Information Architecture: How to Structure a Website or Digital Product
How to restructure a complex website or product: content audit, card sorting, tree testing, sitemap, URLs and breadcrumbs, with a re-org worksheet.
How do you structure a website or digital product?
Information architecture (IA) is how a website's or product's content and features are organised, labelled and connected so people can find things and understand where they are. To structure a complex site, you audit what exists, learn how users group it (card sorting), design a hierarchy and labels, test that hierarchy without visual design (tree testing), then document it as a sitemap with a URL, breadcrumb and redirect plan. Menus and page layouts come after that, not before.
This applies whether you run a marketing site that has grown to hundreds of pages, an ecommerce catalogue, a content library or a SaaS product with dozens of settings screens.
Key takeaway
Key takeaways: audit before you sort; let users' groupings shape the structure; test labels in a tree test before anyone designs a menu; Google reads your structure from links, not URL folders; and a re-org is not finished until the redirect map is done.
IA and Navigation Are Different Things
Nielsen Norman Group describes IA as "the information backbone of the site" and navigation as the "elements in the UI that allow users to reach specific information". A menu redesign won't fix a confusing structure underneath. The website navigation and UX guide covers menus, breadcrumb design and mobile navigation; this guide covers the structure they depend on.
| Component | What it defines | Example |
|---|---|---|
| Organisation systems | How content is grouped | By topic, task, audience or product type |
| Labelling systems | What things are called | "Plans" vs "Pricing", "Help" vs "Support" |
| Navigation systems | How users move through content | Global menu, local navigation, breadcrumbs |
| Search systems | How users find by query | Site search, filters, suggestions |
| Taxonomies and metadata | How content is classified | Product categories, attributes, tags |
The IA Re-org Worksheet
Use this as the plan for a restructure. Each phase has a method, a concrete output, and a decision it settles. If you skip a phase, the decision it settles gets made in a meeting instead, usually by whoever has the most senior job title.
| Phase | Method | Output | Decision it settles |
|---|---|---|---|
| 1. Inventory and audit | Crawl plus spreadsheet review | Content audit sheet: URL, owner, traffic, keep / merge / cut | What is actually in scope |
| 2. Discover groupings | Open card sort | Candidate groups and users' own labels | Which top-level categories make sense to users |
| 3. Validate labels | Closed card sort or tree test | Success and directness by task | Whether people can find key items in the proposed structure |
| 4. Document the hierarchy | Sitemap | Page hierarchy with templates and owners | What lives where, and how deep it goes |
| 5. Plan URLs and breadcrumbs | URL pattern and breadcrumb rules | URL list and breadcrumb trail per template | Which URLs change and how location is shown |
| 6. Map redirects | Old-to-new mapping | Redirect map, one row per changed URL | That no existing link or bookmark breaks |
Step 1: Inventory and Audit the Content
An inventory answers "what is there?"; an audit answers "is it any good, and should it stay?". For an existing site, crawl it to get every indexable URL, then add columns for page type, owner, last updated, traffic over a sensible period (say 12 months, from your analytics) and whether internal links point to it.
Then make a call on each row. A common working method is to flag content as redundant, outdated or trivial (often shortened to ROT) and assign an action: keep, merge into another page, rewrite, or remove with a redirect. For a product, the equivalent is a feature and screen inventory with usage data. For a new site, derive the inventory from user tasks and business requirements instead.
Do not skip this step. Sorting 400 cards when 150 of them should be cut produces a structure designed around content nobody needs.
Step 2: Learn How Users Group Content
Structure should reflect users' mental models, not your org chart. Card sorting is a discovery method used to generate ideas for an IA. Nielsen Norman Group describes three types:
- Sample size (NN/g): at least 15 participants for a qualitative sort, at least 30 to 50 for quantitative results you want to generalise
- Use 30 to 60 representative cards, not the whole site; pick items from across the audit
- Write card labels in plain language so the card's own wording doesn't give away the answer
- Recruit people who match your real audience, not colleagues
- Look for agreement patterns and the labels users invent, not a single "correct" answer
| Type | How it works | Use it when |
|---|---|---|
| Open | Users create their own groups with no predefined categories, and name them | You are building or rethinking the top level and want users' language |
| Closed | Users sort cards into categories you provide | You have a proposed set of categories and want to see if items fit |
| Hybrid | Some predefined categories, plus freedom to create new ones | Rarely. NN/g generally recommends against it, because predefined groups steer users; it may suit a team confident about one or two categories but unsure of the rest |
Step 3: Define Categories and Labels
Group content into categories that are distinct from each other, and name them in users' language. Avoid internal jargon and clever labels that hide meaning. Where items could belong in several places, choose one primary location and use cross-links from the others.
Balance breadth and depth. Too many top-level items overwhelm; too many levels bury content. Check that the most common tasks need the fewest clicks.
Step 4: Validate With Tree Testing
A tree test evaluates a hierarchical category structure. Participants see a text-only version of the menu and are asked to find specific items ("Where would you go to change your billing address?"). Because there is no visual design, the results show whether the structure and labels work on their own.
Nielsen Norman Group's tree testing guide describes the main measures. A qualitative tree test needs only a few participants; a quantitative one needs a larger sample. NN/g's guide does not give a fixed number, so size the study to the decision you need to make.
| Metric | What it tells you | What to do if it is weak |
|---|---|---|
| Success rate | Share of participants who found the correct location for a task | Rename or move the item; check whether two categories overlap |
| Directness | Whether people went straight there or tried other branches first; a proxy for confidence | Find the wrong branch most people tried; that is the label competing with yours |
| Time | How long it took to choose | Long times with high success suggest labels people can decode, but only slowly |
| First click and paths | Which top-level category people chose first | A wrong first click usually means the top level needs rework |
Pro tip
Test the existing tree and the proposed tree with the same tasks. The comparison is more persuasive in a stakeholder meeting than either result on its own. For how these measures sit alongside task success and satisfaction scores, see UX metrics beyond conversion rate.
Navigation Models, Briefly
Once the tree passes testing, choose how the interface exposes it. Most sites combine several of these.
- Mega menu guidance: NN/g on mega menus
- Ecommerce specifics: ecommerce navigation design and faceted navigation and SEO
- Menu, search and mobile navigation UX: the navigation guide
| Model | Fits | Watch out for |
|---|---|---|
| Global menu with local navigation | Most marketing and content sites | Local menus that differ from section to section |
| Mega menu | Large sites with many second-level pages | NN/g advises a hover delay of about 0.5 seconds, no complex widgets inside, and each option shown once |
| Hub pages | Services, guides or product families where an overview page links to each child page | Hubs that are just link lists with no help choosing |
| Faceted navigation | Ecommerce and large catalogues filtered by attributes | Crawlable filter combinations multiplying URLs |
| Tab bar or sidebar | Mobile apps and SaaS products | Too many top-level destinations |
Step 5: The Sitemap Deliverable
A UX sitemap documents the hierarchy everyone has agreed. Keep it practical: each node should show the page name, its template (for example service page, article, category), its owner, and whether it is new, kept, merged or removed. Pair it with user flows for the main tasks; information architecture vs user flow explains how the two fit together, and user flow design covers mapping flows.
Hypothetical example: a services business with a sprawling site might settle on the structure below. It is illustrative, not a template to copy.
Home
├── Services [hub]
│ ├── Bookkeeping [service]
│ ├── Payroll [service]
│ └── Tax returns [service]
├── Who we help [hub]
│ ├── Sole traders [audience]
│ └── Small companies [audience]
├── Pricing [page]
├── Guides [article index]
│ └── (articles, tagged by service)
├── About [page]
└── Contact [page]Step 6: URLs, Breadcrumbs and Redirects
Google says it "generally doesn't look at the structure of URLs to work out the structure of a site"; it analyses the relationship between pages "based on their linkages", and recommends ordinary <a href> links rather than JavaScript events on other elements. So the structure you designed has to show up in menus, hub pages and cross-links. A tidy folder path is not enough on its own.
URLs still matter for people. Google's URL structure guidance recommends readable words rather than long ID numbers, hyphens rather than underscores, words in your audience's language, and trimming parameters that don't change the content. It also notes URLs are case sensitive. A practical rule: make URLs short and readable, and don't change existing ones just to mirror a new menu.
Breadcrumbs show where a page sits. Google describes a breadcrumb trail as indicating "the page's position in the site hierarchy" and uses breadcrumb markup to categorise information in results. Its documentation lists the feature as available on desktop, and since 23 January 2025 Google no longer shows breadcrumbs in the visible URL on mobile results, only the domain. Breadcrumbs on the page still help users on every device, so keep them for anything more than two levels deep.
Finally, every URL that changes needs a permanent redirect from the old address to its closest new equivalent, listed in a redirect map before launch. The URL migration guide covers mapping and testing redirects, and SEO-friendly website development covers the build side.
Search as Part of IA
Search complements browsing, especially for large catalogues or content libraries, but it should not compensate for a structure people cannot browse. Taxonomies and metadata power filters and relevant results. Your site search logs are also research: frequent queries for things that are already in the menu point to labels people don't recognise. See in-app search and ecommerce category page design for applied examples.
IA for Different Products
| Product | IA priority |
|---|---|
| Marketing website | Clear service and audience paths, hub pages, short routes to contact or demo |
| Large content or corporate site | Content audit, ownership per section, governance so the menu stops growing |
| Ecommerce store | Category taxonomy, product attributes for filters, search |
| SaaS product | Task-based navigation, settings organisation, permissions |
| Mobile app | Few top-level destinations, shallow hierarchy |
| Content library | Topics, tags, search and related content |
Common IA Mistakes
- Mirroring the company's internal structure or departments
- Card sorting the whole site before cutting redundant and outdated content
- Vague or clever labels users don't recognise
- Overlapping categories where items could live anywhere
- Very deep hierarchies that bury common content
- Designing menus before testing the structure
- Changing every URL to match the new menu, then forgetting redirects
- Relying on search to compensate for poor structure
Risks and Limits
A full re-org is not always the right answer. If analytics and support questions show problems in one section, fix that section and tree-test it rather than restructuring everything. Small sites with a dozen pages rarely need formal card sorting; a quick review of labels against user tasks may be enough. And research only helps if participants resemble your real audience: a card sort with colleagues mainly tells you how your colleagues think.
Structural change also carries search risk. Moving or merging many pages at once can temporarily disrupt how pages are found, which is why the redirect map and internal linking plan belong inside the IA work, not after it.
Conclusion
Good information architecture is invisible: users find what they need and know where they are. Audit content first, learn users' mental models with card sorting, label in their language, validate with tree testing, then document the sitemap, URL and breadcrumb rules and redirect map before visual design. It is the foundation of the UX design process.
Next steps: copy the worksheet above into a spreadsheet, run the inventory, and agree which sections are in scope before recruiting for a card sort.
Planning a redesign? Map the structure first.
ZSpace Labs is a remote-first technology studio that works with clients globally on UX and information architecture and website development. If useful, we can talk through how to audit, test and restructure your site before design starts.
Sources
Research methods: NN/g: The Difference Between Information Architecture and Navigation (Cardello, 2014); NN/g: Card Sorting (Tankala and Sherwin, 2024); NN/g: Card Sorting vs Tree Testing (Tankala, 2024); NN/g: Tree Testing (Laubheimer, 2023, reviewed 2026); NN/g: Mega Menus Work Well (Nielsen and Li, 2017).
Search: Google Search Central: help Google understand your site structure; Google Search Central: URL structure best practices; Google Search Central: breadcrumb structured data; Google Search Central blog: simplifying the visible URL on mobile (23 January 2025).
Accessed 15 October 2026. Search engine behaviour changes; check Google's current documentation.
Common questions.
The practice of organising, structuring and labelling content and features so users can find what they need and understand where they are.
Information architecture is the underlying structure and labelling. Navigation is the interface, such as menus, links, search and breadcrumbs, that exposes that structure to users.
A research method where participants group content items into categories. Open card sorts let them name categories; closed card sorts use predefined ones; hybrid sorts mix the two.
A method where participants try to find items in a text-only version of a proposed structure, showing whether labels and groupings work without visual design influencing results.
A diagram of a product's pages or screens and how they're organised hierarchically. It's a key information architecture deliverable, and it is different from the XML sitemap file submitted to search engines.
A controlled system for classifying content, such as product categories and attributes in ecommerce, that supports navigation, filtering and search.
Signals include users searching for things that should be easy to browse, high exit rates from navigation pages, support questions about where to find things, and failed tree tests.
It can, mainly through internal linking. Google says it generally doesn't look at the structure of URLs to work out the structure of a site; it analyses how pages link to each other. Clear hierarchy, crawlable links and descriptive labels help both users and search engines. Readable URLs are still good practice, but a folder path alone does not tell Google how your site is organised.
Nielsen Norman Group recommends at least 15 participants for a qualitative card sort and at least 30 to 50 for quantitative results you want to generalise.
Only when there is a clear reason. Every changed URL needs a permanent redirect from the old address, and the new structure should still be expressed through internal links and breadcrumbs, which is what search engines and users actually follow.