• Whistlestop tour of IA for the web
• See the wiki for more in-depth information and links
• Going to concentrate on navigation and card sorting

• IA is the organisation, labelling and structuring of information
• Not just a taxonomy - IPSV is not an information architecture
• IA is equally concerned with the application of taxonomies
• E.g. navigation, page layout, search functionality, metadata, tagging etc
• IA enables logical progression through a site to a target destination

• Common definition - IA is the point where the needs of users, the structure of content and the needs of the organisation overlap
• IA is built on business and user requirements, so these need to be established before tackling IA
• Enables users to quickly, easily and intuitively find stuff
• In a complex organisation such as an LA we need to consider different needs of different groups of users
• IA is a subset of usability processes and attributes

• Progress in web development means we cab do just about anything online now
• Previously user frustration was concentrated aroung not being able to do stuff online
• With the wealth of information and services online now that frustration is concentrated around finding things
• Getting IA wrong devalues your entire online enterprise
• For LAs that means driving users to more expensive channels

• Business - strategic objectives
• Content - inventory, future needs
• Users - surveying and testing
• The first two are fact-gathering exercises that can happen internally. The last deserves closer examination.

• Many users come to the table with a goal and a mental model of how to reach it
• The challenge is to reflect and accommodate that mental model, for all users
• Some users think in a distinctly odd way
• Identify your distinct groups of user and adjust accordingly
• Do not mimic organisational structures in your IA - they change and mean little to users
• The best way to identify potentially effective IA is to involve users
• But you must exercise some professional judgement - you can't abdicate responsibility to a bunch of users

• Users don't always know best
• They may not be able to articulate why something isn't right for them
• So don't place too much stock on individual user's opinions
• Try to establish a consensus among similar users

• Testing method for oganising information into structures
• Provide one or more groups of users with cards, on which is printed the label of an item from your content
• Ask the users to sort them into groups that make sense to them
• Open card sorting lets the users name the groups
• Closed card sorting names the groups and asks the users to sort the items into those pre-defined groups
• Cheap and easy to conduct
• Provides useful data on groupings, and suggested categories
• Also identifies potential labelling problems

• Have a clear objective - is it to test a structure, confirm theories, or give an early lead in establishing labels and groups?
• Objective informs open or closed approach - open is most useful early in the IA process, closed is useful as a validation tool
• Maximum number of cards cited as between 30 and 200!
• Topics should be key pages or key functions with sufficient similarity to be easily grouped
• Use card not paper
• Number cards on the back

• Make sure there's plenty of space - cramped card sorts are no fun
• Recruit the right users (or at least know who you've recruited). You need to know something about their level of experience and exposure to your content inventory
• Individuals or ideally groups (5 groups of 3 is cited as an effective setup)
• Record results using video or camera
• Observe and take notes. A lot of the value is in the group discussion and debate.

The items for the exercise are:

1. apple
2. banana
3. blue
4. dip
5. dive
6. dodge
7. dodge
8. duck
9. ford
10. forth
11. green
12. humber
13. ibm
14. microsoft
15. nile
16. oracle
17. orange
18. peach
19. red
20. severn
21. tower
22. vauxhall
23. waterloo
24. yellow

• Our example is obviously contrived to throw up questions.
• For example Forth could refer to a bridge, a river, a brewery and more.
• And only dodgeball (the movie) fans will spot the 5 Ds of Dodgeball.
• But it serves to illustrate the issues that card sorting can throw up.

• Similar category labels provide a guide to category naming
• Merge categories as you see fit (e.g. "Social Care and Health" is seen on a number of LA sites)
• An affinity diagram shows strength of relationship between items based on number of times grouped together (see next slide)
• Item labelling issues will emerge from the discussion that took place in the user groups and from follow-up discussion
• Take account of context when considering equivocal labelling - a label may make sense in the context of its group label, which is fine for global navigation, but if repurposing for an A to Z index you may need to qualify it
• Also consider the knowledge a specialist user may bring to your site when pursuing a specialist goal - it's important to recognise your specialist audiences and their vocabulary

• In many cases we've already got an IA in place and want to know how good it is, and where improvements could be made
• There are various techniques for evaluating an existing IA
• Best results from user testing
• Generally only qualititative results
• Must try to detach IA from visual and stylistic aspects of site
• If the font, colour, or layout is a barrier to the user then no IA will meet their needs

• Heuristics are agreed design principles which can be measured
• E.g. no single page should link to more than 5 related pages
• E.g. no item should appear in more than 2 categories
• Test formally within development team
• Only useful when IA is stable and effective
• Doesn't measure effectiveness of categorisation, only compliance with design principles

• Generally not visible to the user
• But enriches interaction - search, content filtering, contextual information
• Think beyond traditional keywords - dates (publication, modification, review), categories (taxonomy), geography, ownership, etc

• Tagging content with geographic metadata
• Settlement - allows presentation of all content for a settlement, or with further context through other metadata
• E.g. all press releases about Dollar published in October 2006
• Geocoding allow us to go even further
• E.g. provide a map interface to explore events, news, facilities etc. Think beyond planning!

• Tagging content with demographic metadata
• Would provide hooks for true user-centered content delivery
• Targetted emails, alerts etc
• Event-driven information, e.g. benefits and employment services on loss of job
• Requires serious planning!

• The success of Flickr and del.icio.us shows how users are comfortable formalising their mental models
• It works - the consensual tags emerge for a resource
• Would it work in the public sector?
• Hard to see users being engaged in doing the same for government, but then no-one has tried?

• IA is worthwhile investing in, because it impacts on the effectiveness of your site
• You must involve users when developing and evaluating an IA
• But you must know something about those users too
• Don't get boxed in to rigid taxonomies - use IA to your advantage by developing metadata that supports the business and its users