What IA is: organization × labeling × navigation × search — four systems, one mission
Information Architecture (IA) is 'the design of the four systems of organization, labeling, navigation, and search to help people find and manage information more smoothly' — from Information Architecture by Rosenfeld, Morville, and Arango (known in the industry as the 'polar bear book,' first published in 1998 and a foundational work).
In one line: IA decides 'how content is grouped, what each piece is called, how people move around, and how they search' —
- Organization system: how content is categorized;
- Labeling system: what each piece of content is called;
- Navigation system: how users move through the structure;
- Search system: how users query directly for what they want.
IA is usually an invisible underlying skeleton — when it's done well you don't notice it; when it's done badly, you constantly 'can't find things and don't know where to click.' Note: IA is not what the interface looks like (that's visual/interaction design) — it is structure and meaning.
Hold onto this yardstick: even the most useful feature, buried four menu levels deep with a mislabeled name, might as well not exist to users.
The most practical standard for judging IA isn't how it looks, but: can users find what they need within a reasonable number of steps — measured with tree testing and task success rates, not gut feeling.
How to group things: exact vs ambiguous schemes × hierarchy / hypertext / faceted structures
First, organization schemes (what logic you group by). The polar bear book splits them into two kinds:
- Exact schemes: grouping by objective, indisputable attributes — alphabetical, chronological, geographical. Strengths: unambiguous and easy to maintain, suited to known-item finding where 'I know what I'm looking for' (looking up a contact, browsing bills by date).
- Ambiguous schemes: grouping by subjective meaning — topic, task, audience, metaphor. Harder to build (categorization is debatable), but more valuable for exploratory browsing where 'I can't quite say what I want; I just want to look around and learn' — most discovery and learning happens here.
In practice most sites need both: an e-commerce site lets you browse by category (ambiguous) and also sort by price or alphabetically (exact).
Next, organization structures (the shape of relationships between content). Three main ones:
(1) Hierarchy / tree — the most common: top-down, broad categories → subcategories. The design trade-off is breadth vs depth: too many options per level gets wide and messy, while too many levels causes click fatigue — usually better to be moderately wide than too deep.
(2) Hypertext — lateral connections via associative links: flexible but easy to get lost in, usually a supplement to a hierarchy rather than the main structure.
(3) Faceted / database — tag content with multiple independent attributes so users can filter along several dimensions (like an e-commerce filter combining 'brand + price + size + color'), especially suited to large, heterogeneous collections.
Which structure to pick depends on the volume of content and how users look for things.
Label in the user's words, know your three navigation types, validate with card sorting — and dodge two big myths
A label is the name you give each category / link / section — it 'speaks for' the whole chunk of content. Two iron rules:
(1) Consistency — one meaning, one term, site-wide; don't say 'My Orders' here and 'Purchase History' there;
(2) Speak the user's language — use words your target users know, not internal jargon or department names. This is exactly usability heuristic #2, 'match between system and the real world,' landing on IA.
The cost of bad labels is very real: even when the content is right there, the wrong name makes it invisible and unclickable to users. How to choose labels? Look at users' search terms, run card sorting, and follow established industry conventions — don't just make them up.
Navigation answers two user questions: 'Where am I, and where can I go?' Three main types:
- Global: site-wide, on every page (top main menu / footer), providing the overall frame;
- Local: sub-navigation inside a section (sidebar), for moving within one big category;
- Contextual: related links embedded in the content ('Related articles,' 'Frequently bought together'), for jumping along lines of meaning.
Plus supplemental navigation: search, sitemap, A-Z index, and breadcrumbs (showing your current location). The principle: let users always know where they are, where they came from, and where they can go; don't just pile up links — layer them and give them emphasis (echoing visual hierarchy).
Don't decide structure and labels by fiat — let users' mental models decide. Card sorting: write content items on cards and ask users to group them.
- Open: users group freely and name the groups themselves — revealing how they categorize and what words they use; best for building a structure from scratch;
- Closed: you provide the categories and users sort cards into them — best for validating or optimizing an existing structure.
Its value: categories and labels that fit user expectations instead of mirroring the org chart — 'organized by department' is often exactly why users can't find things. Then validate with tree testing: give users your structure and see whether they can find specified items in it.
自测 · 学完检查一下
想真正动手做题、记进度、攒连胜？到互动课里练。
The polar bear book (*Information Architecture*) defines IA as the design of four systems that help people find and manage information. Which four?
答案：Organization, labeling, navigation, and search
The polar bear book (Rosenfeld, Morville, and Arango, *Information Architecture*, first published 1998 and a foundational work) defines IA as 'the design of the four systems of organization, labeling, navigation, and search to help people find and manage information more smoothly' — that is, how content is grouped (organization), what each piece is called (labeling), how people move around (navigation), and how they search (search). IA is the invisible underlying skeleton, not what the interface looks like (that's visual/interaction design). The other options are visual styling, a development process, and page types — none of them the four systems. (Source: Rosenfeld/Morville/Arango, Information Architecture, 4th ed., O'Reilly; the Information Architecture Institute's definition of IA)
True or False: Information architecture is basically drawing the nav bar and producing a boxes-and-arrows sitemap — once the navigation is drawn, the IA work is done.
答案：False
False. This is the first big myth about IA: navigation is only the visible tip of IA; the body of IA is the underlying organization and labeling logic (how content is grouped and what it's called), derived from content, users, and context, and often invisible — done well you don't notice it, done badly you constantly 'can't find things and don't know where to click.' Equating IA with drawing a nav bar or sitemap skips the two core systems of organization and labeling. (Source: Rosenfeld/Morville/Arango, Information Architecture, 4th ed.; NN/g — IA is not navigation)
The polar bear book splits organization schemes into exact and ambiguous. Which statement about what each is suited for is correct?
答案：Exact schemes (alphabetical/chronological/geographical) suit known-item finding where 'I know what I'm looking for'; ambiguous schemes (topic/task/audience) are more valuable for exploratory browsing where 'I can't quite say what I want'
Exact schemes group by objective, indisputable attributes (alphabetical, chronological, geographical): unambiguous, easy to maintain, suited to known-item finding (looking up a contact, browsing bills by date). Ambiguous schemes group by subjective meaning (topic, task, audience, metaphor): harder to build but more valuable for exploratory browsing — most discovery and learning happens there. They are not either/or: in practice most sites need both, like an e-commerce site that lets you browse by category (ambiguous) and sort by price or alphabetically (exact). (Source: Rosenfeld/Morville/Arango, Information Architecture, 4th ed., Ch.6 Organization Systems)
An e-commerce site with hundreds of thousands of products wants users to filter freely by 'brand + price + size + color.' Which organization structure fits this need best?
答案：A faceted / database structure — tag content with multiple independent attributes so users can filter along several dimensions; especially suited to large, heterogeneous collections
A faceted / database structure tags content with multiple independent attributes so users can combine filters across dimensions (exactly 'brand + price + size + color'), and is especially suited to large, heterogeneous collections. Hypertext connects laterally through links — flexible but easy to get lost in, usually a supplement to hierarchy rather than the main structure; a subcategory for every attribute combination would make the tree so deep it causes click fatigue; and dumping hundreds of thousands of products on one page abandons organization altogether. Pick the structure based on content volume and how users look for things. (Source: Rosenfeld/Morville/Arango, Information Architecture, 4th ed. — hierarchy/hypertext/faceted classification; NN/g — Faceted Navigation)
True or False: When designing a hierarchy, the deeper the better — keep subdividing content level after level; the deeper users click, the clearer the structure must be.
答案：False
False. Hierarchies involve a breadth vs depth trade-off: too many options per level gets wide and messy, but too many levels adds click fatigue — usually better to be moderately wide than too deep. 'Clicking deep' doesn't mean 'clearly organized'; it just means users must click many times to reach content. Add Morville's findability lens: a feature buried four menu levels deep might as well not exist to users. The standard is whether users can find their target within a reasonable number of steps, not how finely the levels subdivide. (Source: Rosenfeld/Morville/Arango, Information Architecture, 4th ed. — breadth vs depth trade-off; Morville — Ambient Findability)
An article page shows a row of 'Related articles' at the bottom; a product page shows 'Frequently bought together.' Under the polar bear book's navigation taxonomy, which type is this?
答案：Contextual navigation — related links embedded in the content, jumping along lines of meaning
'Related articles' and 'Frequently bought together' are classic contextual navigation: related links embedded in the content that let users jump along lines of meaning. The three main types each have a job: global navigation is site-wide and on every page (top main menu / footer), providing the overall frame; local navigation is the sub-navigation inside a section (sidebar) for moving within one big category; contextual navigation is embedded in the content itself. Breadcrumbs belong to supplemental navigation (along with search, sitemap, A-Z index) and show your current location. (Source: Rosenfeld/Morville/Arango, Information Architecture, 4th ed., Ch.8 Navigation Systems)
True or False: It's fine to name categories and links with internal company jargon or department names — the content is there anyway, and users will find it after a few extra clicks.
答案：False
False. The labeling system has two iron rules: (1) consistency — one meaning, one term, site-wide; don't say 'My Orders' here and 'Purchase History' there; (2) speak the user's language — use words your target users know, not internal jargon or department names, which is exactly usability heuristic #2 'match between system and the real world' landing on IA. The cost of bad labels is very real: even when the content is right there, the wrong name makes it invisible and unclickable to users. Choose labels from users' search terms, card sorting, and established industry conventions — not off the top of your head. (Source: Rosenfeld/Morville/Arango, Information Architecture, 4th ed. — labeling systems; Nielsen — Heuristic 2: Match between system and the real world)
A team is building a brand-new site's category structure from scratch and wants to learn how target users would group the content and what words they'd use to name the groups. Which research method fits best as the first step?
答案：Open card sorting — users group cards freely and name the groups themselves, revealing how they categorize and what words they use
For building a structure from scratch, choose open card sorting: write content items on cards and let users group them freely and name the groups themselves, revealing their mental models and vocabulary. Closed card sorting — where you provide the categories and users sort into them — suits validating or optimizing an existing structure. Copying the org chart is the classic anti-pattern: 'organized by department' is often exactly why users can't find things. And tree testing doesn't generate structures either: it's a validation method where you give users an existing structure and see whether they can find specified items — you need a structure before you can test it. (Source: Nielsen Norman Group — Card Sorting & Tree Testing; the polar bear book — deriving IA via card sorting)
True or False: By Peter Morville's notion of findability, even the most useful feature — buried four menu levels deep with a mislabeled name — might as well not exist to users; content users can't find is worthless no matter how good it is.
答案：True
True. Morville proposed findability: if users can't find content, it doesn't matter how good it is — one of IA's core missions is making things 'findable.' His UX honeycomb lists 7 facets of good experience: useful, usable, desirable, findable, accessible, credible, and valuable — and IA mainly carries 'findable + usable.' The most practical standard for judging IA is whether users can find what they need within a reasonable number of steps (measured with tree testing / task success rates). (Source: Morville — Ambient Findability, O'Reilly, 2005; Morville — User Experience Honeycomb)
Someone claims: 'The more categories and options you list in the navigation, the easier it is for users to find things.' What's wrong with this claim?
答案：Too many top-level options actually slows decisions down (Hick's Law: the more choices, the longer the decision time) and dilutes emphasis; good IA is just-right grouping + clear labels + strong search
'More categories means easier to find' is the second big IA myth: too many top-level options actually slows decisions down — Hick's Law says the more choices, the longer the decision time — and dilutes emphasis, while overly deep hierarchies add click fatigue. Good IA isn't laying everything out flat; it's just-right grouping + clear labels + strong search. The test is always 'can users find it,' not 'did I list everything.' (Source: Rosenfeld/Morville/Arango, Information Architecture, 4th ed.; Hick W.E. (1952), On the rate of gain of information)