Information Architecture: How to Structure Content and Navigation

TL;DR: What You Will Learn

• Information architecture is how you organise, label, and connect content so people can find what they need without thinking too hard.
• The core building blocks are your content inventory, grouping (how items cluster together), labelling (what you call things), and navigation (how people move around).
• You do not guess at IA. You test it with real users using methods like card sorting and tree testing, then iterate.
• Good IA lifts conversions, reduces support load, and builds trust, which is why marketers, product owners, and designers all need to care about it.

If you own a website or app, this is the groundwork that makes everything else easier. Let us build it properly.

What Information Architecture Really Means

When I first started structuring content for clients, I thought good design was mostly about how a page looked. I was wrong.

Information architecture is the practice of organising, structuring, and labelling content so that both people and search engines understand it and can find what they need fast. It is the invisible skeleton underneath every website you have ever navigated without frustration.

Think about the CPF portal. It holds an enormous amount of dense information: retirement schemes, housing withdrawals, MediSave, employer contributions. A fresh graduate checking their first contribution and a 55-year-old planning a withdrawal have completely different goals, yet both need to find their answer without drowning. That is IA doing its quiet work.

The Difference Between IA and UX/UI

People often lump IA in with design, but they are not the same thing.

IA decides how content is grouped and named. UX is the overall experience of using the site, and UI is the visual layer you actually tap and click. If you want a clearer picture of that last distinction, we cover the difference between UI and UX in detail elsewhere.

Here is how I explain it: IA is deciding that HDB resale content belongs in one section and rental in another. UI is choosing the button colour and font. UX is whether a first-time buyer actually completes the journey without giving up.

The Four Components

Every solid IA rests on four systems working together:

• Organisation: how you group content, for example by user type, topic, or task.
• Labelling: the words you use for categories and links, like “Withdrawals” versus “Cashing Out”.
• Navigation: the menus, breadcrumbs, and links that move people between groups.
• Search: the on-site search that catches users who do not want to browse.

A useful lens practitioners borrow from information architect Dan Klyn’s ontology, taxonomy, and choreography model, developed at The Understanding Group what your terms mean, how they are grouped, and how users move through them. On the HDB site, “flat” must mean the same thing everywhere (ontology), resale must sit logically beside BTO (taxonomy), and the path from eligibility to application must flow (choreography).

Why IA Affects SEO and Findability

Good IA is not just kind to users. It directly shapes how you rank.

Google’s own link best practices guidance confirms that internal links are a signal Google uses both to judge how pages relate and to discover which pages to crawl next. When related content is grouped and clearly labelled, crawlers can follow that structure to the right pages for the right queries.

Poor IA buries important pages several clicks deep, splits topics across confusing labels, and leaves both Google and your visitors lost. Fix the architecture, and findability improves for everyone.

Starting With a Content Inventory and Audit

Before you touch navigation menus or draw a single sitemap, you need to know exactly what you already have. I have watched too many teams jump straight into redesigning information architecture only to discover, weeks later, that a dozen important pages quietly vanished during migration.

Picture a Singapore SME, say a heating and air conditioning services company, moving its ageing website from a hand-coded setup to a modern CMS like WordPress. Without a full catalogue of every URL, half their service pages and old blog posts risk becoming broken links or 404 errors that tank their search rankings.

That is why I always start with the least glamorous step: the content inventory and audit.

Building The Inventory Spreadsheet

A content inventory answers a simple question: what exists? It is a spreadsheet listing every page, PDF, image asset, and blog post on your current site.

I keep the columns practical so the sheet stays usable rather than becoming a graveyard. For a typical SME migration, I track:

• URL and page title
• Content type (service page, blog, landing page, PDF)
• Content owner or department responsible
• Last updated date
• Key performance data (pageviews, entrances, conversions)

Pull the URL list from a crawler like Screaming Frog, then enrich it with performance figures from Google Analytics and Search Console. That data is what turns a dry list into a decision-making tool.

Auditing for Quality, Duplication, and Outdated Content

The inventory tells you what exists. A content audit tells you how good it is. These are two different jobs, and skipping the second one is where most projects go wrong.

Here you review each page for quality, accuracy, and relevance. In that HVAC example, you might find three near-identical pages about aircon servicing that compete with each other in search, plus a promotions page advertising a Chinese New Year deal from two years ago.

Search Console shows you which of those pages actually earn impressions and clicks. If a duplicate page gets no traffic and ranks below its sibling, that is a clear signal.

Deciding Keep, Revise, Merge, or Retire

Once every page has a quality verdict, run a ROT analysis to remove content that is Redundant, Outdated, or Trivial. Each page gets one of four labels: keep, revise, merge, or retire.

A high-performing service page stays. Those three duplicate aircon articles merge into one authoritative guide with 301 redirects. The expired promotion retires. Thin pages with potential get flagged to revise as part of a content marketing strategy that boosts your SEO.

This unglamorous groundwork is the foundation for every later IA decision, because you cannot structure content sensibly until you know precisely what you are structuring.

Grouping and Labelling Content So People Understand It

Once you have a pile of content, the real work of information architecture begins: sorting it into groups that make sense to the people using your site, not just to the team that built it.

I have seen sites where the navigation mirrors the internal org chart, with sections named after departments. Visitors do not think in departments, so they get lost.

Good grouping does two jobs at once. It reduces the mental effort a visitor spends hunting for things, and it signals to Google how your topics relate, which strengthens your topical relevance.

Choosing an Organisation Scheme

Before you name anything, decide how content is clustered. There are a few common schemes, and each fits a different situation.

• Topical: grouping by subject, like a polyclinic separating “Health screening”, “Vaccinations”, and “Chronic disease management”. This suits content-rich sites where people browse by interest.
• Task-based: grouping by what someone wants to do, like “Book an appointment” or “Collect medication”. A GovTech service site often leans this way because citizens arrive with a job to complete.
• Audience-based: grouping by who you are, such as “For patients” versus “For healthcare professionals”. Use it only when the audiences genuinely need different content.
• Alphabetical: useful for glossaries or long directories, but rarely for primary navigation.

Most real sites blend schemes. A polyclinic might use topical groups at the top level and task-based labels inside each one.

Writing Labels in the User's Vocabulary

A label only works if it matches the words people already use. On a GovTech or HealthHub-style service, I would label a section “Get your appointment reminders”, not “Notification preference management”.

The trick is to steal your vocabulary from real users. Pull the exact phrases from your site search logs, from support enquiries, and from what people type into Google to reach you.

If citizens keep searching for “MC” or “medical certificate” while your menu says “Attendance documentation”, the label has failed. Match their word, not your internal jargon.

Avoiding Ambiguous or Overlapping Categories

Overlapping categories are where users stall because they cannot tell which group holds what they need.

Imagine a polyclinic with both “Services” and “Treatments” in the menu. A visitor with a sore knee has no idea which to pick, so they guess, backtrack, and lose trust.

Test every label the same way. Take your ten most common support enquiries and your top search queries, then check whether each one points cleanly to a single, obvious group.

If a query could reasonably sit in two places, your categories are ambiguous. Merge them, rename them, or add a clear cross-link so nobody has to guess.

Designing Navigation Structures That Guide Users

Grouping your content is only half the job. If users cannot find those groups, your information architecture stays invisible.

Navigation is how you expose the structure you built. Most sites use several types working together, not one.

• Global navigation: The main menu that appears on every page, showing top-level sections.
• Local navigation: Menus within a section that reveal its subpages, like a sidebar inside a product category.
• Contextual navigation: In-text links that point to related content, such as a related products strip.
• Utility navigation: Small links for search, account, cart, or language, usually tucked into a corner.
• Breadcrumbs: A trail showing where a page sits in the hierarchy.
• Footer: A catch-all for policies, contact, and secondary links.

Before you design menus, pick an underlying structure. A hierarchical structure suits most sites, where categories branch into subcategories. A sequential structure suits step-by-step flows like a checkout or a course enrolment. A matrix structure suits sites where users want to browse the same content by different attributes, for example, filtering by brand, price, and category all at once.

Primary Versus Secondary Navigation

Your primary navigation carries the highest-priority paths, the ones most visitors need. Keep it short and lead with what people actually came for.

Take an electronics retailer like Challenger. As its business has shifted toward a much larger online catalogue and a leaner store footprint, its site’s mega menu has to expose thousands of SKUs without overwhelming a shopper. Primary navigation shows broad categories such as Laptops or Audio, while secondary navigation inside each category surfaces brands and use cases.

The mistake I see is cramming every subcategory into the top menu. Push the long tail into local navigation and on-page filters instead.

Flat Versus Deep Hierarchies and The Impact on Clicks

A flat hierarchy means fewer levels, and a deep hierarchy buries content, but what actually costs you, users, is less the raw click count than whether each step feels obvious to the person clicking it.

You’ll often hear “three clicks or fewer” cited as a rule. Treat it as a rough sanity check, not a target to engineer for. The research it’s loosely based on doesn’t hold up well: in a widely cited 2003 study, UX researcher Josh Porter analysed over 8,000 clicks and found no correlation between click count and task success or satisfaction, so long as each click gave people clear “scent” that they were headed the right way a finding Nielsen Norman Group has echoed since.

A five-click path where every choice is obvious beats a three-click path where users have to guess. Use “can I reach this in roughly three clicks?” as a flag to go investigate depth, then confirm with tree testing (more on that later) rather than a pass/fail rule. A marketplace balancing breadth against depth should still widen categories before it deepens them; depth is a real cost, just not one measured in click counts alone.

Flat structures also help SEO. Google’s SEO Starter Guide advises grouping topically similar pages so both users and crawlers can understand how content relates the same shallow-structure principle that makes it easier to strengthen internal linking for local SEO.

Breadcrumbs and Wayfinding Cues

Breadcrumbs answer a simple question: where am I, and how do I go back? They cut clicks and reinforce your hierarchy for both users and search engines.

On mobile, screen space forces hard choices. Collapse global navigation into a hamburger, prioritise search, and use bottom bars for the few actions people repeat most.

Test every navigation change against real tasks, not opinions, because a menu that looks tidy can still hide the page people need.

Using Card Sorting to Learn How Users Think

When I build an information architecture, the biggest risk is that I structure content the way my team thinks, not the way real users think. Card sorting is the method I lean on to close that gap.

The idea is simple. You give people a set of content items on cards, ask them to group the cards in ways that make sense to them, and often to name each group. What comes back is a window into their mental models.

Open Versus Closed Versus Hybrid Sorts

There are three flavours, and choosing the right one depends on what you already know.

• Open Sort: participants create and name their own groups. I use this early, when I have no fixed structure yet and want to discover how people naturally cluster things.
• Closed Sort: participants place cards into categories I define. This is for validating an existing or proposed structure, for example, when a bank already has “Save, Spend, Borrow” and wants to check whether products land where users expect.
• Hybrid Sort: participants use my predefined groups but can also add their own. This is my go-to when I have strong candidate categories but suspect I have missed one or two.

Recruiting and Running a Session

Recruit people who resemble your actual users, not colleagues. If I were structuring a SkillsFuture course directory, I would recruit mid-career professionals actively searching for courses, plus a few fresh NUS graduates, because they categorise “digital marketing” and “data analytics” differently.

For in-person sessions, I write each course or programme on a physical card and let participants shuffle them on a table. It is tactile and surfaces great “thinking out loud” moments.

For remote or larger studies, I use a tool like OptimalSort, which lets participants drag cards into groups online and aggregates the data automatically. This scales far better across Singapore’s dispersed audiences.

As a rule of thumb, around 15 to 20 participants per group is usually enough to see patterns stabilise for a closed sort, and open sorts often need a touch more to settle.

Interpreting The Results

Tools output a similarity matrix, showing the percentage of participants who paired any two cards together, and a dendrogram, a tree diagram that clusters cards by how often they were grouped.

I read the dendrogram for strong clusters and hunt for cards that participants split, which signals conflicting mental models.

When models conflict, I reconcile them by favouring the majority pattern, then supporting the minority through cross-links or search rather than forcing one rigid tree.

The payoff is that your grouping and labelling decisions from earlier now rest on evidence, not internal assumptions.

Testing and Validating Your Information Architecture

Here is the uncomfortable truth I keep coming back to: your information architecture is only as good as your users’ ability to find things in it, and no amount of internal debate settles that. You validate IA with data, not opinion.

I have watched teams argue for an hour over whether a page belongs under “Claims” or “Support”, when a 30-minute test with real customers would have answered it. So let me walk you through how I check an IA before and after launch, using a Singapore insurer redesigning its customer portal as the running example.

Tree Testing and Success and Directness Metrics

Tree testing checks findability without any visual design getting in the way. You give participants a bare text version of your navigation tree and ask them to find where they would go to complete a task.

For our insurer, one task might read: “Your car was scratched in a car park. Where would you go to start a claim?” Participants click through the tree until they land on the page they think holds the answer.

Two numbers matter most here. Success rate tells you how many people reached the correct destination, and directness tells you how many got there without backtracking.

If success is high but directness is low, your grouping is broadly right, but a label is misleading people down the wrong branch first. That points you straight back to your labelling choices, not a full restructure.

First-Click and Task-Based Usability Testing

Once you have a design, first-click testing tells you where people click first on a given screen. The first click is a strong predictor of task success; an original 2006-2009 analysis by Bob Bailey and Cari Wolfson found users were nearly twice as likely to complete a task when their first click was correct.

I like to pair this with moderated usability tests where you watch a real customer try to find, say, the fund fee schedule on a bank portal. Hearing them mutter “I thought fees would be under Investments, not Account” is worth more than any survey.

These sessions surface the same skills hiring managers probe in UX designer interview questions to expect, so they are good practice for your team too.

Analytics and Search Console Signals After Launch

Testing does not stop at go-live. Post-launch, your analytics and Google Search Console quietly flag IA problems every day.

Watch for pages with high entries but immediate exits, internal searches for terms your navigation should already surface, and Search Console queries where people land on the wrong page. If customers keep searching “claim status” on-site, that page is buried.

When a task fails, iterate: rename the label, regroup the section, or shorten the path, then re-run the same tree test so you can benchmark the new success and directness against your baseline. Improvement you can measure beats a redesign you merely feel good about.

How Information Architecture Supports SEO and Discoverability

Most marketers think of information architecture as a usability concern, something for designers to worry about. In my experience, it is one of the most underrated levers for organic visibility and lead generation.

When your content is structured well, Google crawls it faster, understands how your pages relate, and rewards you with rankings for terms you could not touch before.

URL Structure and Site Hierarchy Signals

Your URL structure is a map of your site that both people and search engines read. A logical path like /services/aircon-servicing/gas-top-up tells Google exactly where a page sits and what it supports.

Compare that to a jumble of pages like /page-2847 or /aircon-final-v3. Those give crawlers no signal about relationships or importance.

Breadcrumbs reinforce this hierarchy on the page itself. When a user lands on a deep service page from search, breadcrumbs show them (and Google) the parent categories, which strengthens the topical context of the whole branch.

Topic Clusters and Pillar Pages Built on IA

Here is where IA pays off directly in rankings. A pillar page covers a broad topic, and supporting cluster pages each tackle a specific subtopic, all linked back to the pillar.

Let me give you a Singapore example. Say you run a home cleaning and disinfection business, and you have twenty thin blog posts scattered across the site, each targeting a slightly different keyword like “office cleaning” or “post renovation cleaning”.

None of them rank because they compete with each other, and none has authority. So you consolidate.

You build one strong pillar page on commercial cleaning services, then restructure the thin posts into proper cluster pages under it: post renovation cleaning, office disinfection, upholstery cleaning. Each links up to the pillar, and the pillar links down to each.

This IA-first consolidation is often what moves a local service business from page three to page one for competitive terms. You stop diluting your authority and start concentrating it.

Internal Linking and Crawl Efficiency

Internal linking is how you tell Google which pages matter and how they connect. A flat site with no clear hierarchy forces crawlers to waste budget on low-value pages while your money pages sit undiscovered.

Good IA fixes this by design. Important pages sit close to the homepage, and every page has contextual links pointing to related content.

This matters more now that AI-driven search is reshaping discovery. Well-structured content with clear headings and logical relationships is exactly what gets surfaced in AI answers, so it is worth learning how to optimise content for Google AI Overviews alongside your IA work.

The takeaway is simple. Clean IA is not polish; it is the foundation your entire organic strategy stands on.

Common Information Architecture Mistakes and How to Avoid Them

Over the years, I have audited plenty of sites and intranets where the information architecture looked tidy on a slide but fell apart the moment a real person tried to find something. The same handful of mistakes come up again and again.

Building The Structure Around Your Org Chart

The biggest one is organising content by how your company is structured internally rather than how users think.

I once reviewed a Singapore public sector intranet that had been reorganised strictly by department: HR, Finance, Corporate Comms, IT. It looked logical to management, but staff did not think in departments. They thought in tasks like “claim my medical expenses” or “book leave”, and those tasks were scattered across three different sections.

The fix is to lead with user language, and the card sorting exercise from earlier sections is exactly how you surface it. When you let real staff group and label content, task-based clusters emerge that no org chart would ever predict.

Menus Too Deep, Labels Too Clever

Two related pitfalls hurt findability quietly. Overly deep menus force people to click four or five levels down before they reach anything useful, and jargon labels make them guess what a section even contains.

If your navigation uses internal terms like “Stakeholder Enablement” or “Resource Hub”, a tree test will show people wandering or giving up. Rename those items in plain words, and flatten the hierarchy where you can, not to hit a specific click count, but because every extra unclear level is another place someone can guess wrong and bail.

Orphan Pages and Skipped Research

Orphan pages are pages that exist but are not linked from any menu, so only a direct URL or search reveals them. Your content audit is where you catch these, and every orphan either needs a home in the structure or needs to be retired.

The biggest mistake, though, is skipping user research entirely and designing the IA from assumptions. Card sorting and tree testing exist precisely so you are not guessing.

Letting Structure Decay Without Governance

Here is the one that catches fast-growing Singapore SMEs most. They launch a clean site, then over eighteen months, every team bolts on pages, and nobody owns the whole picture, so the structure quietly rots.

Governance prevents this. Assign a content owner for each major section, and schedule a light IA review, perhaps quarterly, as the site grows.

Run this quick checklist mindset on your own site:

• Does every top-level label match a word your users actually use?
• Can key tasks be reached quickly, with a clear, obvious choice at every step?
• Are there any orphan pages hiding from your navigation?
• Who owns each section, and when did you last review it?

If you cannot answer the last question, that is where I would start.

Conclusion

Remember the frustration we started with: a visitor lands on your site, cannot find what they came for, and leaves within seconds. That abandonment is not just a bad experience; it quietly erodes your conversions, your credibility, and your search visibility all at once.

Good information architecture is what stands between that lost visitor and a customer who trusts you enough to enquire, buy, or come back.

Here is the whole workflow in one tight arc. Start by taking stock: inventory every page and audit what is working and what is not. Then group related content and label it in words your users actually use, not internal jargon. From there, design navigation that mirrors those groupings, and finally validate it with card sorting and tree testing, both before you build and after you launch.

The part many teams miss is that IA is not a one-off project. It is ongoing governance. For a growing Singapore business, whether you are a fast-scaling e-commerce brand or an SME adding new service lines every quarter, your content keeps expanding. Without regular review, even a beautifully structured site drifts back into chaos.

So do not wait for a full redesign. This week, open a simple spreadsheet and start a content inventory of your top pages. That single habit will surface more problems, and more quick wins, than you expect.

The real leap comes when structuring content and testing it with users becomes second nature rather than guesswork, and that shift usually happens faster with structured practice than trial and error alone. If you want to build that muscle properly, Equinet Academy’s WSQ Advanced UX UI Design Course is a good next step. Over two hands-on days, you’ll learn to turn user feedback into validated design decisions, build user flows and wireframes, and run real usability tests on interactive prototypes the exact skills behind the IA workflow above. Singapore-based professionals may also qualify for SkillsFuture funding that brings the course down significantly from its S$990 fee.