a project six months later.
The takeaway: a school website is a content system first
The real test of a school website is not launch day. It is year three, when the admissions office updates the fee schedule, a new teaching post goes up and the newsletter goes out in the same week, and none of it needs a developer. The design gets a school noticed. The structure is what keeps the site right.
A launch is also an audit. Every item in the section above was invisible from the outside, and it was found because somebody followed every redirect, submitted every form and read every page’s code on the live address. None of it needed a clever fix. It needed someone to look.
If you are planning web design in Cape Town for a school, that is the order we would put things in.
School website rebuild: frequently asked questions
What does a school website need that a business website does not?
Mostly, a structure for content that changes all the time. A school publishes vacancies, newsletters, term dates, fee schedules and news every week, through many people, for years. Each of those should be a form with fields rather than a page layout, so the people updating it cannot break the design. On Herschel, vacancies, publications and newsletters, Council and committee, and page introductions all became structured content.
Can school staff update the new site without breaking the design?
Yes. That was the main brief. The layout lives in shared templates, and the school enters content into fields. A vacancy is a form with a switch for an on-site application. A Council member is an entry. Nobody has to open a page layout to change a date or a name.
Do we lose our news archive or old links when the website is rebuilt?
You should not. On Herschel, 69 pages and 244 news posts were carried over, and every legacy address was redirected. Before launch, each redirect was followed hop by hop, which is how we found four defects, including an infinite loop, that a normal check would have missed.
What should a website launch check?
At minimum: every redirect followed hop by hop, every form submitted for real, every address in the sitemap checked on the live domain once the cache has been cleared, the sitemap holding only real content, the structured data naming the right organisation, the social sharing tags set once, and a speed test of the live site. On Herschel, several of those found something worth fixing.
Why did the mobile score go up after launch rather than at launch?
Because the first test of the live site found layout shift worth fixing. The new build scored 65 on that test. A logo with no height and body text with no fallback font were making the page jump. Fixed on launch day, the score went to 81 and layout shift to 0.0000253.
Does structured data help a school show up in AI assistants?
It helps assistants describe the school accurately, which is the part a school controls. Structured data states the school’s name, address, contact details and logo in a form machines can read without guessing. It does not guarantee a mention. Readable pages, plain facts and a clean sitemap matter as much.
Thinking about your school’s website?
If your school’s team finds the website hard to keep current, or you are not sure what it is telling search engines, send us the address. We will run the same checks we ran on Herschel, from page speed to redirects to what the structured data says about you, and tell you plainly whether you need a rebuild or a tidy-up.



