Jump to

Skill · Information architecture

Information architecture.

Design the structure that holds the content.

Design the structure that holds the content, across marketing sites, product surfaces, knowledge bases, and ecommerce. Six layers build on each other: mental models, sitemap, URL structure, navigation, taxonomy and metadata, and labeling. A well-designed IA makes every downstream decision easier; a poor one forces each one to fight the structure.

The structure follows how the audience thinks, not the org chart, and the cheap validation methods (card sorts, tree tests, first-click tests) catch the big problems before launch.

Audience: designers, content strategists, and product teams planning a new site or section, restructuring an existing one, or auditing an IA for problems.

View the skill on GitHubBrowse the full catalog

The framework

Six layers, each built on the one below.

IA has six layers. Get the mental models right and the rest follows; get them wrong and every layer above fights the structure.

  1. 01Mental models: how the audience thinks about the domain (what they group, the words they use, their frame of reference), surfaced through card sorts, tree tests, and first-click tests.
  2. 02Sitemap: the map of all pages and how they relate, following a dominant pattern (hub-and-spoke, tree, faceted, or flat), with page types and cross-references documented.
  3. 03URL structure: user-facing and indexed, so reflect the hierarchy, stay lowercase and predictable and stable, with one pattern per content type.
  4. 04Navigation: primary (5 to 7 recognizable items reflecting audience priorities), secondary, utility, breadcrumbs, and footer, with the most important pages one click from home.
  5. 05Taxonomy and metadata: a small controlled category list (5 to 15), a larger governed tag set for cross-cutting discovery, and consistent metadata fields.
  6. 06Labeling: audience language over internal jargon, consistent across the site, and validated with a closed card sort or tree test.

The sitemap layer

Four sitemap patterns, usually blended.

Most sites blend these, so pick the dominant pattern and document the exceptions rather than forcing a single shape.

  1. 01

    Hub-and-spoke

    Cornerstone content plus supporting content linking back to it. Common for content marketing.

  2. 02

    Tree

    A strict hierarchy where every page has exactly one parent. Common for product documentation.

  3. 03

    Faceted

    Content living in many overlapping categories at once. Common for ecommerce.

  4. 04

    Flat

    Everything reachable directly from the home page. Common for small sites.

How to structure it

By the audience's model, not the org chart.

Design the IA around how the audience thinks, not the org chart. 'Engineering' and 'Marketing' sections make sense to the company and confuse everyone else, so the structure follows the audience's mental models, surfaced through card sorts, tree tests, and first-click tests, which are cheap and catch large problems early.

Hold categories at 5 to 15 and govern the tags, because categories proliferate as every team adds one for its thing, and ungoverned tags sprawl into a junk drawer. Use audience language for labels rather than internal jargon: 'Solutions' and 'Resources' are the vague catch-alls everything eventually ends up in.

Pick one URL pattern per content type and keep the most important pages one click from home. Search supports the IA, it does not replace it, so 'just use search' is not a structure. And IA evolves with the content, so plan for revision rather than treating it as a one-time deliverable.

Reference files

Two references that go alongside the SKILL.md.

  • references/ia-document-template.md

    The template for the IA deliverable.

  • references/url-pattern-library.md

    URL pattern conventions for common content types.

Browse all reference files on GitHub

Bridges to other skills

What feeds the structure and what builds on it.

IA needs the audience and content defined first, and hands its structure to the skills that compose and fill the pages.

  • Composition on top

    vertical-site-conventions

    Models the nav tree and URL structure that vertical composition builds on. That skill decides which paths are prominent on the page; IA owns the structural model behind them.

  • What content exists

    content-strategy

    Defines the content the structure has to hold. If the content scope is unclear, run content strategy first, then structure what it produces.

  • Who it is for

    brand-discovery

    Surfaces the audience whose mental models the IA follows. If the audience is unclear, discovery comes before any structure.

  • Search planning

    seo-keyword

    SEO-driven content planning is a different lens on what to build. IA organizes what exists or is planned; keyword research decides what to add for search.

Open source under MIT

Read the SKILL.md on GitHub.

The skill source lives in the rampstackco/claude-skills repository alongside dozens of other skills covering the full lifecycle of brand and product work. This page is a structured overview; the SKILL.md is the source. MIT licensed.

View SKILL.mdBrowse the full catalog

Frequently asked questions.

What are the six layers of IA?
Mental models (how the audience thinks about the domain), sitemap (the map of pages and their relationships), URL structure (the user-facing, indexed paths), navigation (primary, secondary, utility, breadcrumbs, footer), taxonomy and metadata (categories, tags, and fields), and labeling (what you call things). Each builds on the one below, so getting the mental models right makes the rest follow, while getting them wrong forces every downstream decision to fight the structure.
How many categories should I have?
A small, controlled list, typically 5 to 15, mutually exclusive where possible, and used for structural navigation. Hold the line, because categories proliferate as every team adds one for its thing until the set becomes unscannable. Tags are the separate, larger, multi-assignment set used for cross-cutting connections and long-tail discovery, and they need governance too, since ungoverned tags sprawl into a junk drawer of thousands that no one maintains.
How should I structure URLs?
Reflect the content hierarchy, keep them lowercase and hyphen-separated, predictable (one pattern per content type), stable (no changes without redirects), short where possible, and descriptive. Inconsistent URL patterns, where some posts sit at /blog/[slug], some at /[slug], and some at /articles/[slug], confuse users, crawlers, and analytics alike, so pick one pattern per content type and stick to it.
How do I validate the structure?
With cheap tests that surface large problems early. Open or closed card sorts reveal how the audience groups content; tree tests check whether a proposed structure lets users find specific items; first-click tests show whether the labels point people the right way, since a wrong first click means the labels and structure are wrong. Skipping validation is a failure pattern, and 'just use search' is not a structure, because search supports the IA rather than replacing it.
Should the navigation match the org chart?
No. IA designed by org chart, with 'Engineering' and 'Marketing' as top-level sections, makes sense to the company and not to the audience. Primary navigation reflects what the audience cares about, in 5 to 7 recognizable items of two or three words each, with the most important pages reachable one click from home. Use audience language for labels rather than internal jargon, and validate the wording with a closed card sort or tree test before shipping it.