Our own website is a project like any other, so we held it to the standards we'd set for a client. This is what's under the hood, and why we made the choices we did.
Content in a database, pages that ship as HTML
Almost everything you read on this site can be edited from a private admin panel: the services, the process, the industries, the team, the case studies and the Journal you're reading now. The content lives in MongoDB.
The pages, though, aren't assembled in your browser from an API. They're rendered on the server with Next.js and arrive as finished HTML. That matters for two reasons. Pages appear quickly, even on a slow phone. And search engines and AI assistants can read every word without running any JavaScript.
When something changes in the admin panel, only the affected pages are refreshed. If the database is ever unreachable, pages still load instead of showing an error: page sections fall back to their default text, and lists of services, case studies and articles show an empty state.
Motion with manners
The site moves a lot: headlines that settle into place line by line, scroll-linked sections, and a WebGL backdrop built with three.js that changes shape from page to page. Animation runs on GSAP, and smooth scrolling on Lenis.
Movement should never be the price of reading, so every animation checks your system's reduced-motion setting first. If you've asked your device for less motion, you get a calm, still version of the same page. Smaller screens get simpler layouts rather than a squeezed desktop experience.
One detail we care about: headlines that animate line by line are still written as ordinary sentences in the HTML. A screen reader or a search crawler reads "We build intelligent digital products that move businesses forward", not a string of fragments.
How do search engines and AI assistants read the site?
- a sitemap, and a robots.txt that welcomes search engines and AI crawlers while keeping private areas out
- structured data describing the company, each page, breadcrumbs, case studies and articles
- an llms.txt file, a plain-text brief for AI assistants generated from the same data the pages use, so it can't drift out of date
- IndexNow notifications when the site is deployed, so search engines hear about new pages quickly
None of it invents anything. Everything the structured data and llms.txt say comes from content that's already on the site.
A contact form that doesn't lose leads
When you send an enquiry, it's saved to the database first and emailed to us second. If email delivery ever fails, the enquiry is still waiting in the admin panel. It sounds obvious, but plenty of contact forms quietly drop messages when email breaks.
Private by default
The public site runs no analytics and no advertising trackers, and browsing it sets no cookies. Our cookie policy lists the two small settings your browser can store, and why.
Writing tools for the Journal
Articles are written in a block editor: headings, paragraphs, lists, quotes, images, code, callouts and video, each rendered in the site's own typography. A draft can be pasted in from a document and split into blocks automatically, and an article can be scheduled for a future date.
If you'd like something built with the same care, have a look at our web development services or start a project.
- #Next.js
- #CMS
- #SEO
- #Web performance
- #Accessibility
- #Animation
