Boiler Decoder is our own content site and the first case study we publish, because we can show every decision behind it. It is a UK reference for boiler fault codes, built on a custom CMS our team wrote, and it is where we test the site architecture, technical SEO and publishing methods we later use for clients. This website runs on the same CMS.
What Boiler Decoder Covers
When a UK boiler shows a code such as F28 or EA, the homeowner wants three answers quickly: what the code means, whether there is anything safe they can check themselves, and whether they need a Gas Safe registered engineer. Boiler Decoder answers those three questions for the major UK brands: Vaillant, Worcester Bosch, Ideal, Baxi, Glow-worm, Viessmann, Vokera and Potterton.
The site is live at boilerdecoder.co.uk. Its sitemap lists 82 URLs: 52 fault-code pages across those 8 brands, 8 brand hubs, 14 guides (6 troubleshooting guides and 8 cost guides, such as new boiler and powerflush costs), plus the homepage, two section hubs and five static and legal pages. Content is taken from manufacturer documentation, such as user manuals and installation and servicing instructions, cross-checked across sources, and every page is dated.
The fault-code page template
Every fault-code page follows the same template. The H1 gives the code and its meaning. A causes table lists each likely cause, its signs and who fixes it. A numbered list of safe checks covers what a homeowner can do (such as checking the pressure gauge or a frozen condensate pipe), followed by clear criteria for when to call a Gas Safe engineer, a UK repair cost table, prevention tips, a short FAQ, links to other codes for the same brand, related guides and breadcrumbs.
Action labels and safety
Every code carries one of three action labels: DIY fix, Safe checks or Engineer. Brand hubs show a filterable code table (code, meaning, who can fix it) with a summary line counting how many codes fall under each label. Because gas work is dangerous, a gas emergency banner with the national gas emergency number appears on every page, and the Gas Safe requirement is repeated wherever a job goes beyond safe checks.
The Challenge of Fault-Code Content at Scale
Fault-code searches are a good fit for a content site: the intent is specific, the questions repeat across brands, and the existing results are often manufacturer PDFs or forum threads that are hard to read on a phone. The difficulty is doing it at scale without producing thin, repetitive pages that search engines ignore.
We also wanted to answer a question clients ask us often: does a content site need WordPress? A typical WordPress build for this job means a theme, a page builder, an SEO plugin, a caching plugin, a schema plugin, an analytics script and a cookie banner, all updated separately. We wanted to see how much faster, lighter and easier to maintain a purpose-built CMS could be.
What We Built for Boiler Decoder
- A content silo of brand hubs and fault-code pages. The site follows the LLMS silo structure we use on client sites. Each brand has a hub page that introduces the brand, lists its codes and links to every fault-code page for it. Each fault-code page links back up to its hub and across to related codes, so both visitors and crawlers can move through the site logically.
- A custom PHP 8 and SQLite CMS in one folder. No WordPress, no plugins and no separate database server. The whole site, including the admin area, lives in one folder that can be copied to any PHP host. SQLite keeps content, settings and analytics in a single file, which makes the site fast and simple to move.
- Drip-feed publishing. Pages are written ahead of time and queued. The CMS publishes a set number per day at randomized times inside a daily window, so the site grows at a steady pace rather than dumping hundreds of pages at once.
- Hidden links to unpublished pages. Content can link to pages still in the queue. The CMS hides those links until the target page goes live, so visitors and crawlers never hit a 404 and the internal linking fills in automatically as the site grows.
- IndexNow pings on publish. When a page goes live, the CMS notifies search engines that support IndexNow, such as Bing. Google does not use IndexNow, so for Google the site relies on the XML sitemap and Search Console.
- Structured data and AI crawler support. Pages carry JSON-LD for WebSite, Organization, BreadcrumbList, Article and FAQPage where relevant. The CMS also generates an llms.txt file that describes the site and its main sections for AI assistants and answer engines.
- Automatic sitemap and robots. The XML sitemap and robots.txt are generated from the live content, so they never list unpublished pages or miss new ones.
- Cookie-free analytics. Visitor analytics are built in without cookies. Visitors are counted with a hash that rotates daily, raw IP addresses are never stored, and the visitor's country comes from an offline lookup database, so no data is sent to a third party. Normal visitors never receive a cookie, so analytics work without tracking anyone across sites or days.
- A keyword screen in the dashboard. The admin area pulls query data from Google Search Console and Bing Webmaster Tools, so we can see which searches each page appears for without leaving the CMS.
- One-click backups. Because the whole site is one folder and one SQLite database file, a full backup is a single download from the admin area, and restoring it means replacing one file.
Boiler Decoder Results So Far
We do not publish traffic, ranking or revenue numbers for Boiler Decoder yet. A content site in a new niche needs months of Search Console data before any numbers mean something, and we would rather wait than show an early spike out of context. Results will be published here, with Search Console named as the source and the date range stated, once enough data is in.
What we can report now is qualitative. The site is live with 52 fault-code pages and 14 guides, and we applied for Google AdSense in September 2026, because the site is funded by advertising. The drip-feed queue, hidden-link system and IndexNow pings have run as designed. Pages are light and fast because there is no plugin stack to load. The keyword screen shows which queries each page is starting to appear for, which tells us where to expand content next.
What Boiler Decoder Proves for Client Projects
Boiler Decoder is a working example of three services we sell.
- Content architecture. The brand hub and fault-code silo is the same structure we plan for ecommerce categories, service areas and knowledge bases. The principle is the same everywhere: one page per search intent, clear parent and child pages, and internal links that follow the hierarchy.
- Technical SEO. Clean URLs, automatic sitemaps, structured data, fast pages and no crawl traps are the foundations of our technical SEO services. We have built all of them from scratch, so we know how each one works rather than only which plugin setting to change.
- Custom CMS builds. If WordPress is more than your site needs, or a plugin stack has become slow and fragile, custom CMS development can give you a lighter system built around your content model. The Boiler Decoder CMS is the proof that this does not have to be a large, expensive project.
The llms.txt file, the structured data and the plain question-and-answer format of each page also reflect how we approach AI search optimization: pages that answer one question clearly are easier for AI assistants to understand and cite.
If you are planning a content site, a knowledge base or a move away from a heavy CMS, tell us about the project and we will explain what we would build and why. You can also compare this with our other product case study, AM Jarvis.