Content-Centric Design The Future of Online User Assistance

1
Content-Centric DesignThe Future of Online User
Assistance

• Presented by
• Edward Galore, University of Washington /
  Microsoft
• David Farkas, University of Washington
• Janet Galore, Microsoft
• Mark Huentelman, Microsoft
• Bruce Keever, Microsoft
• Zherina Salamanca, University of Washington

2
A contemporary help scenario

• User searches for invitation template
• User is looking for a way to create an invitation
  using Word

3
This isnt quite right...
But where to go next?
4
What well cover

• The problem is discoverability
• Traditional models of user assistance inhibit
  discoverability for large online help systems
• Examples from Office Online
• Content-centric design
• Applications
• Conclusions

the problem office online content-centric
design applications conclusions
5
The problem
6
The problem is discoverability

• The trick is getting the right content to the
  right user at the right time, even when they
  dont know the right terms
• Site hierarchies are designed based on an
  educated guess as to where the user will go
• Hierarchies dont adapt to new content types and
  unexpected user behavior
• We have good intentions and excellent content
• Every new feature requires new help content and
  each help article dilutes discoverability

the problem office online content-centric
design applications conclusions
7
Coming in deep
Users do not always come in through the front
door via a Web sites homepage
Users often come in deep and are lost without
the context provided on a homepage they may never
see
the problem office online content-centric
design applications conclusions
8
Example external referrals to Office Online

• Over a 3-month period
• 99M referrals (75) were to the top 5 home pages
• 24M referrals came to other pages and content
• 6,502 different entry points

Number of visits
Entry pages
the problem office online content-centric
design applications conclusions
9
Invisible hierarchies

• Web sites are structured according to a designer
  imposed hierarchy
• Search technologies obviate hierarchy and
  decontextualize content
• Site architecture becomes largely invisible to
  site users

the problem office online content-centric
design applications conclusions
10
Limits of traditional help system Information
Architecture

• Traditional help system IA
• Is modeled on a hardcopy table of contents
• Tends towards linearity
• Is inadequate for large, complex products
• Ignores realities of continuous publishing and
  user provided content and user tagging

the problem office online content-centric
design applications conclusions
11
Content needs context

• Search ignores hierarchy
• In the absence of an explicit hierarchy, other
  methods must be employed to provide context
• Encapsulating context within content is
  content-centric design

the problem office online content-centric
design applications conclusions
12
Content-centric design

• The primacy of Search is recognized and supported
• Hierarchy becomes just another piece of metadata
• Empowers users by allowing them to validate
  content and forge new paths among content

the problem office online content-centric
design applications conclusions
13
Microsoft OFFICE ONLINE
14
What is Office Online?

• Office Online is the broad customer connection
  for Office. Its where Office customers go to get
  help, software updates, training, clip art,
  templates, services, and more
• Goals
• Increase customer satisfaction
• Increase customer productivity
• Increase revenue through upgrades, sales, and
  advertising

office.microsoft.com
the problem office online content-centric
design applications conclusions
15
How big is Office Online?

• Over 70M unique users per month worldwide
• In the top 30 sites on the Web (US unique users)
• 48 markets, 34 languages
• Over 800M page views per month worldwide
• Supports over 25 Office branded products
• Over 150,000 clips, 1.2M help articles, 4300
  templates, 150 training courses, 87 homepages,
  and more
• Over 250 writers, editors, site managers,
  production engineers, localization engineers

the problem office online content-centric
design applications conclusions
16
Who uses Office Online?
Globally, the majority of Office Online users are
from business segments.
Source Office Online worldwide customer
satisfaction survey, 2005
the problem office online content-centric
design applications conclusions
17
Continuous publishing on Office Online
the problem office online content-centric
design applications conclusions
18
Customer ratings and comments on articles
the problem office online content-centric
design applications conclusions
19
Context mismatch near miss

• Users often almost find the content they are
  looking for
• Near misses represent one of 5 types of context
  mismatches, right idea but
• Wrong application/version
• Wrong user intent
• Wrong user type
• Wrong terminology
• Wrong language

the problem office online content-centric
design applications conclusions
20
Context mismatch wrong application

• A topic may satisfy the users intent, but may
  not match the application about which the user
  has a question
• Scenario
• User searches all Office Online to find how to
  make an invitation
• A topic for MS Publisher comes up first, along
  with some templates in Word
• User clicks on first topic, not thinking about
  the application

the problem office online content-centric
design applications conclusions
21
Context mismatch wrong application

• The user does not notice the visual cues about
  what application the topic applies to
• There are no links to find similar content for a
  different application
• The user clicks on No and writes a verbatim
  comment

the problem office online content-centric
design applications conclusions
22
Context mismatch wrong application
The writer examines customer verbatim feedback on
that topic to dig deeper
Many users are not looking for Publisher help
the problem office online content-centric
design applications conclusions
23
Context mismatch wrong application

• Possible remedies
• Create similar topics for other applications
• Create a general invitation topic that talks
  about multiple applications
• Create search best bets for this query
• Include links to similar topics in the see also
  area of the Publisher topic

the problem office online content-centric
design applications conclusions
24
Context mismatch user intent

• A topic may look like the answer the user is
  looking for, but the topic does not satisfy the
  users intent
• Symptom Frequently failed search query
  signature led to deep analysis of existing
  signature topics

the problem office online content-centric
design applications conclusions
25
Context mismatch user intent

• Scenario
• User searches on digital signature
• User intent A visual representation of their
  signature in a document
• Top search results
• Digitally sign a file
• Insert a picture
• Create a signature for messages
• User action Clicks on first result but finds
  its not what they want
• Marks the topic as No and leaves negative
  comments

the problem office online content-centric
design applications conclusions
26
Context mismatch user intent
Selected verbatim comments reveal some users
intent
the problem office online content-centric
design applications conclusions
27
Context mismatch user intent

• Possible remedies
• Modify the insert picture topic so that it also
  mentions that inserting a picture can be used for
  adding a visual signature
• Add a link to the insert picture topic from the
  digital signature topic

the problem office online content-centric
design applications conclusions
28
Context mismatch user type

• One writing style or approach does not work for
  all users
• Office Online serves primarily information
  workers, but all audiences come to the site
• There are different content types for different
  purposes and learning styles (help topics,
  training courses, videos, etc.)
• Investigate if users might identify themselves in
  their verbatim comments, and if there is user
  mismatch happening

the problem office online content-centric
design applications conclusions
29
Just a few types of users...
access expert administrative assistant amateur asp
iring web page designer audio learner baby
boomer beginner user big computer
freak biochemistry instructor
freelance editor gas service engineer genius good
VB programmer grandmother graphic artist hair
dresser hands on learner
disabled user disabled veteran doctor dyslexic
experienced user faculty member first time user
knuckle head lawyer legal secretary librarian lice
nsed minister luddite Mac person mainframe
programmer manual laborer
slow learner small business owner social
worker software test engineer special education
student student system administrator
teacher total beginner
blind computer user branch chief brand new
FrontPage 2003 user brand new user business
owner case manager certified applications
instructor chef
the problem office online content-centric
design applications conclusions
30
Context mismatch user type

• Example comments

the problem office online content-centric
design applications conclusions
31
Context mismatch user type

• Possible remedies
• Consider tagging content by audience type
• Have clear destinations for different audiences

the problem office online content-centric
design applications conclusions
32
Content-centric design
33
Apples and oranges

• When developing content taxonomies, designers
  create buckets, categories to store content
• Apples go in the apple bucket, oranges go in the
  orange bucket
• Great! A time and place for everything under
  heaven

the problem office online content-centric
design applications conclusions
34
Bananas

• Expect the unexpected
• New, unanticipated content types are constantly
  being generated
• We need dynamic site architecture to accommodate
  dynamic help content

the problem office online content-centric
design applications conclusions
35
What are bananas, anyway?
If only user assistance content were as orderly
the problem office online content-centric
design applications conclusions
36
Three stages of design philosophy

• Rational design (designer centered)
• User centered
• Content-centric

the problem office online content-centric
design applications conclusions
37
Rational design

• Designer knows best
• Rational, not empirical, design methods
• Design optimized for the system rather than the
  userusability is secondary
• Finite pathways

the problem office online content-centric
design applications conclusions
38
User centered design

• User is incorporated into the design
• Task oriented
• Benefits from empirical studies of user behavior
• Design for typical users

the problem office online content-centric
design applications conclusions
39
Limitations of user centered design

• Task or transaction oriented
• Assumes finite number of user actions
• Doesnt necessarily account for dynamic content
• Employs a top-down approach based on anticipated
  user behavior

the problem office online content-centric
design applications conclusions
40
Content-centric design

• The content-centric approach does not imply that
  UCD is no longer valid
• On the contrary, it assumes that user input is
  invaluable, so much so that content should adapt
  to user needs dynamically
• Web sites are forever changing new content is
  being added continuously
• Every new piece of content challenges the
  existing structurecontent-centric design is
  adaptive to users and content

the problem office online content-centric
design applications conclusions
41
Applying content-centric design
42
Applying content-centric design

• How might content-centric design improve the
  experience for a user wanting to edit a comment
  in MS Word?
• Relate content and provide context
  non-hierarchically
• Aggregate and incorporate user feedback
  dynamically
• Provide multiple contexts for See Also
  Near missesincluding visual context

the problem office online content-centric
design applications conclusions
43
Current UI hierarchical breadcrumbs

• Breadcrumbs give the user context in terms of
  where they are in the systems information
  architecture
• But breadcrumbs reflect the designersnot the
  usersperspective of the site

the problem office online content-centric
design applications conclusions
44
Current UI hierarchical breadcrumbs
Home gt Products gt Excel gt Excel 2003 Help and
How-to gt Sharing Information gt With Other People
gt Comments
Home gt Products gt Excel gt Excel 2003 Help and
How-to gt Sharing Information gt With Other People
gt Comments
Home gt Products gt Excel gt Excel 2003 Help and
How-to gt Sharing Information gt With Other People
gt Comments
Home gt Products gt Excel gt Excel 2003 Help and
How-to gt Sharing Information gt With Other People
gt Comments
Home gt Products gt Excel gt Excel 2003 Help and
How-to gt Sharing Information gt With Other People
gt Comments
Home gt Products gt Excel gt Excel 2003 Help and
How-to gt Sharing Information gt With Other People
gt Comments
Home gt Products gt Excel gt Excel 2003 Help and
How-to gt Sharing Information gt With Other People
gt Comments

• Context provided via typical breadcrumbs only
  shows systems linear, hierarchical path to the
  article
• Do not show the users actual path
• In order for users to see where theyve been,
  they must use the back button
• Users do this because on the Web, as in real
  life, a sound way to discover where you are is to
  see where youve been

the problem office online content-centric
design applications conclusions
45
Content-centric UI My Path

• Pathways users take to article are unique and
  reflected in UI as My Path
• User can trace their navigation pattern without
  using the Back button

the problem office online content-centric
design applications conclusions
46
My Path real breadcrumbs
My Path reflects the users actual,
non-hierarchical path to the article.
My Path Home gt Search All Office Online edit a
comment gt Edit a comment (Excel 2003) gt Modify a
comment (Word 2003)
the problem office online content-centric
design applications conclusions
47
Current UI Article rating

• Users can give input as to whether the article
  was what they were looking for or not

the problem office online content-centric
design applications conclusions
48
Content-centric UI User suggested articles

• Related article suggestions can dynamically
  populate according to user behavior and feedback

the problem office online content-centric
design applications conclusions
49
Current UI See Also

• See Also boxes appear in some articles, providing
  suggestions for related content

the problem office online content-centric
design applications conclusions
50
Content-centric UI More related content

• Two new ways to show related content
• Related Search Results provide limited context
  through text
• A Visual Map can use graphics to help the user
  visualize how content is related

51
Visual map vector based navigation

• Tags for the topic are shown as vectors to
  additional content
• Tag vectors could be color coded
• Vectors could reflect customer provided tags,
  aggregated user behavior, or categories as
  assigned by content authors, or all three

the problem office online content-centric
design applications conclusions
52

• This Slide Intentionally Left Blank

53
Conclusions
54
Conclusions

• Users dont always use the front door, prepare
  for their arrival by other means
• If you want content to be discoverable
• Make search work
• Make it easier to get to content from a related
  topic
• Dont create the perfect taxonomy, allow it to
  evolve

the problem content-centric design office
online applications conclusions
55
Conclusions

• In the absence of an explicit hierarchy, other
  methods must be employed to provide context
• Hierarchies are just another type of metadata,
  not always a meaningful navigation tool for users
• When content authors and users add new content,
  they also indirectly affect information
  architecture

the problem content-centric design office
online applications conclusions
56
Edward Galoreedward_at_washington.edu