Architecture is the set of relationships, not the folder names

Website architecture is how pages are grouped, labeled, reached, and related. URLs are one visible expression of that structure, but architecture also includes main navigation, subnavigation, breadcrumbs, contextual links, hub pages, search, filters, and rules for adding content. A site can have tidy-looking folders and still strand important pages. It can also have a short flat URL scheme while its navigation clearly groups services and guidance. Plan the information relationships first; decide the route format second.

Google recommends organizing a site logically because it can help users and search engines understand how pages relate. The same official guidance warns against overreacting: search engines can often understand an existing site even when its organization is imperfect. Rebuilding every URL solely to make folders look cleaner can create migration risk with little user benefit. Our SEO and AI-search services begin with business goals, content inventory, crawl evidence, and user pathways before recommending structural change.

Inventory decisions before drawing a tree

Five inputs for a defensible architecture

01

List business capabilities

Document the services, products, locations, audiences, industries, support needs, and transactions the business can actually deliver. Separate current offers from ideas.

02

List customer tasks

Capture what people need to learn or do: understand fit, compare approaches, estimate cost, check availability, see proof, meet requirements, book, buy, or get support.

03

Inventory current URLs

Record status, canonical, indexability, traffic, links, conversions, content owner, last review, and likely action for every page. Do not design a replacement without knowing what already works.

04

Study search language

Use Search Console, customer calls, proposals, site search, support requests, and primary research to understand how people distinguish needs. A query list informs labels; it does not dictate one page per phrase.

05

Define constraints

Include CMS limits, permissions, localization, regulatory requirements, product feeds, integrations, publishing volume, team capacity, and migration tolerance. Architecture must be operable after launch.

Add a proposed purpose to each future URL in one sentence: This page helps [audience] decide [task] by providing [distinct information]. If two rows produce nearly the same sentence, they may belong on one stronger page. If a page serves several incompatible tasks, split it only when each new page can stand on its own and has a clear discovery path. This exercise prevents a keyword spreadsheet from creating hundreds of thin pages the team cannot distinguish or update.

Choose hubs that reflect stable business concepts

A compact service-business hierarchyThe tree is shallow enough to understand, but the right depth is determined by meaningful groups rather than a fixed number.
01HomeDefines the company and routes major audiences
02Services hubExplains the offer family and links specific services
03Specific serviceCovers scope, fit, process, proof, and action
04Evidence and decisionsCases, pricing, comparisons, requirements, and FAQs
05Location or industry contextAdded only where genuinely distinct information exists

A hub is more than a card grid. It introduces the group, explains how choices differ, and helps a person select the next page. A services hub might distinguish strategy, design, development, ecommerce, maintenance, and SEO. A resource hub might organize buyer questions into pricing, hiring, redesign, search visibility, conversion, and platforms. The SEO, GEO, and AI-search hub is an example of using one clear category to orient readers before they choose a specific guide.

When a hub deserves its own indexable page

SignalCreate or keep a hubUse navigation without a hub
User needPeople must understand or compare the group before selectingMost users already know the exact destination
ContentThe overview provides unique definitions, criteria, and choicesThe page would only repeat child-page summaries
ScaleSeveral durable child pages need a stable parentOnly one or two short-lived destinations exist
Search purposeThe category itself answers a distinct broad needThe category label is merely an internal filing term
OwnershipA team can maintain the overview and child relationshipsNo one can keep a separate summary accurate

Keep important pages easy to reach without worshiping depth

You may hear that every page must be within three clicks of the home page. Google does not publish a universal three-click ranking rule. Click depth is still a useful diagnostic: if a primary service requires opening an archive, selecting a year, paging through results, and expanding a script-only control, people and crawlers may struggle to find it. The solution is not to place all 500 URLs in the main menu. Give priority pages clear routes from relevant hubs, navigation, body links, breadcrumbs, related content, or search.

Artificially flatUnderstandably layered
Main menuEvery service, city, industry, and article listed togetherA few durable choices with descriptive subnavigation
Home pageHundreds of links added to reduce crawl depthMajor routes, proof, and high-value next steps
Article archiveOne chronological stream with no useful categoriesSearch, categories, related guides, and contextual paths
Service familyNo overview; every URL competes for attentionA hub explains the group and differentiates child services

W3C's WCAG 2.2 guidance on multiple ways says users should have more than one way to locate pages within a set, except for steps in a process or results of one. Depending on site size, that might mean navigation plus related links, a table of contents, a sitemap, or site search. A blog-only search can help people who remember a topic but not its category. Search should complement, not replace, crawlable category and article links.

VISUAL CHECKPOINT · SearchA compact service-business hierarchy

The tree is shallow enough to understand, but the right depth is determined by meaningful groups rather than a fixed number.

Design URLs for humans and long-term operation

Google's URL guidance recommends simple, descriptive structures with readable words, the audience's language, and hyphens between words. URLs are case sensitive to clients and search systems, so choose lowercase conventions and enforce them. Avoid session IDs, unnecessary parameters, fragments used to swap indexable page content, and filter combinations that create vast numbers of near-duplicate URLs. Use root-relative internal links when appropriate so a misplaced parent-relative path cannot generate nonsense routes.

URL decisions with operational consequences

DecisionPreferReason
Words/services/website-redesignA person can recognize the destination
Separatorswebsite-redesignGoogle recommends hyphens rather than underscores
Caselowercase routesConsistent casing avoids accidental duplicate URLs
Date foldersUse only when date is structurally meaningfulEvergreen guides should not look obsolete or require moves each year
ParametersAllow only those needed for real functionsSorting, filters, sessions, and tracking can multiply crawlable variants
Future changeStable labels based on durable topicsAvoid routes tied to staff names, campaign slogans, or temporary org charts

Coordinate navigation, breadcrumbs, links, and sitemaps

  • Main navigation exposes a small set of durable, high-value sections
  • Subnavigation distinguishes sibling pages within a service, product, or resource group
  • Breadcrumbs show one understandable path upward and use real crawlable links
  • Body links connect related decisions across categories when the relationship helps
  • Related-content modules are curated or based on strong editorial rules rather than tags alone
  • Site search covers content people reasonably need to find and returns useful no-result guidance
  • The XML sitemap lists canonical, indexable URLs and excludes redirects, errors, and blocked duplicates
  • HTML navigation remains the primary route; the sitemap helps discovery but does not guarantee indexing

These systems should tell a compatible story without being identical. A breadcrumb expresses hierarchy, while a contextual link can cross between related branches. An XML sitemap can include all preferred indexable pages even though the main menu includes only a few. Google says a sitemap can improve crawling for large, new, complex, media-heavy, or poorly linked sites, but proper internal linking usually lets it discover most pages. The detailed small-business internal linking guide explains how to build those contextual routes without a link quota.

Scale locations, industries, and articles with admission rules

Growth plans fail when a page type has no admission standard. A location page should require a real service relationship plus location-specific operating details, proof, staff, projects, regulations, or useful local context. An industry page should require distinct needs, language, examples, requirements, workflows, or evidence. An article should resolve a researched question not already answered. These pages may link to shared service pages, but each needs a distinct reason to exist beyond inserting a place or industry name.

A content-type admission checklist

01

Define required evidence

Specify the facts, examples, subject-matter input, and source material a page must contain before publication.

02

Define the parent

Every page type needs a hub, navigation path, or contextual parent that helps people discover and interpret it.

03

Define related links

State which services, cases, locations, industries, or guides may be linked and what makes each relationship legitimate.

04

Define metadata and schema

Map unique titles, descriptions, canonicals, Open Graph images, breadcrumbs, and eligible structured data to maintained fields.

05

Define review and retirement

Assign an owner, review trigger, archive or redirect policy, and response when the page no longer meets its standard.

The rule is not 'never create many pages.' A business with hundreds of distinct products, public locations, support documents, or researched questions may need a large site. The rule is to preserve distinct purpose, truthful content, usable discovery, and ownership at that scale. A smaller set of complete pages is often more valuable than a giant collection that no visitor or employee can tell apart.

Prototype and test the architecture before migration

From draft map to production structure

01

Run task-based tree testing

Give representative users a text-only hierarchy and ask where they would find key information. Record wrong turns and labels that mean something only inside the company.

02

Prototype responsive navigation

Test menu order, labels, current-page indication, keyboard access, focus, and mobile behavior. W3C recommends semantic nav regions and structurally marked menus.

03

Crawl the staging site

Find orphan pages, excessive depth, broken links, redirect chains, duplicate routes, canonical conflicts, and URLs missing from or incorrectly included in the sitemap.

04

Build the redirect map

For every old canonical URL that changes, select the closest relevant new destination. Do not redirect unrelated removed pages to the home page.

05

Monitor after launch

Check server errors, indexing, sitemaps, canonicals, traffic by page group, and key user paths. Keep the old map and launch annotations for diagnosis.

How many levels should a small-business website have?

Use as many meaningful levels as people need to understand the choices, usually a compact home → hub → detail structure with supporting pages. There is no universal ranking rule for two, three, or four levels. Avoid both buried priorities and an unreadable flat menu.

What is an SEO silo?

The term often describes grouping related pages and linking them through a hub. Topic grouping can be useful, but strict silos that forbid helpful cross-links are unnecessary. Link across groups when the destination genuinely helps the reader.

Should URLs match the navigation hierarchy exactly?

Not always. URLs should be readable, stable, and manageable, while navigation reflects how people explore. They can align, but a deep folder path is not required to prove every parent-child relationship. Breadcrumbs and links also express hierarchy.

Do we need an XML sitemap if the site is well linked?

A small, comprehensively linked site may not require one for discovery, according to Google, but a clean automated sitemap is still useful for monitoring and growth. It does not replace internal links or guarantee crawling and indexing.

Should an architecture redesign change all existing URLs?

No. Preserve working URLs unless a change creates a meaningful long-term benefit. Every move carries redirect, recrawl, link, analytics, and implementation costs. When URLs must change, map and test redirects carefully.