Why architecture, not pages
The instinct after "our site underperforms" is to fix pages: rewrite the homepage, redesign the services page. But underperforming B2B sites usually have a sounder diagnosis at the architectural level — the pages exist in isolation, with no engineered relationships between them. The blog never mentions the services. The service pages cite no work. The work has no detail. Every journey dead-ends.
What follows is the pattern language we use — on this site, visibly, if you want to see it in practice.
Layer 1: the content system
Pattern: one hub per topic, no orphan articles. Content organized around a small number of clusters (topics the business can credibly own), each with a hub page that defines the topic and links its articles, services and proof. Articles without a cluster die in isolation; clusters without articles are empty promise. The test: every article can name its cluster, and every cluster can list its articles without embarrassment.
Pattern: answer-first structure. Each page answers its implicit question in the first screen — what it is, why it matters, who it is for — then earns the depth. This serves buyers (most skim, few scroll) and retrieval systems alike, which extract the early, clear statements. The anti-pattern is the throat-clearing introduction that says nothing until paragraph four.
Pattern: publish on evidence, not on calendar. A smaller corpus of pages only the company could write — from real projects, real decisions, real measurements — outperforms a large corpus of competent generic prose. Quality is a growth strategy; volume without evidence is a cost center.
Layer 2: the entity system
Search engines and AI answer engines do not read pages so much as resolve entities: who is this company, what do they do, what have they built, who stands behind the claims. Engineering that legibility is now a first-class concern:
Pattern: a stated identity, everywhere identical. One organization description, one founder identity, one set of facts — repeated consistently in the about page, the profile pages, the author bylines and the structured data. Contradictory self-descriptions across pages quietly cost trust with every system trying to model the company.
Pattern: schema that only states facts. Organization, WebSite, Person for the real author, Article with authorship and dates, Service for offerings, SoftwareApplication for products, BreadcrumbList for position. Ratings, reviews, awards, client counts that do not exist must never appear in markup — invented evidence is discoverable and poisonous.
Pattern: real authorship. Articles carry a person — name, role, profile link — and the profile links back to verifiable work. Anonymous "admin" bylines and fake editorial teams both fail the same test: there is no human whose reputation the claims touch.
Pattern: machine-readable structure without exclusivity. Clean semantic HTML, stable canonical URLs, a complete sitemap, honest robots rules, and a plain-text identity file (an llms.txt) summarizing the entity for retrieval systems. None of it is a hack; all of it is legibility.
Layer 3: the conversion system
Pattern: no dead ends. Every high-value page ends with a contextually matched next step — and "matched" means matched to intent. A performance article ends with an assessment offer; a WooCommerce article ends with a commerce-engineering offer. A generic "Contact us" slapped onto every page is the anti-pattern that trains visitors to ignore the footer of every page.
Pattern: proof adjacency. Claims sit next to evidence. The service page links the project that demonstrates it; the project links the articles that explain its engineering; the article links both. Proof adjacency is what separates a website that asserts expertise from one that demonstrates it.
Pattern: the primary CTA is a product, not a page. The strongest B2B conversion paths lead into a structured process — an intake, an assessment, a scoping conversation with defined inputs — rather than a contact form that opens an unstructured inbox. The process itself is a signal of how the engagement will run.
The anti-pattern catalogue
- The isolated blog — articles published into a void, linked from nothing, linking to nothing.
- The services page with no work behind it — capability claims with zero demonstrated evidence.
- The portfolio of screenshots — work shown without decisions or outcomes, indistinguishable from a template.
- The everywhere-CTA — identical contact blocks on pages with completely different intents.
- The schema mismatch — markup claiming what the page does not support; search systems notice, and so do buyers.
Seeing the pattern language applied
This site runs all three layers in public. Clusters with hubs (this article belongs to one), articles with a named author who has a verifiable profile, a project archive with real systems, schema that states only facts, and every article ending in an intent-matched next step — including this one.
If your site needs its three layers engineered into that shape, the assessment is where we would start: what exists, what connects, what is missing — and the shortest path to a site that works as a system.