Information Architecture

Information Architecture

Author
Louis Rosenfold, Peter Morville and Jorge Arango
Year
2015
image

Review

What is it about? Was it worth reading?

This book blew the doors off a world I knew nothing about. It helped me become a better product manager. I gained a stronger understanding of taxonomy, navigation and search. Many of the concepts here will be familiar to engineers, this book brought be closer to that discipline and enabled me to have conversations at a much deeper level. I believe that Product Managers should read around their discipline, a lot of the magic happens at the edges and this is a good place to start.

You Might Also Like…

image

Key Takeaways

The 20% that gave me 80% of the value.

  1. Information Architecture is the art and science of shaping products to support usability, find-ability and understanding. Increasing amounts of content, metadata and complexity are making it increasingly important.
  2. Information ecology is interested in the intersection of users, context and content
    • Context: Business goals, funding, politics, culture, technology, resources and constraints
    • Content: Data objects of different types & their metadata
    • Users: Audience, tasks, needs, information-seeking behaviour, user experience
  3. There are 4 main types of information seeking
    1. Known item seeking: You know what you're looking for. E.g. What's the population of London?
    2. Exploratory seeking: You don't know exactly what you're looking for, you need a few results to achieve your goal: What are the best restaurants in Kings Cross?
    3. Exhaustive Research: You want to see everything.
    4. Re-Finding: Finding information you've found before. Think 'pin to top' or 'favorite'.
  4. Information seeking behavior can include:
    • A combination of 'search', 'browse' and 'ask'
    • Search and browse are often used together. Users often only 'ask' when stuck
    • Berry picking behavior: users iterate their approach as information is revealed when interacting with the product
    • Pearl growing: users see content that resonates, and follow 'more like this' prompts (sideways?)
  5. IA helps create consistency and coherency. Even though content changes, and channels have different capabilities and limitations, users become familiar with semantic structures and become disorientated if they change.
  6. Components of information architecture:
    • Organisational Systems: how we organise information (by subject, chronology)
    • Navigation Systems: how we help users move through content (click hierarchy)
    • Search Systems: allow users to search content (executing a search against an index)
    • Labelling Systems: how we represent information (titles, abstracts, link)
  7. Information architecture can be determined top down (shaped by user journeys) or bottom up (shaped by the information/content structure)
  8. A single user action (like searching) can touch many IA decisions (query builders, indexing decisions, display decisions, search tradeoffs) which impact the user experience.
  9. Information Architecture is challenging. Language is ambiguous. Information can be diverse and seem unrelated. People see things differently. Company politics can get in the way.

Organisation Systems:

  1. Organisation schemes can be exact (alphabetical, geographical, chronological) or ambiguous (topic, task, audience specific, metaphorical, hybrids). Supermarkets use topic/task hybrids
    • Exact schemes are best for known-item searching. They're also easy to maintain and automate.
    • Ambiguous schemes are difficult to design and maintain. However, they're more useful for exploratory seeking, associated learning and serendipity.
      • Libraries offer author, title and subject schemes. Subjects are more popular because customers often don't know exactly what they're looking for.
  2. Hierarchies (Top Down) are typically mutually exclusive groups with parent-child relationships. They give users an overview of the information space and their place in it.
    • think about the tradeoff between exclusivity and inclusivity. It's hard to make ambiguous organisational schemes mutually exclusive. Hierarchies that cross list items are called polyhierarchical.
    • think about balance between breadth and depth. Breadth refers to options per level. Depth refers to number of levels. Avoid both 'narrow and deep' and 'broad and shallow'. Don't overwhelm the user with options or make them click through lots of levels. If in doubt, start with broad and shallow, levels can be added as content increases.
  3. Databases (Bottom up). Without defining a hierarchy, its still possible to leverage the metadata we have on information objects to make them findable. Using tags to create powerful searching, browsing, filtering and dynamic linking. This approach is scalable, and best paired with a more traditional top-down approach
  4. Hypertext (Sideways) Links are flexible and can be used creatively to make useful relationships between content, unlocking new pathways. However links don't provide the user with a sense of place, or help them understand the nature of the relationship. Used in isolation, they are confusing and overwhelming, but can be a powerful supplement to a more traditional hierarchy
  5. Social classification (Sideways)Users generate content and tag it (in text or tag fields) creating 'folksonomies'. Public tags allow users to move freely between objects (that wouldn't otherwise be grouped together) enabling discovery. Large numbers of engaged people are needed to make tagging work.

Navigation Systems:

  1. Labels are representations of concepts. 'Contact us' triggers a powerful understanding of the information that likely sits behind it (without us thinking).
    • Spoken language is essentially a labelling system for concepts and things
    • Its harder to convey meaning through products than conversation, its one way, you have no live feedback
    • Use language familiar to your users, that makes sense in context and alongside content.
  2. Create better labels:
    • Labels are hard because language is ambiguous (synonyms, homonyms, context). Content, users and context affect the perception of a labels meaning.
    • Apply a consistent labelling system across the product. Style, presentation, syntax, granularity, comprehensiveness, audience comprehension levels (medical PHDs?)
    • Audit labels in a single document. Audit comparable products and competitors. Adopt industry standard topologies or controlled vocabularies.
    • For inspiration: look at your content, look at your search terms, ask authors, ask user advocates, test with users (card sorting, free listing). Label tomorrows product features now, they should impact today's choices.
    • Links mean different things to different people. They require trust to action. Surround them with clear context. Make it clear where the link takes the user. Provide guidelines to authors.
    • Headings describe content that follows them. Headings establish hierarchy within content. Hierarchy between headings is established with styling. Users follow reading paths dictated by heading hierarchy
    • Navigation labels are best when repeated and kept consistent throughout the environment. Avoid using the same label for two different things.
    • Icons are aesthetically pleasing, but users have a limited icon vocabulary. Text labels are more clear. If you don't have space, use icons, but only common ones.
  3. Global Navigation: shown on every page, shows the user where they are, gives them quick access to major tasks
  4. Local Navigation: shows what's nearby (hierarchy levels, steps), compliments global
  5. Contextual navigation: shows what's related to what's here. Create new connections and pathways. Where does the user want to go next?
  6. Supplemental navigation: Only important in large information environments. Used as a backup to main navigation
    • Sitemaps reinforce information hierarchy, good for known-item seeking. Strip out content and provide access.
    • Indexes: alphabetical keywords. Granularity must be sensible. Consider term rotation "Car hire" and "Hire car"
    • Guides: best for new users, or sneak peaks of paid products. Keep them short, allow users to exit, keep navigation consistent, allow users to step forward and backward.
  7. Configurators used for complex decision trees. Show the impact of each choice as you go (change product images, price, shipping dates etc)
  8. Search: Users use their own terms. Allows specificity. However, there's ambiguity in language which causes problems.
  9. Personalisation is guessing what the user wants (based on behaviour, needs). Customisation lets the user specify what they want (presentation, navigation, content). Both are hard to do well. Users don't spend much time customising. Customising only applies to power users who return frequently, and even they don't always know what's best for them.
  10. Social: when actions of related individuals has value. Feed are ordered based on social graph. Good for discovery. Dynamic social navigation systems are becoming increasingly complex and useful. Beware the echo chamber!
  11. Navigation tips:
    • Innovate selectively. Use interaction norms and platform conventions.
    • Let users know they've arrived (brand identity). Provide a sense of where they are (information hierarchy) and what they can do (actions). Stress test it.
    • Balance the flexibility of movement, with user overwhelm from navigational clutter
    • Plan how your global, local and contextual navigation work together

Search Systems:

  1. Search often has a high engineering cost, so be sure you need it. Don't build search because navigation is broken, fix that first.
  2. Search works great when there's lots of content, or content is dynamic or fragmented.
  3. Search logs help you understand what your users want
  4. Indexing most things allows for richer results. Exclude documents or content components that users don't need to see, or that will hinder search results (e.g reviews of restaurants could mention competitors and confuse results).
  5. Only use search zones when users have shown an interest in the segment, would expect results only from that segment and when segmenting would improve results.
    • Segment ideas: content type, audience, roles, subject, topic, chronology, geography, author or business unit
    • Tradeoff between improving results and introducing complexity. Many users will ignore search zones.

3. There's a tradeoff between precision and recall.

  • Precision is what proportion of the records returned were useful.
  • Recall is what proportion of the useful results in the system were returned
  • Often, you can only increase recall at the expense of precision (and vice versa).
  • Balance knowing your user needs, are they 'known-item seeking', 'exploratory seeking' or doing 'exhaustive research'

4. Use query builders to increase the effectiveness of queries

  • Spell checkers correct spelling before returning results
  • Phonetic tools (soundex) expand queries like 'Smith' to include "Smyth'.
  • Stemming tools retrieve documents by finding words that use the same stem
  • Natural language processing tools take 'how to' and 'who is' and use that to narrow retrieval results.
  • Controlled vocabularies and thesauri leverage semantic nature of a query by including synonyms within the query

5. What to show in search results:

  • Pick content components based on user needs.
  • Known-item searches benefit from less content and more results. Exploratory seeking benefits from more content (abstracts)
  • Consider providing user choice for result density and display type (list, map)
  • Use pagination but don't expect users to venture past the first page.
  • If the documents you search over are flat (no headings) show the sentences that surround your search term for context
  • Show the search query & the number of results, helping users iterate on searches
  • Including key call to actions in search results (e.g. 'Get' in AppStore)
  • Consider allowing users to save searches or 'pin' specific results to read later
  • Explain if you've edited the search term
  • Explain where the results came from if you've done something non-obvious
  • Integrate searching with browsing: look for opportunities to connect your search and browse experience

6. Default to ranking results by relevance. Relevancy ranking is often a combination of:

  • how many of the query terms appear in the doc
  • how frequently they appear in the doc
  • how close together they are
  • where the terms occur (title or body)
  • the popularity of the document.
    • popularity should play a larger role if you have the data (Google · PageRank)

7. When to deviate from ranking by relevance:

  • Alphabetically for names
  • Chronologically for time bound results (news, tweets)
  • Allow sorting if it helps with a task (sort by price)
  • Human editorial
  • Ads
  • Rankings / Reviews (sort best first)
  • Consider grouping results by type

8. Consider query language support and advanced search interfaces if you have a captive audience of frequent power users.

9. Keep search limited to a single box if you can, and keep it away from other boxes

10. AutoComplete and AutoSuggest help identify potential matches based on partial or incomplete information. Can give users hints as to how the system is structured, help them iterate on searches

Theasuri, Controlled Vocabularies, and Metadata

  1. Metadata describes the thing. Examples include: means of creation, purpose, time & date of creation, author, location, standards used, tags
    • tags are used to describe documents, pages, images, software, video and audio files and other content objects for the purposes of improved navigation and retrieval
    • HTML has a <meta> tag, which authors can stuff with keywords for search engines
  2. Metadata-driven systems support powerful navigation and discovery environments. In large metadata-driven products, controlled vocabularies act as the glue that holds the system together.
  3. Types of controlled vocabularies:
    1. Synonym Ring: connects a set of words that are defined as equivalent for the purposes of retrieval.
    2. Authority Files: List of preferred terms or accepted variables. They are synonym rings in which one term has been defined as the preferred term
    3. Classification schemes: They are authority files, but also introduce the concept of a term hierarchy.
      • Example: The Dewey Decimal Classification {1876} (DDC) book classification used in libraries. Starts with 10 master categories, and has others underneath
      • Netflix classification scheme helps movie discovery. Macro genres (Drama, comedy), micro genres (based on real life) and micro tags (#happyEnding) that are used to inform the categorisation process
    4. Theasuri: All of the above and more. A controlled vocabulary in which equivalence, hierarchical and associative relationships are identified for purposes of improved retrieval
      • Each preferred term becomes the centre:
        • Equivalence = manages synonyms
        • Hierarchical = classification into categories and sub categories
        • Associative = meaningful connections that aren't handled by hierarchical or equivalence relationships
    5. Implementation Tips:
      • you may confuse users, if you use these in search results without explaining it
      • you may reduce the relevance of search results (precision and recall tradeoff)
      • Synonym rings can dramatically improve recall (from 20-80%)
      • A good tradeoff might be to use synonym ring recall by default, but order exact matches towards the top
image

Deep Summary

Longer form notes, typically condensed, reworded and de-duplicated.

1) Problems that IA addresses
Information Overload
  • Find-ability: Yahoo was a curated hierarchical directory in the 1990s. Things that worked then don't work now as the volume of stuff is expanding
More ways to access information
  • Information is decoupled from the artefact that contains it, but also from the contexts in which we access it.
  • We can gather information about their usage, that we could never before
  • We can view information on more devices and in more ways than ever before
Places made of information
  • Its a challenge to make a coherent experience across different devices and context
  • We interact with the product through the use of language: labels, menus, descriptions, visual elements, content, and their relationships
  • These create an environment that differentiates these experiences and facilitates understanding
  • The words and images we use, help us understand what can and can't be done in the environment
Coherence across channels
  • When you serve users through multiple channels, the users' experience across them should be consistent and familiar.
  • Users should experience the same semantic structures.
  • The capabilities and limitations of each channel are different, the semantic structures should be familiar and consistent
Systems thinking
  • In the era of ecosystems, seeing the big picture is more important than ever, and less likely. We must go from boxes to arrows, tomorrow belongs to those who connect
  • You need to start with a basic understanding of what products and services you offer across what channels, and how they all interact with each other. What the capabilities are
  • If you're laying bricks, you need to know if you're building a Cathedral or a Garage
2) Defining IA
  • 4 definitions:
    • The structural design of shared information environments
    • The synthesis of organisation, labelling, search and navigation systems within digital, physical and cross-channel ecosystems
    • The art and science of shaping information products and experiences to support usability, find-ability and understanding
    • An emerging discipline and community of practice focused on bringing principles of design and architecture to the landscape
  • Concepts:
    • Information
    • Structuring, organising and labelling
    • Finding and managing
    • Art and science
  • Just because you can't see it, doesn't mean that it isn't there.
    • Chess can be played with physical pieces, or deconstructed back to its concepts so that it can be played digitally or via post.
    • Nobody set out to create an information architecture for chess, its just become emergent over the years
  • Towards a dammed good IA
    • Users. Context. Context. They form the basis of our model for practicing effective information architecture design. You can't design IA in a vacuum.
  • Information Ecology: Users, context and content
    • Context: Business goals, funding, politics, culture, technology, resources and constraints
    • Content: Documents, data types, content objects, volume, existing structure
      • Ownership, format, structure, metadata, volume, dynamism
        • Metadata: Manual or automatic, complete or partial, quality and consistency, controlled vocabulary or user submittable?
    • Users: Audience, tasks, needs, information-seeking behaviour, experience
3) Design for finding
The 4 main different types of information needs:
  1. Known item seeking
    • You know what you want
    • Example: Whats the population of London
  2. Exploratory seeking
    • You don't know exactly what you're looking for
    • You're happy to have a few good results returned (having only 1 wouldn't be enough)
    • Example: Best restaurants in Kings Cross
  3. Exhaustive research
    • You want to see everything, no stone unturned
    • Example: Conducting research for a PHD. Grid optimisation using ML.
  4. Re-finding
    • Information that you've happened on before.
    • That you'd prefer never to lose track of.
    • Pin/favourite/tag/follow/save for later
Information-Seeking Behaviours
  • Searching, browsing and asking
  • Integration and iteration
    • We often integrate searching, browsing and asking in the same session
    • Browse, Browse, Search, Ask for help
    • Berry Picking Model: formulate an initial query, move iteratively through an information system along potentially complex paths, picking bits of information along the way. In the process, they modify their information requests as they learn more about what they need and what information is available from the system
      • if Berry Picking is common, you want to make it easy for your users to move from search to browse and back again
    • Pearl growing model: start with one or two documents that are exactly what they need, they want to get 'more like this one'
    • 2 step model: search and then browse
How to learn about information seeking needs and behaviours:
  • Search analytics: what not why
  • Contextual inquiry: why
4) Design for understanding
  • We want the products we design to be understandable and usable.
  • Humans are grounded in a sense of place. Digital products are still places.
  • Cultural convention and patters of use change the way that we design places.
  • You should design in a way that follows certain standards and conventions but also takes into account your specific needs (an architect knows what a minimum ceiling height should be, but should also know he's designing a bank, and the bank needs a vault)

Structure and Order

  • Building entrances are often highlighted with porticos that serve as visual indicators to the way in
  • Semantic structures in information architecture also have hierarchies that indicate the relative importance of individual components
    • Top-Level links - limited to the highest level elements in a hierarchy
    • Rhythm and density - search results for different things maybe delivered in different ways

Typologies:

  • Digital products have (even though young) evolved typologies. Banks, airlines, newspaper websites have all evolved to look similar.
  • Choice of abstract architecture is important:
    • tells users what type of place they are in
    • makes it easier to understand and navigate the environment
    • having a standard environment, makes it easier to differentiate on the grounds of information: tone, words, numbers stand out more when architecture is similar and familiar

Modularity and Extensibility

  • Most information is dynamic and ephemeral, they are subject to constant change
  • Content changes faster than visual design which changes faster than information structures (which are often stable)
  • Buildings are composed of 6 laters, which change at different rates over time:
    • Site: the setting of the building
    • Structure: the skeleton of the building
    • Skin: The exterior surface of the building
    • Services: The working guts (electrical systems, heating etc)
    • Space plan: The internal layout - partitions and doors
    • Stuff: Furnishings, appliances, day-to-day objects
  • A well designed building, can accommodate many different use cases in its lifetime.
  • Digital products can change, but the semantic structures tend to be consistent
  • Users become used to the semantic structures, and become disorientated if they change too quickly
  • Information Architecture can be flexible or brittle. Flexible isn't always good, sometimes flexible can be less clear. We want somewhere in the middle, can accommodate change but is also clear and crisp.

Reminded me of this

The Happiest Place(s) on Earth

  • DisneyLand could be really complicated, instead it uses a hub and spoke model. DisneyLand in the centre, Main Street, Frontierland, AdventureIsland, FantasyLand, TomorrowLand on the periphery. All of the spokes contain attractions, restaurants, shops and services
  • Note the Disney copy choices: Customers = Guests, Staff = Cast members
  • Consistency between channels is important, but the information hierarchy must also be tailored to serve the specific information needs of each channel

Recap:

  • The structure of information environments influences more than how we find stuff, it also changes how we understand it
  • We experience information environments as places where we go to transact, learn, and connect with other people, among many other activities.
  • When designing information environments, we can learn from the design of physical environments

Basic Principles of IA

5) Anatomy of IA
Components of Information Architecture:
  • Organisational Systems: how we organise information (by subject, chronology)
  • Navigation Systems: how we help users move through content (click hierarchy)
  • Search Systems: allow users to search content (executing a search against an index)
  • Labelling Systems: how we represent information (titles, abstracts, link)
Browsing Aids
  • Organisational Systems: taxonomies and hierarchies
  • General navigation systems: help users understand where they can go
  • Local navigation systems: help users understand where they are, and where they can go within an information environment
  • Sitemaps/tables of content: supplement primary navigation, condensed overview and links to major content areas
  • Indices: Supplementary navigation systems that provide an alphabetised list of content
  • Guides: specialised information on specialist topics, and links to subset content
  • Walkthroughs and wizards: supplementary navigation systems, lead users through sets of steps
  • Contextual navigation: consistently presented links to related content, often embedded in text and generally used to connect specialised content within an information environment
Search Aids
  • Search interface: entering and revising a search query and configuration options
  • Query language: The grammar of a search query
  • Query builders: Ways of enhancing a query's performance: spell checkers, stemming, concept searching and drawing in synonyms from thesaurus
  • Retrieval algorithms: relevance ranking
  • Search zones: subsets of sites separately indexed for narrower searches
  • Search results: presentation of content that matches query, what type, how many, how they are ranked, sorted, clustered
Content and Tasks
  • These are the ultimate destination of the user, not the stepping stones.
  • Headings Labels for content that follows
  • Embedded links links within text
  • Embedded metadata information that can be used as metadata but must first be extracted, an ingredient can be in a recipe, but the ingredient is also its own thing
  • Chunks logical units of content, vary in granularity, and can be nested (sections and chapters)
  • Lists groups of chunks or links to chunks, these are important because they've been grouped together
  • Sequential aids: help a user understand where they are in a series of tasks
  • Identifiers: clues that suggest where the user is in an information system (breadcrumbs)
Invisible components:
  • Algorithms
  • Best bets and curation
  • Controlled vocabularies and thesauri
Top Down Information Architecture
  • Determining the most common questions or journeys, and designing the product to meet those
    • Where am I? I know what I'm looking for, how do I search for it? How do I get around the site? Whats important and unique about this organisation? Whats available on this site Whats happening there? How do I engage with them via various channels
Bottom up information architecture
  • Content can have information architecture embedded.
  • Instead of top-down, the architecture is suggested by the systems content (and inherent in the systems content)
  • More important, as often, users arrive at a page from search results, adds, external websites..
  • So knowing the following is important:
    • Where am I
    • What's here
    • Where can I go from here?
    • See this Stress test:
  • Example the iOS Photos app:
    • Its almost all information architecture. Its structure is defined almost entirely by the metadata and deep contextual links embedded in the content.
    • You can see your in collections, and can go back to years
    • Photos are separated by events and sorted chronologically
    • You can switch to shared or album
    • i
Invisible information architecture
  • Search on the BBC - returns articles, sorted by relevance, editors choice first, then surfaces images, titles and abstracts from those, as well as source 'news' 'world service'
  • All of these decisions are unknown to users,
6) Organisation Systems
Challenges to organising information

We're all becoming librarians. Technology has changed information forever, we need to innovate in content organisation

  • Ambiguity: Classification systems are made from language, language is ambiguous. 'Pitch' has 15 different meanings. Using words as labels is a shaky foundation. Is a tomato a fruit, vegetable or berry? Or all of them?
  • Heterogeneity: an object or collection composed of unrelated or unlike parts. The heterogeneous nature of modern information makes it difficult to impose any single structured organisation system on the content. Doesn't usually make sense to classify documents of varying levels of granularity side by side (articles shouldn't appear next to magazines). Assuming one size fits all is often a flaw in information hierarchies
  • Differences in perspectives: Finding something on somebodies desktop or personal filing system is difficult. The creators perspective hugely influences the outcome
    • So rely on users. How do they group information, what labels do they use, how do they navigate
  • Internal politics: Positioning for influence or respect. Organisation org charts being reflected in products.
Organising information environments
  • Organisational systems are composed of schemes and structures.
    • A scheme: the shared characteristics of content items and influences the logical grouping of those items
    • A structure: the types of relationships between content items and groups
  • Organisation is closely related to navigation, labelling and indexing.
    • Although its possible and useful to separate organisational systems from labelling and navigation systems
Organisation Schemes
  • We navigate through organisation schemes everyday.
    1. Exact organisation schemes
      • alphabetical order: e.g sorting works alphabetically in a dictionary
      • chronological
      • geographical
      • Useful for known-item searching, if you know the name of the thing you're looking for
      Ambiguous organisation schemes
      • Divide information into categories that defy exact definition. Ambiguity of language and organisation, human subjectivity involved.
      • supermarket topic/task hybrid. Marshmallows in snacks or baking ingredients?
        • Topical: by subject/topic (e.g newspapers (sport, travel, news)
        • Task orientated:
          • Good when there is a limited number of high priority tasks
          • Common in desktop and mobile apps - especially those that support creation and management of content (MS Word)
          • On the web, common when customer interaction takes centre stage, often embedded into other topical schemes
        • Audience-specific: Best when seperating an audience that has different needs. E.g Cern website: Students · Scientists · Cern people
          • Can be open or closed (restricted to login)
        • Metaphor driven schemes:
          • Desktop computer. Material Design. Skeuomorphism
          • Use with caution:
            • metaphors have to be common to succeed
            • unwanted baggage or limitations
        • Hybrids:
          • Blend too many elements or schemes and you can create confusion.
          • Mixing schemes at the same level causes confusion, breaks the mental model
          • Shallow hybrid schemes can be OK, deep ones are often not
            • Topics & Tasks mixed in surface level primary nav on a website is OK
            • Using a hybrid model breaks when you have large volumes of content though
      Pros and Cons
      • Exact organisation schemes:
        • are easy to maintain and automate
        • easy to use for known item searching
        • no good if you don't know what you're looking for
      • Ambiguous organisation schemes:
        • difficult to design (requires rigorous user testing)
        • difficult to maintain (requires human input)
        • helpful when you don't know exactly what you're looking for
        • helpful for associated learning, serendipity and recommendations
        • more useful and used more
          • Libraries often have author, title and subject schemes
          • subject schemes are used more often, as people often don't know what they want
          • are difficult to design,
    2. Information seeking is often iterative and interactive. What you find at the beginning of your search influences what you look for and find later
    3. Ambiguous organisation supports this serendipitous mode of information seeking by grouping items in intellectually meaningful ways
      • grouping of related items supports associated learning, process that may enable the user to make new connections and reach better conclusions
Organisation Structures
  • We rarely think about them
  • Movies are linear in their physical structure - frame by frame, beginning to end
  • The plotts themselves can be non-linear.
  • Maps are linear, they have spacial structure. Many maps cheat, sacrificing accuracy for clarity

The Hierarchy: A top-down approach:

  • Hierarchies are ubiquitous, have been for a longtime and are the foundation of most organisation structures
  • Users understand and are comfortable with hierarchies, they help provide a model of the information space to the user, and where they are in it
  • Mutually exclusive subdivisions & parent-child relationships are simple and familiar
  • The top-down hierarchy allows you to gain a quick understanding of the scope of the environment

Designing Hierarchies

  • Rules of thumb:
  • Mutually exclusive categories (if you can)
    • Tradeoff between exclusivity and inclusivity
    • Hierarchies show cross-listing are polyhierarchical.
    • Ambiguous organizational schemes make it difficult to follow mutually exclusive
    Balance breadth and depth
    • Breadth = options per level
    • Depth = number of levels
    • Looking to avoid both 'narrow and deep' and 'broad and shallow'
    • Recognize the danger or overloading the user with too many options
    • Group and structure information at page level
    • Test your designs with users
    • If worried about conversion, be concerned about depth. Limit to 2 or 3 levels if you can
    • For new information environments, broad and shallow is more favorable as its more scalable. Easier to add levels to secondary levels of hierarchy than to the main page.

The Database Model: A Bottom-up approach:

  • Metadata links database structures to information architecture. We can benefit from the structure and power of relational databases and apply it to the heterogeneous, unstructured environments of websites and intranets.
  • By tagging documents and other information objects with powerful metadata, we enable powerful searching, browsing, filtering and dynamic linking.
  • Relationships between metadata elements can become complex.
  • You need to understand how metadata, controlled vocabularies and database structures can be used to enable:
    • Automatic generation of alphabetical indexes
    • Dynamic presentation of associative "see also" links and content
    • Fielded searching
    • Advanced filtering and sorting of search results

Hypertext (Hyperlinks)

  • Highly nonlinear way of structuring information. Provide flexibility but can also be confusing.
  • A sense of place, and relationships between content can be less obvious to some. Without context, users can become overwhelmed and frustrated.
  • Hypertext is rarely a food candidate for organization structure, but can be used to compliment others
  • Allows for useful and creative relationships between content.
  • First design the hierarchy, then see how hypertext can compliment it

Social Classification:

  • User-generated content and tagging.
  • Free tagging (a.k.a: collaborative categorization, mob-indexing, and ethno-classification)
    • in text fields, or in special tag fields
  • Tags are public, and serve as pivots for social navigation. Move freely between objects, authors etc
  • When large numbers of people get involved, tagging can be powerful
  • No single person or team create created a taxonomy to create these relationships
  • Social classification (folksonomies) haven't replaced traditional top-down taxonomies, but they do augment them well

Creating Cohesive organization structures:

  • The first step, is to explore its organization. Consider a variety of exact and ambiguous schemes. By topic, task, or by audience? Chronological or geographical? What about using multiple organization structures.
  • Then think about user navigation. Hierarchy or database? Loose hypertext?
  • All information based retrieval systems work best when applied to narrow domains

Things to remember:

  • Difference between exact and ambiguous schemes
  • Exact schemes are best for known item searching. Ambiguous are better then users only have a vague idea of what they want
  • Whenever possible, use both types of schemes.
  • Language is ambiguous, content it heterogeneous, people have different perspectives, politics can be a challenge. Using multiple schemes can be a good way to deal with these challenges
  • Large systems often need multiple types of structure
  • Top-level umbrella architecture should be hierarchical
  • Keep eyes out for collections of structured and homogenous information. These are good for database models.
  • More creative relationships, can be created by users with hypertext or tagging
7) Labeling Systems
  • Labels are representations of concepts and thoughts
  • "Contant Us" is a label that represents a chunk of content
  • Works as a shortcut that triggers the right association, without having to present all the information
  • The goal of a label is to communicate information effectively, to convey meaning without taking up space or requiring the user to think too hard
  • Spoken language is essentially a labelling system for concepts and things
  • When in conversation we rely on user feedback and signals, which enables us to adapt. Its harder to communicate through digital products, labels are more important
  • Aim is to design labels that speak the same language as our environment's users, while reflecting its content.
  • Labels should educate people about new concepts and help them quickly identify familiar ones
  • Testing the labels on a main page: Do the prominent labels standout? If so why/

Varieties of labels:

Contextual links: links to chunks of information on other pages or locations on the page
  • the basis for exciting interconnections thats driven the webs success
  • these are often not systematic, personal to the authors of content
  • links mean different things to different people, if we saw 'shakespeare' linked, where would we expect that to go?
  • Links are better understood when you draw meaning form the authors surrounding text, provides more context
  • Non-representational links require trust. Clarity & Trust makes something more clickable
  • Ask: what kind of information would the user expect to be taken to?
  • Usually content authors are responsible for contextual links, so provide them with guidelines
Headings: describe the content that follows them
  • Used to establish hierarchy within content (categories from sub categories)
  • Hierarchy between headings is established by numbering, font sizing, colours, styles, indentation and whitespace.
  • You often need the user to follow a reading sequence, when you do heading hierarchies are useful
Navigation system choices: the options in a navigation system
  • Need to be consistent, more so than any other label
  • Navigation is repeated throughout the environment, so navigation problems are magnified through repeated exposure
  • They should be consistent, behave rationally.
  • They should be in a consistent location, and provide a sense of familiarity when moving from page to page
  • Be careful when using the same label text to describe two different things
Index terms: Keywords, tags, subject headings that represent content for searching or browsing.
  • Help with search, can also help with browsing
  • Sometimes index terms are not visible in the UI. You can add a bunch of terms to your homepage for Google to scrape
Iconic labels: Icons
  • Icons can represent (a few ubiquitous) concepts too
  • Best when option set is small (icon vocab is limited)
  • Repeated exposure can help users learn an icon language
  • Useful for navigation, when space is limited
  • Default back to text if concepts aren't ubiquitous, or there are lots of options.
  • Icons are aesthetically pleasing, but you risk usability. Be wary of putting form ahead of function

Designing labels:

  • Language is too ambiguous for your to feel confident you've perfected a label.
    • Synonyms, homonyms and changes in context are pitfalls
  • General Guidelines:
    • Content, users and context affect the perception of a label
    • How can we help:
      • Narrow scope wherever possible - the narrower the context, the more clear the label
      • Develop consistent labeling systems, not labels. Consistency = predictability. Style, presentation, syntax, granularity, comprehensiveness (otherwise lose faith), audience (doctor facing language or patient facing language)
    • Start with what you have. Bring them into a single document.
    • Look at comparable or competitive environments. There are industry specific typologies.
    • A great source of labels is existing controlled vocabularies and thesauri. These are great for indexing content. Seek out narrowly focused vocabularies that help specific audiences to access specific types of content. Think like the user
      • Creating new labeling systems, where to get them from:
        • Content analysis - labels can come directly from content
        • Content authors - ask them to submit labels
        • User advocates and subject matter experts - allow them to input
        • Users:
          • Card sorting
          • Free listing (user brainstorm words to describe a term)
          • Search term analysis
      • Tuning and tweaking:
        • Get all of your favorites into a list, review them for consistency.
        • If you have gaps, or future requirements, fill in those labels, as they may impact your current choices
8) Navigation Systems

Types of Navigation Systems:

Global, local and contextual navigation
  • Provide context and flexibility, helping users understand where they are, and where they can go
Global = Where am I?
  • Present on every page (top nav), allowing direct access to key areas and functions, no matter where the user travels in the sites hierarchy
  • Most navigation bars contain a logo, homepage link, search and links to major areas
  • Headers and footers are now also widely used for global navigation.
    • Headers often mega menus, can have 2nd and 3rd levels
    • Fat footers are abridged site maps
  • Heavily test global navigation.
  • Local = What's nearby? (side nav)
    • Compliments the global
  • Contextual navigation = What's related to what's here? (inline links)
    • See also links. Wikipedia links out to other definitions and concepts.
    • Support associative learning. Users learn by exploring relationships.
    • Contextual navigation allows you to create a web of connective tissue that benefits users and the organization.
    • Often determined by the author, subject matter expert or editor... once the content is known.
    • Use inline links for non-critical points of interest.
    • Imagine your user has landed directly on this page. Where might they want to go next? How might contextual links help with that?
Supplemental navigation

Are critical in large information environments, but less so in smaller ones. They are often implemented badly. The hierarchy and taxonomy will always fail for some use cases. Supplemental navigation systems can be the emergency backup for when the global navigation system fails.

  • Sitemaps, indexes and guides
  • Sitemaps
    • Sitemaps are good for large systems with a clear hierarchy
    • Good site maps: 1. Reinforce the information hierarchy 2. Provide direct access for users who know what they want 3. Strip out content
    Indexes
    • Presents keywords or phrases alphabetically
    • Great for known item finding, bypasses the hierarchy entirely
    • Key challenge is with granularity - base on what the users are usually looking for. Look at search logs, conduct research
    • Can be curated, small and manual, or large and automated.
    • Try using term rotation in an index "Car hire" and "Hire car" to increase findability
    Guides
    • Guided tours, tutorials or targetted walkthroughs
    • Good for introducing new users.
    • Good if you're behind a paywall, you can give a sneak peak
    • Typically follow a linear navigation pattern, with limited interaction
    • Tips for guides:
      • Keep them short
      • Allow users to exit
      • Keep navigation consistent, allowing users to move back & forwards
      • Designed to answer questions
      • Should be crisp, clear and optimized with enlarged details of key features
    • Many people may never use the guide.
    Configurators
    • Use for complex decision trees
    • Show the impact of the choices (product image changes, price changes)
    Search (see chapter 9 also)
    • Searching is a central part of supplemental navigation
    • Puts users in the driver's seat, they can use their own words to drive navigation
    • Enables specificity.
    • Ambiguity of language causes huge problems. We all use different words for the same thing.
    • This is super difficult, so see chapter 9
    Personalisation and Customization
    • Personalisation: we guess what the user wants. VS. Customisation: The user tells us what they want
    Personalisation: Choose what to serve based on user behaviour, meeds or preferences.
    Customization: user has control over an aspect of presentation, navigation and content.
    • Play important but limited roles, require solid foundation of structure and organization, are really difficult to do well, can make it more difficult to collect metrics.
    • Customization: users don't want to spend much time customising.
    • Customisation only really works for power users and repeat visitors.
    • Customisation is problematic, as users often don't know what will be useful to them later.
    Visualisation
    • Useful when you can identify what you want from how it looks (clothing)
    Social
    • When value for the user can be derived form observing the actions of other users, especially those with a meaningful relation to that individual
    • Helps users discover content based on popularity. Good for content aggregation and discover
    • Feeds that are ordered based on the social graph
    • Dynamically generated social navigation systems will become increasingly complex, sophisticated and useful.
    • Need to avoid the echo chamber problems.
  • Exist outside of the content-bearing pages
  • Similar to search, provide a birds-eye view of the environment
Navigation tips:
  • Think about how your product appears. If its in a web browser (back & forward), or if its an app (touch features). Follow the norms.
  • Placemaking: what the environment is, what you can do in it - makes information more easily understandable. Create a sense of place through clear language, and clear paths to explore.
  • Users need to know they've arrived. They need a 'You are here' before they know where to go next
    • Users should always know what site or app they're in (even if they arrive direct to a page). Extending the organisations logo and brand identity throughout is an obvious way to do this
    • The navigation structure, should present the information hierarchy clearly, indicating the users current location. This helps people build a mental model of the organization scheme, which facilitates navigation and makes them feel comfortable. Use the
  • Hierarchy can form the basis of IA, but don't restrict navigation to moving up, down and within a hierarchy. Balance flexibility against navigation clutter, add hypertext to provide key quick shortcuts
  • Balance the flexibility of movement, with user overwhelm from navigational clutter
  • You need to take a holistic design approach, and work out how global, local and contextual navigation can work together.
  • Text labels are more clear. If you don't have space, use icons, but only common ones
  • Follow the conventions: Websites: Global nav at the top. Mobile: icons at the bottom.
9) Search Systems
Do you need search?
  • Many users are natural browsers. Be careful of using just search
  • Consider the following:
    • How much content do you have? Search is more valuable the more you have
    • Don't build search because other means of navigation are broken, fix those first
    • Engineering complexity to build, maintain and optimise
    • Other alternatives: index is easier to implement.
    • User preferences: know if your users want to search or browse
    • Search helps when there is too much information to browse
    • Search helps fragmented sites
    • Search is a learning tool. Will search logs be valuable to you?
    • Is search expected by your users?
    • Search helps tame dynamic environments, where things are changing. E.g: News
Choosing what to index
  • Start with everything
  • Think about adding search zones. When users may only want to search in a given silo.
  • Search zones are pockets of homogeneous content, reduces the apples and oranges effect and allows users to focus their searches
  • Each document or record has a structure. Content atoms. You can choose to leave out parts from the index.
  • If there's obviously more important content, consider filtering for that by default and allowing users to expand search beyond it by exception (E-commerce search only returning product results, not site wide results including products)
Determining Search Zones:
  • Zones are subsets of information that have been indexed separately.
  • The user should have already identified themselves as interested in that particular information.
  • As the search relates only to his needs, it will result in a better search experience. Retrieving fewer, more relevant results
  • Types of search zone:
    • Content type, audience, roles, subject or topic, geography, chronology, author, department or business unit
  • Tradeoff between improving results and introducing complexity. Many users will ignore search zones.

Navigation versus destination: You probably want to remove navigation pages from results

Indexing for specific audiences: If you have an audience-orientated organization scheme, then consider having search zones by audience as well.

Search zones should only be used if there's little overlap, and a large reduction in filtered out content / results.

Indexing by topic: helpful for reducing large numbers of results

Indexing recent content: Content that is chronological by nature like news, it can be useful to provide filters by date. You may want to filter out news from previous decades.

Selecting content components to index:
  • If documents have administrative information, that isn't meaningful, consider excluding it
  • Exclude things that could confuse your results. Reviews of restaurants could contain names of competitor restaurants. We may choose to exclude these from search
  • Documents structures can help us render the search results in a more enriched way. Indexing more content components allows us to have a better experience once the search results need to be rendered,
Search Algorithms
  • Their task is to return the best pool of documents. Best is subjective, so know the users needs
  • There are many information retrieval algorithms (40+), choosing the right one depends on the problem
Pattern-Matching Algorithms: Precision vs Recall tradeoff
  • Compare user query with an index of, typically, full text of search documents, looking for the same string of text.
    • When a matching string is found, that document is added to the retrieval set
  • Recall vs precision: they're inversely related
    • Some return a few high-quality results, some return numerous results with a relevancy ranking
      • Precision ratio = relevant documents / documents retrieved
      • Recall ratio = relevant documents / total relevant documents in the system
    • The balance between these two, will depend on the nature of the search:
      • Exhaustive research: searching for your name, or doing research for a PHD, you want it to return lots of results
      • Otherwise, a user may just want 2-3 best articles on a topic, or definitions of a term. They don't need more than a handful of the best responses.
Complexities
  • A search tool may provide automatic stemming - taking the stem of a word, which expands a term to include other terms that share the same root or stem
  • How structured is the content? Can you increase the relevance of search results that are in headings of HTML vs the body
  • Cited by, related documents, related documents from co-citation

Query Builders
  • Tools that soup up a query's performance
  • Spell checkers: correct spelling before returning results
  • Phonetic tools: Soundex, expand on a query like 'Smith' to include 'Smyth'
  • Stemming tools: Enter a term and retrieve documents by finding words that use the same stem
  • Natural language processing tools: take 'how to' and 'who is' and use that to narrow retrieval results.
  • Controlled vocabularies and thesauri: leverage semantic nature of a query by including synonyms within the query
Presenting results
  • The 2 big questions:
    • What content components to include in search results? (atomic components)
    • How to group the search results?
Which content components to include?
  • Display less information to users who know what they're looking for.
  • Display more information to users who don't know what they're looking for
  • Show users, who know what they're looking for only representational content. components, like title or author, to help them distinguish between results
  • Users who don't know what they want will benefit from descriptive content abstracts, summaries and keywords
  • You may provide user choice - result density, or display type (list or map)
  • Depends on the number of results - if you only have a few, you maybe able to display more
  • Users never venture past the first screen of results
  • What components are useful to the user. Include phone numbers in results for a phone directory
  • If your documents/content doesn't have a hierarchy within it, e.g. just flat text. Consider showing the part of that document that matched the search result in the search results itself... the preceding and following sentences. Helps the users scan for relevance
How many to include?
  • Depends on the content components you'll include, and how much space they take
  • Let the users know how many documents you've retrieved
  • Consider providing a results navigation system to help them move through results (pagination)
  • Show the original search in the box, allowing users to iterate and try again easily
Listing results (Sorting and ranking options)
  • In what order should they be displayed?
  • Sorting or ranking?
    • If users have a specific task or action in mind, sorting might be useful. E.g: Sort by price
    • Ranking is more useful when there's a need to understand the information or learn something. Relevance, from most to least is the most common.
  • Sort by alphabet - good for names
  • Sory by chronology - good for news, tweets or time sensitive stuff
  • Ranking by relevance:
    • How many of the query terms are in the document
    • How frequently the query terms are in the document
    • How close together the terms occur
    • Where the terms occur (title or body)
    • The popularity of the document
  • You can add human indexing, like the BBC do with editors choice
  • Ranking by popularity: how google works. How many links to a page. Quality of those links. PageRank.
  • Ranking by users or experts: User ratings for restaurants for example. Most sites don't have sufficient user engagement for this
  • Ranking by pay for placement: ads and auctions, perhaps going to result in most relevant ones otherwise they're going to go out of business paying for irrelevant terms

Grouping results
  • Clustering retrieved results by some common aspect.
    • Different types of information, or different topics etc
Acting on results"
  • Call to actions: 'Get' on Appstore
  • Select a subset of results: add to list, save
  • Save a search: Save search link in upper right corner
Designing the search interface

Relevant variables to deciding on the search interface:

  • Level of searching expertise and motivation
    • Do they know query language? Are they willing to iterate? Are they happy with good enough, or do they need a certain result?
  • Type of information needed:
    • A taste or comprehensive research? What components help them decide if they should click through? Should results be brief or rich? How detailed a query are they willing to write?
  • Types of information being searched:
    • Structured documents or full text? How many results is right?

In the absence of amazing search, if you have a captive audience you may want to support a query language

  • Users make assumptions about how search works, you may want to check yours:
    • They may expect natural language processing, 'AND/OR' support, that the query will search the entire site, or just a subset.
  • You can provide a help page, but nobody will check it.
  • You can educate users about search, after they've searched.
    • Typing watch into ebay yields 1.6m results, so you now know you have to be more specific
    • Users will learn to refine their search if you give them too many or too few results
    • Think about supporting revision of the search after showing the results
  • Keep search limited to a single box if you can, and keep it away from other boxes
  • If you need additional fields to make a search useful (flights) then make them prominent and have a button marked 'search' to trigger the search

AutoComplete and AutoSuggest

  • A list of results is presented alongside the search box, prompts the user to select a specific term, shows them that there are results for those terms.
  • Help users identify potential matches based on partial or incomplete information. Can give them hints as to how the system is structured.

Consider advanced search interfaces or support if you have frequent power users

Supporting revision
  • Display the initial search query with the results, so users know what they searched for and have a quick place to edit it
  • Explain where results came from: if you searched across multiple zones, let people know where the results came from
  • Explain what the user did - if you get no results, explain what the query did, and what results were returned
  • Integrate searching with browsing: look for opportunities to connect your search and browse systems
  • If the user gets too many, make it clear they've got loads of results, let them filter them ~
Presenting results
  • The 2 big questions:
    • What content components to include in search results? (atomic components)
    • How to group the search results?
Which content components to include?
  • Display less information to users who know what they're looking for.
  • Display more information to users who don't know what they're looking for
  • Show users, who know what they're looking for only representational content. components, like title or author, to help them distinguish between results
  • Users who don't know what they want will benefit from descriptive content abstracts, summaries and keywords
  • You may provide user choice - result density, or display type (list or map)
  • Depends on the number of results - if you only have a few, you maybe able to display more
  • Users never venture past the first screen of results
  • What components are useful to the user. Include phone numbers in results for a phone directory
  • If your documents/content doesn't have a hierarchy within it, e.g. just flat text. Consider showing the part of that document that matched the search result in the search results itself... the preceding and following sentences. Helps the users scan for relevance
How many to include?
  • Depends on the content components you'll include, and how much space they take
  • Let the users know how many documents you've retrieved
  • Consider providing a results navigation system to help them move through results (pagination)
  • Show the original search in the box, allowing users to iterate and try again easily
Listing results (Sorting and ranking options)
  • In what order should they be displayed?
  • Sorting or ranking?
    • If users have a specific task or action in mind, sorting might be useful. E.g: Sort by price
    • Ranking is more useful when there's a need to understand the information or learn something. Relevance, from most to least is the most common.
  • Sort by alphabet - good for names
  • Sory by chronology - good for news, tweets or time sensitive stuff
  • Ranking by relevance:
    • How many of the query terms are in the document
    • How frequently the query terms are in the document
    • How close together the terms occur
    • Where the terms occur (title or body)
    • The popularity of the document
  • You can add human indexing, like the BBC do with editors choice
  • Ranking by popularity: how google works. How many links to a page. Quality of those links. PageRank.
  • Ranking by users or experts: User ratings for restaurants for example. Most sites don't have sufficient user engagement for this
  • Ranking by pay for placement: ads and auctions, perhaps going to result in most relevant ones otherwise they're going to go out of business paying for irrelevant terms

Grouping results
  • Clustering retrieved results by some common aspect.
    • Different types of information, or different topics etc
Acting on results"
  • Call to actions: 'Get' on Appstore
  • Select a subset of results: add to list, save
  • Save a search: Save search link in upper right corner
Designing the search interface

Relevant variables to deciding on the search interface:

  • Level of searching expertise and motivation
    • Do they know query language? Are they willing to iterate? Are they happy with good enough, or do they need a certain result?
  • Type of information needed:
    • A taste or comprehensive research? What components help them decide if they should click through? Should results be brief or rich? How detailed a query are they willing to write?
  • Types of information being searched:
    • Structured documents or full text? How many results is right?

In the absence of amazing search, if you have a captive audience you may want to support a query language

  • Users make assumptions about how search works, you may want to check yours:
    • They may expect natural language processing, 'AND/OR' support, that the query will search the entire site, or just a subset.
  • You can provide a help page, but nobody will check it.
  • You can educate users about search, after they've searched.
    • Typing watch into ebay yields 1.6m results, so you now know you have to be more specific
    • Users will learn to refine their search if you give them too many or too few results
    • Think about supporting revision of the search after showing the results
  • Keep search limited to a single box if you can, and keep it away from other boxes
  • If you need additional fields to make a search useful (flights) then make them prominent and have a button marked 'search' to trigger the search

AutoComplete and AutoSuggest

  • A list of results is presented alongside the search box, prompts the user to select a specific term, shows them that there are results for those terms.
  • Help users identify potential matches based on partial or incomplete information. Can give them hints as to how the system is structured.

Consider advanced search interfaces or support if you have frequent power users

Supporting revision
  • Display the initial search query with the results, so users know what they searched for and have a quick place to edit it
  • Explain where results came from: if you searched across multiple zones, let people know where the results came from
  • Explain what the user did - if you get no results, explain what the query did, and what results were returned
  • Integrate searching with browsing: look for opportunities to connect your search and browse systems
  • If the user gets too many, make it clear they've got loads of results, let them filter them ~
10) Theasuri, Controlled Vocabularies, and Metadata
  • Products are a collection of interconnected systems with complex dependencies
  • A link can be part of the products structure, organisation, labelling, navigation and searching systems
  • In large metadata-driven products, controlled vocabularies act as the glue that holds the system together.
Metadata: data providing information about one or more aspects of the data
  • Means of creation, purpose, time and date of creation, author, location, standards used
  • Metadata tags are used to describe documents, pages, images, software, video and audio files., and other content objects for the purposes of improved navigation and retrieval.
  • HTML has a <meta> tag, which authors can stuff with keywords for search engines
  • Many companies are now leveraging metadata to crate metadata-driven systems that support distributed authoring and powerful navigation.
  • Metadata asks the question: How do I describe this thing?
Controlled Vocabularies: defined subset of natural language
  • a list of equivalent terms, in the form of a synonym ring, or a list of preferred terms in the form of an authority file.
  • Define relationships between words and you have a classification scheme. Model associative relationships between concepts and you're working on a thesaurus
image
  • A full-blown thesaurus integrates all the relationships and capabilities of the simpler form
  • Synonym Ring: connects a set of words that are defined as equivalent for the purposes of retrieval
    • you may confuse users, if you use these in search results without explaining it
    • you may reduce the relevance of search results (precision and recall tradeoff)
      • After a while, you can only increase one at the expense of the ther
      • Precision = relevance of documents within the set
      • Recall = portion of relevant documents in the results vs the system
    • Synonym rings can dramatically improve recall (from 20-80%)
    • A good tradeoff might be to use synonym ring recall by default, but order exact matches towards the top
    Authority Files: List of preferred terms or accepted variables
    • Traditionally used by libraries and governments to define the proper names ofr a set of entities within a limited domain
    • Authority files are inclusive of both preferred and variant terms. They are in essence synonym rings in which one term has been defined as the preferred term or acceptable value.
    • 2 letter codes that denote the shorthand for US states are an example.
      • CT Connecticut, Conn, Connecticut, Constitution State

    Why bother?

    • Authority files can be useful tool for authors and indexers, enabling them to use approved terms efficiently and consistently.
    • The preferred term can serve as the unique identifier for each collection of equivalent terms, allowing for more efficient addition, deletion and modificatoin of variant terms.
    • You can educate users, to misspelling. Or explaining industry terminology or brand names.
    • Its an opportunity to nudge everyone towards speaking the same language, which will help with this interaction and future ones with your company/industry.
    • Preferred terms are also important as the user switches from searching to browsing mode. Think taxonomies, navigation and indexes - we need consistency
    • You can include pointers from one word to another that used in your vocabulary. This is called term rotation.
    Classification schemes: arrangement of preferred terms
    • Many people prefer to use taxonomy instead
      • A FE browsable hierarchy thats visible
      • or a BE tool used by authors and indexers for organising and tagging docs
    The Dewey Decimal Classification (DDC) book classification used in libraries
    • First published in 1876.
    • 000 Computers, information and general reference
    • 100 Philosophy & phycology
    • 200 Religion
    • 300 Social sciences
    • 400 Language
    • 500 Science
    • 600 Technology
    • 700 Arts and recreation
    • 800 Literature
    • 900 History & geography
    • Netflix uses a sophisticated classification scheme to help customers find new movies (Drama, comedy) etc. They also have thousands of micro-genres like 'based on real life' and 'with strong female lead'. Movies are analysed and given a number of micro tags #happyEnding - these can be used to inform the categorisation process
    • Search results can also present things like "Departments" categories, which reinforces users' familiarity with Walmart's classification scheme
    • Classification schemes are not tied to a single view or instance

Thesauri: a controlled vocabulary in which equivalence, hierarchical and associative relationships are identified for purposes of improved retrieval
  • builds upon the constructs of the simpler controlled vocabularies, modelling these three fundamental types of semantic relationships
  • image
  • Each preferred term becomes the centre:
    • Equivalence = manages synonyms
    • Hierarchical = classification into categories and sub categories
    • Associative = meaningful connections that aren't handled by hierarchical or equivalence relationships
Technical lingo:
  • PT = Preferred Term
  • VT = Variant Term (synonyms)
  • BT = Broader Term (parent of preferred)
  • NT = Narrower Term (child of the Preferred Term)
  • RT = Related Term (useful connection, 'see also', not hierarchy or synonym)
  • U = Use (VT use PT) or (VT see PT)
  • UT = Used for (indicated the reciprocal relationship of PT to VT)
  • SN = Scope Note (a specific type of definition of the PT, to restrict the meaning of that term in order to rule out ambiguity as much as possible)
Different types:
  • Classic: Used for indexing and searching.
  • Indexing only: (if you can't add to search)
    • You can build browsable indexes of preferred terms
  • Searching (if indexing isn't practical):
    • If you can't index the documents that you're using
    • Leverages it at the point of searching but not indexing

Getting IA Done

11) Research
  • Do things in this order:
    • Research > Strategy > Design > Implementation > Administration
    • Research:
      Whats included
      • Essentially, reviewing existing background materials, meeting with stakeholders, exploring information ecology.
      • Try to gain an understanding of:
        • Goals and business context
        • The existing information architecture
        • The intended audiences
        • The content
        • Explore information ecology

      Areas and tools:

      Context: business goals, funding, politics, culture, tech, human resources

      Whilst conducting research try to get buy-in for your project. Explain who you are and why you're asking the questions, what is information architecture and why they should care, whats your methodology and how it relates to their work

      Background Research

      Questions:

      • What are the short and longterm goals?
      • Whats the business plan? What are the politics?
      • Whats the schedule and budget?
      • Who are the intended audiences?
      • Why will people come? Why will they come back?
      • What types of tasks should users be able to perform?
      • How will content be created and managed? by whom?
      • What's the technical infrastructure?
      • What worked in the past? What didn't?

      Materials:

      • Documents of the sites vision, mission, goals, intended content and audiences
      Presentations and meetings
      Introduce yourself
      • What is information architecture and why is it important?
      • How will the information architecture relate to the other component of the information environment and to the organisation itself?
      • What are the major milestones and deliverables?
      Strategy questions
      • What are the goals for the system?
      • Who are the intended audiences?
      • What is the planned content and functionality?
      • What channels will people use to access the system?
      • Who will be involved in this effort?
      • When do you need to show results?
      • What obstacles do you anticipate?
      Content questions
      • Formal and informal policies regarding content inclusion?
      • Is there a CMS that handles authoring and publishing?
      • DO those systems use controlled vocabularies?
      • How is content entered and by whom?
      • What technology is being used?
      • What content does each owner handle?
      • What is the purpose of the content? What are the goals and vision behind this content area?
      • Who is the audience?
      • How will the audience access the system?
      • What is the format of the content? Dynamic or static?
      • Who maintains it?
      • What future content or services are planned?
      • Where does content originate?
      • What legal issues impact the content management process?
      Tech questions
      • Can we leverage the CMS?
      • How can we create the infrastructure to support tagging?
      • Does the CMS handle automated categorisation of docs?
      • What about automated index generation?
      • What about personalisation?
      • How flexible is the search engine?
      • Will the search engine support integration of a thesaurus?
      • How so we get regular access to search logs and usage analytics?
      Stakeholder questions
      • What is your role? What do your team do?
      • Competitive advantage of your product?
      • Key challenges?
      • What company wide intiatives should we know about?
      • DO you use the existing product? If not why?
      • How do you access the product?
      • What incentives exist to use the product?
      • Success factors, how to measure them?
      • Top 3 priorities for the intranet redesign?
      • If you could tell the product strategy team one thing, what would that be
      • What questions should we have asked that we didn't?
      Technology assessment
      • Understand whats in place, what you want, do a gap analysis
      • Look at commercially available tools to close the gap
      Content: documents, datatypes, content objects, metadata, volume, existing structure
      • The stuff in your information environment.
      • Information must be found before it can be used.
      • Study your information. What makes one different from the other? How do their structure and metadata impact find-ability?
      Heuristic evaluation
      • Try to learn from the existing environment and structure. What works what doesn't?
      • A heuristic evaluation is an expert critique testing a product against a set of design guidelines. An expert reviews the information architecture, and identifies major problems and opportunities for improvement.
        • Guidelines that they may look at:
          • The environment should provide multiple ways to access the same information
          • Indexes and sitemaps should be emploted to supplement the taxonomy
          • The navigation system should provide users with a sense of context
          • The environment should consistently use language appropriate for the audience
          • Searching and browsing should be integrated and reinforce each other
      Metadata and content analysis
      • Review of the stuff / documents in our system
      • Can be an information survey or a detailed audit
      • Gathering content: Gather a few of each thing:
        • Note the format, document type, source, subject, existing architecture
      • When you stop finding new interesting things, stop looking. Be wary of diminishing returns
      • Analysing Content:
        • Structural metadata: hierarchy
        • Descriptive metadata: topic, audience, format,
        • Administrative metadata: Business context, ownership, creators, removal timespan
      • Questions to ask:
        • What is this object?
        • How can I describe it? for people? for machines?
        • What distinguishes it from others?
        • How can I make this object findable for people and machines?
      Content mapping
      • Visual representation of the existing information environment
      • A tool for understanding, rather than a concrete design deliverable
        • help you understand the structure, organisation, location of content.
      image
      Benchmarking
      • point of reference to make comparative measurements or judgements
      • systematic identification, evaluation and comparison of IA features of the stuff in your information environment.
        • Competitive benchmarking: Borrow what works. Don't borrow whats broken.
          • Pros of the approach:
            • Laundry list of new features
            • Makes conversations more specific and measurable
            • Challenges embedded assumptions
            • Establishes position vs competitors
        • Before-and-after benchmarking:
        • Types of questions
          • Time to find documents
          • Success of finding documents
          • Have we had a negative impact anywhere?
          Pros of the approach
          • Helps prioritise IA features in the existing environment
          • Encourages transition to data and specifics
          • Creates a point of reference against which you can measure improvement
      Users: audiences, tasks, needs, information-seeking behaviour, experience, vocabularies
      • Users are the ultimate judge of the information environment
      • Consider the unique nature of the product and the people that will be using it
      • Its much better to conduct 5 interviews and 5 usability tests, that to run one test 10 times. Each approach is subject to the law of diminishing returns
      Search log and clickstream analysis
      Usage analysis
      • Content performance:
        • The number of visits and interactions with content.
          • What is popular? What is helpful? What isn't?
      • Visitor information:
        • Sources of visitors, countries, capabilities of web browsers etc
      • Path users take through website. How long they spend on each page?
      • What you really need to know is why users came to the site
      Search log analysis
      • What queries are people searching?
      • Get a monthly report
        • Which popular queries are getting 0 results?
          • User error? Looking for stuff that doesn't exist?
        • Which popular queries retrieve too many results?
          • What are these people actually looking for?
        • Which queries are becoming more popular? Less popular?
      Customer support data
      • What confuses people?
      • What generates a lot of support interest?
      Surveys:
      • Which content and tasks users find most valuable?
      • What frustrates users most about the current product?
      • What ideas users have for improvement?
      • The current level of user satisfaction?
      Contextual inquiry
      • Filed study. Seeing the environment in which your product is used.
      • Observe people performing normal tasks and workflows.
      User interviews and user testing
      User interview questions

      Background: role, background, tenure

      Information use: information needs, hardest to find, what do you do when you can't find something?

      Product use: Do you use it? Impression of it? Easy or hard to use? How do you find information on it? DO you use customisation or personalisation features?

      Document publishing: Do you create products that are used by other people or departments? Tell us about the document lifecycle? What tools help you publish them?

      Suggestions: What would you change? What would you add? What would you tell the strategy team? What didn't we ask that we should have?

      Card sorting:
      • Find out peoples mental models. How they group sort and label.
      • Can use for qualitative: why people do what they do (discovery)
      • Can use for quantitative (validation): % 2 cards placed together, % car in category x
        • You can represent relations in an affinity model
      • Start qualitative, then make it quantitative once you have a view
      Open/Closed:
      • Open: Users write own cards and categories.
      • Closed: Cards and categories are pre-labelled
      • Open sorts are used for discovery while closed sorts are used for validation
      Phrasing:
      • Labels can be words, phrases, sentences, categories, sub categories
      • Card labels might be questions or answers., or use the topic / task orientated words.
      • Granularity: High level or detail
      • Heterogeneity: start with apples & oranges, qualitative, then move to quant
      • Cross-listing: If testing navigation, allow duplicates (maybe not for hierarchy)
      • Random: random selection or specific labels for specific tests
      User testing / usability testing
      • Start with easy, end with impossible
      • Known-item to exhaustive (ask them to find a specific thing, and then find everything they can)
      • Topic to task: Find something on this topic. Complete this task
      • Artificial to real: Give them a task to do, based on a hypothetical scenario that you tell them about.
      In defence of research:
      • Learn about business goals, users, the information helps develop a good strategy
      • Create, present and refine the strategy - work towards consensus
      • Ideally have a similar set of people doing research and strategy
      • Overcoming research resistance:
        • We don't have time or money
        • We know what we want
        • We've already done research
      • Support:
        • You're likely to save money doing research
        • Managers don't know what the users want
        • We need to do information architecture research: Unique questions in unique ways, its unlikely you can get all the information you need from stuff thats lying around already
      Strategy:
      Design
      Implementation
      Administration
12) Strategy
  • Research can be addictive, the more you learn, the more questions you have.
  • We don't have that luxury though, we often have to time-box research phases
  • The bridge between research and design is an information architecture strategy
  • The line between research and strategy is blurred. You may have to move forwards and backwards from between research, strategy and design.

What is an information architecture strategy?

  • High level conceptual framework for structuring and organising an information environment.
  • Providing a sense of direction and scope.
  • Get people on the same page before moving into the more expensive design and build phases.
  • High level recommendations for:
    • Information architecture administration: Developing and maintaining the IA. Central or decentralised. Can you trust authors to provide metadata?
    • Technology integration: What tools and tech. Gap analysis. Options
    • Top-down or bottom-up emphasis: Focus on a top-down hierarchy, or bottom up?
    • Organisation and labelling systems: Defining the organisation schemes, identifying the dominant organisation scheme and primary hierarchy .
    • Document identification (bottom up): Suite of object types, and attributes. Collaborate with the authors and management teams
    • Metadata field definition: Descriptive metadata fields, global ones, local ones, particular document types.
    • Navigation system design: How do they leverage the top-down and bottom up. Search zones (within a hierarchy). Categories in search results.
  • Have a view on strategy going into research, then refine and test it.
  • Strategy Development process:
    • Think > articulate > communicate > test
      • Think: convert research data into creative ideas
      • Articulate: diagrams, metaphors, stories, scenarios, blueprints, wireframes
      • Communicate: present, react, brainstorm
      • Test: closed card sorts, prototypes
    • Strategy phase deliverables
      • IA Strategy Report (detailed strategy direction and scope)
      • IA Strategy Presentation (high-level strategy direction and scope)
      • Project Plan for Design (teams, deliverables, schedule, budget)
Work Products and Deliverables
Metaphor Exploration
  • Powerful tool for communicating complex ideas.
  • Suggesting creative relationships, or mapping the familiar onto the new, metaphors can be used to explain, excite and persuade.
  • Gore used the 'information superhighway' simplification
Organisational metaphors:
  • Have a physical presence? Use that as inspiration. At a car dealership you go to new car sales, used car sales, repairs or parts. If you were making a website for this, it would make sense to follow the same model
Functional metaphors:
  • Make a connection between the tasks you can perform in a traditional vs a new environment. When you enter a traditional library, browse the shelves, search the catalog, or ask a librarian for help. Websites may these tasks as options for users, thereby employing a functional metaphor.
Visual metaphors:
  • Familiar graphic elements, images, icons and colours to create a connection to the new elements.
Functional metaphors
Visual metaphors
  • Scenarios can help people understand how users will navigate around the new things and what's different. Who is using the product and how?
  • Case studies and stories can also be used to make your work more accessible. What did and didn't work on past projects
  • Conceptual diagrams can bring the abstract concepts to life. E.g. if building an intranet, show all the other ways the user gets information at work in a bubble diagram
  • Sitemaps and wireframes: bring clarity and precision to a messy process
  • Strategy report:
    • Executive summary
    • Audiences, mission and vision for the site
    • Lessons learned from user research, past, stakeholders
    • Architectural strategies and approaches
    • Content management approach
  • Project Plan
    • How will we accomplish that
    • How long will it take
    • Who will do it
    • What kinds of deliverables will be required
    • What are the dependencies?

13) Design and Documentation
Move to a delivery mindset- focus presenting a well-defined IA
  • Deliverables are useful. Force the team to pause, capture and review work.
  • You can use deliverables to mitigate the risk o an out-of-control project.
Guidelines for Diagramming Information Architecture
  • Pressure to clearly represent the product our our work
  • We rely on visual representations to explain our work
  • IAs are abstract and conceptual. Hard to represent on a whiteboard
  • How to present work visually, that isn't visual?
    • Provide multiple views of an IA (don't try and show everything in 1)
    • Develop views for specific audiences and needs
      • Determine what others need from your diagram, before creating one
      • Upstream diagrams (leadership) and downstream (developers / designers)
  • Whenever possible present diagrams in person
Communicating Visually
  • Diagrams are good at communicating 2 things:
    • Content Components:
      • What is a unit of content. How they are grouped and sequenced
    • Connections between components:
      • How content components are linked to enable actions like navigating between them
  • The goal of the diagram is communicate information or a concept
  • There have been a number of attempts at creating diagrams to explain IA.
  • Jesse James Garrets (is simple and well used) See here.
  • Play to your own communication strengths, do what you're good at

Sitemaps
  • To show relationships between pages and other IA components, show organisation, navigation and labelling systems.
High Level Architecture Sitemaps
  • Start with the main page, you might user the process of developing a sitemap to iteratively flesh out more and more of the architecture, adding subsidiary sections, increasing detail, working out the navigation
  • Shaping ideas into a sitemap keep things realistic and practical
  • Useful for exploring primary organisation structures, mapping out the navigation and labelling of major areas.
  • Good for stimulating discussion on organisation and management of content, as well as on the desired access pathways for users.
  • Use Visio or OmniGraffle. Quickly lay things out and look professional.
  • Sitemap example:
image
Example SiteMap Narration
  • The building block of this architecture is the sub-site. In this organisation, there are lots of small pages owned by different authors in different teams. Don't fight that, design an umbrella around it.
    • Moving up from the sub-sites, we see a directory of sub-site records serving as a card catalog providing easy access to the sites. Record for each sub-site, records have title, author, description, keywords, audience, format and topic. Creating a standardised record, creates a database, enabling powerful known-item searching and exploratory browsing.
    • Search is represented, also are the options to browse by audience, title, format or topic
    • A dynamic news billboard rotates the display of featured news headlines and announcements.
  • The narrative about should show that sitemaps don't speak for themselves
  • Consider augmenting the sitemap with a text doc that explains thinking and answer the most obvious FAQs

Digging deeper into sitemaps
  • Don't get locked into a certain layout, allow form to follow function
  • You can show levels of hierarchy and filters on a sitemap
  • You can focus on content and information, or on tasks
  • You can focus on diagrams that show only a certain aspect of the journey

Keeping Sitemaps Simple:

  • Now we are designing, we need to do all the corners, we need to iterate and move quickly
  • Develop a simple, condensed vocabulary of objects that can be explained in a brief legend.
    • Page, Content Component, Content Group, Link (a sensible amount for a legend)

Detailed sitemaps:

  • The deeper you get into implementation the more that your focus shifts to internal.
  • You may want to show local vs remote pages, you may want to show navigation in a sidebar

Organising sitemaps:

  • Needs to accommodate more than top level pages.
  • Modularize your sitemap. Top level, and subsidiary sitemaps.
Wireframes
  • Sitemaps are for where content should go and how to navigate it
  • Wireframes depict how a page should look, connecting IA to interaction design
  • Wireframes, are constrained in space as per the finished product, so they help us prioritise what to show.
  • Also helps clarify the grouping of content and prominence of functionality
  • Only need for the major pages, or the ones that are complex
  • Include a disclaimer that real visual design is still needed
Types of wireframes
  • Paper
  • Adobe illustrator
  • Different levels of fidelity
  • High fidelity:
    • Content and colour bring it to life, helping capture the attention of the clients
    • simulating actual page with width font and size, helps you think about constraints
    • Greater effort, time and cost
    • You may end up talking about visual design not infromation

architecture

Guidelines for wireframes
  • Consistency is key when presenting multiple wireframes
  • Visio and other charting tools allow you to reuse components saving time
  • Callouts are an effective way to provide details. Leave room for them at the side of your wireframes
  • Like any other deliverable, wireframes should be usable professionally, label them, date them, project titles and revision dates
  • If working in a team, workout how you're going to create a portfolio of these together

Content mapping and inventory
  • Content mapping is where top-down information architecture meets bottom up.
  • Breakdown or combine existing content into chunks
  • Content must be mapped onto the information architecture so that its clear what goes where in the production process.
    • BREAK CONTENT INTO CHUNKS
    • USE A MAPPING TABLE TO MATCH THEM WITH THEIR DESTINATION

Content models, why do they matter?

They are the information architectures made up of small chunks of interconnected content.

A recipe is a good example of a content model.

  • Ingredients list
  • Directions
  • Title

Why do they matter?

  • Supporting contextual navigation
    • Say you're looking for a blue shirt on a clothing website. It makes sense for the retailer to show you things that might go with that shirt, maybe from different categories. Much more reasonable than expecting you to back up to the main page and navigate the taxonomy towards them
  • Coping with large amounts of content
    • Help us deal with scale
    • Automating the creation of links between chunks is scalable, rather than having 1000 site editors.
    • Very valuable when we have a large number of high-value content chunks that are similar to one another and aren't well linked.
Information architecture style guides
  • The why stuff
    • Documenting lessons learned
    • Decisions made
  • The how stuff
    • Standards : stuff that needs to be followed for things to work
    • Guidelines: suggest not mandate what they should look like.
    • Maintenance procedures: When and how to add new terms to vocabulary
    • Pattern library: documents and provides access to reusable aspects of product design, such as a navigation widget.