Enterprise Information Architecture: Because users don't care about your org chart

Enterprise Information Architecture
Because Users Don’t Care About Your Org Chart

March 30, 2010
Merit Network

Louis Rosenfeld
www.louisrosenfeld.com

©2010 Louis Rosenfeld LLC (www.louisrosenfeld.com). All rights reserved. 1

About Me

Independent IA consultant and blogger
(www.louisrosenfeld.com)
Founder, Rosenfeld Media, UX publishing
house (www.rosenfeldmedia.com)
Work primarily with Fortune 500s and other large
enterprises
Co-author, Information Architecture for the World
Wide Web (1998, 2002, 2006)
Founder and past director, the Information
Architecture Institute (www.iainstitute.org) and User
Experience Network (www.uxnet.org)
Background in librarianship/information science


Seminar Agenda

Welcome/Introduction
Topic: Top-Down Navigation
Break
Topic: Bottom-Up Navigation (content modeling)
Exercise #1: Metadata
Topic: Bottom-Up Navigation (metadata)
Lunch
Topic: Search
Exercise #2: Search Analytics
Break
Topic: Research Methods
Topic: Governance and Organizational Change


Introduction


Introduction:
IA in one slide
Definition: the art and science of
structuring, organizing and labeling
information to help people find and
manage information
•  Balances characteristics
and needs of users,
content and context
•  Top down (questions)
& bottom up (answers)


Introduction:
Only one IA rule
Pareto’s Principle (“the 80/20 rule”)
•  20% of content satisfies 80% of users’
needs
•  20% of possible IA options address 80% of
content
•  20% of IA options address 80% of users’
needs
IA’s goal: figure out which 20%
No other rules, just guidelines


Introduction:
IA is about priorities


What an Enterprise Is

Large, distributed, decentralized organization
made up of multiple business units
Distributed
•  Functionally in many different “businesses” (e.g.,
HR vs. communications, or hardware vs. software)
•  Geographically
Decentralized
•  Large degree of authority and responsibility
resides in hands of business units in practice (if
not officially)
•  Business units often own significant infrastructure
(technical, staff, expertise)


IA and EIA:
The differences
The “enterprise challenge”: providing
centralized access to information in a
large, decentralized, distributed
environment
Information often organized by business
function (e.g., “org chart”), not in ways
users think
Not “textbook” IA; highly dependent on
business context


The Challenge of EIA:
Competing trends
Trend toward autonomy
•  Cheap, easy-to-use democratizing technology
•  Human tendency toward autonomy
Trend toward centralization
•  Users’ desire for single-point of access
•  Management’s desire to control costs and
communications
These tend to cancel each other out, getting us
nowhere
Result: content “silos” and user confusion


Indicators of Problematic EIA:
Intranet glitches
“How come I didn’t know your department
was developing a product similar to
ours?”
“Why couldn’t we find any relevant case
studies to show that important
prospect?”
“Why do our sales and support staff keep
giving our customers inconsistent
information?”


Indicators of Problematic EIA:
External-facing site glitches
“Our customers think we’re still in the
widget business; after all these M&As,
why don’t they realize that we’ve
diversified?”
“We have so many great products that go
together; why don’t we cross-sell more?”
“Customers keep asking for product
support through our sales channel; why
don’t they use the site’s FAQs and tech
support content?”

The Holy Grail:
Cutting against the political grain


Example: Expense Reporting


So How Do We Get There?

Let it go
•  There is no single solution
•  Redemption lies within phased, modular, evolving
approaches that respect 80/20 rule
Your friends
•  Straw men
•  Your colleagues and professional networks
This seminar provides straw men for
•  EIA design
•  EIA methods
•  EIA team design and governance


Top-Down
Navigation


Top-Down Navigation Roadmap
Main page

Site hierarchy

Site map

Site index

Selective navigation


Top-Down Challenges

Top-down IA
•  Anticipates questions that users arrive with
•  Provides overview of content, entry points
to major navigational approaches
Issues
•  What do we do about main pages?
•  Portals: the answer?
•  Other ways to navigate from the top down
•  The dangers of taxonomies


Top-Down Evolution:
Univ. Michigan example 1/2


Top-Down Evolution:
Univ. Michigan example 2/2


Portal Solutions:
Why they fail 1/2
Organizational challenges
•  Fixation on cosmetic, political
•  Inability to enforce style guide changes, portal
adoption
•  Lack of ownership of centralizing initiatives, or
ownership in wrong hands (usually IT)
Information architecture challenges
•  Taxonomy design required for successful portal
tool implementation
•  Always harder than people imagine
•  Taxonomies break down as they get closer to local
content (domains become specialized)


Portal Solutions:
Why they fail 2/2
Challenges for users
•  Portals are shallow (only one or two levels deep)
•  Poor interface design
•  Users don’t typically personalize

More in James Robertson’s “Taking a business-
centric approach to portals” (http://
www.steptwo.com.au/papers/
kmc_businessportals/index.html)


Top-Down Navigation:
Design approaches
Main pages
Supplementary navigation
•  Tables of contents
•  Site indices
•  Guide pages
Taxonomies for browsing
•  Varieties: product, business function,
topical
•  Topic pages


Main pages
Often 80% of discussion of EIA dedicated to
main page
•  Important real estate
•  But there are other important areas
•  Navigational pages
•  Search interface
•  Search results
•  Page design (templates, contextual navigation)
Divert attention from main pages by creating
alternatives, new real estate: supplementary
navigation


Supplementary navigation
Examples
•  Site maps/TOC
•  Site indices
Benefits:
•  Create new real estate
•  Can evolve and drive evolution from org-chart
centered design to user-centered design
•  Relatively low cost to initially implement
Drawbacks:
•  Often unwieldy for largest enterprises (not at IBM,
Microsoft, failure at Vanguard)


Site maps
Condensed versions of site hierarchy
•  Hierarchical list of terms and links
•  Primarily used for site orientation
•  Indirectly cut across subsites by presenting multi-
departmental content in one place
•  But still usually reflects org chart
Alternative plan
•  Use site map as test bed for migration to user-
centric design
•  Apply card sorting exercises on second and third
level nodes
•  Result may cut across organizational boundaries


Site Map:
Visually


Site Map: State of Nebraska

Majority of
links reflect
org chart


Site Map: State of Kentucky

Evolving
toward
more user-
centered,
topical
approach


Site indices
Flat (or nearly flat) alpha list of terms and links
Benefits
•  Support orientation and known-item searching
•  Alternative “flattened” view of content
•  Can unify content across subsites
Drawbacks
•  Require significant expertise, maintenance
•  May not be worth the effort if table of contents and
search are already available
Specialized indices may be preferable (shorter,
narrower domain, focused audience)


Site Index:
Visually


Site Index:
Am. Society of Indexers example
Full site index
•  @1000
entries for
smallish site
•  Too large to
easily browse
•  Replace with
search?


Specialized Site Index:
CDC example
Not a full site
index
Focuses on health
topics
•  Narrow domain
•  Specialized
terminology
•  Possibly still too
large to browse


Specialized Site Index:
PeopleSoft example
Product focus
•  A large
undertaking
at
PeopleSoft
•  High value
to users


“Mature” Site Index:
Informed by search analytics


Guides
Single page containing selective set of important links
embedded in narrative text
Address important, common user needs
•  Highlight content for a specific audience
•  Highlight content on a specific topic
•  Explain how to complete a process
•  Can work as FAQs (and FAQs can serve as interface to
guides)
Benefits
•  Technically easy to create (single HTML page)
•  Cut across departmental subsites
•  Gap fillers; complement comprehensive methods of
navigation and search
•  Can be timely (e.g., news-oriented guides, seasonal guides)
•  Minimize political headaches by creating new real estate

Guides:
Visually


Guides:
Vanguard example 1/2


Guides:
Vanguard example 2/2


Guides:
IBM example


Topic Pages
“Selective taxonomy improvement”
•  Portions of a taxonomy that expand
beyond navigational value
•  Help knit together enterprise content
deeper down in taxonomy
New “real estate” can be used by
•  Individual business units (to reduce
pressure on main page) or…
•  Cross-departmental initiatives


Topic Pages:
CDC example

Subtopics
now
comprise
only a small
portion of
page


Taxonomies & portals
Can a single taxonomy unify an enterprise site?
•  First: can one be built at all?
•  Software tools don’t solve problems (see
metadata discussion)
Approaches
•  Multiple taxonomies that each cover a broad
swath of enterprise content: audience, subject,
task/process, etc.
•  “Two-step” approach:
1.  Build shallow, broad taxonomy that will answer “where
will I find the information I need?”
2.  Rely on subsite taxonomies to answer “where in this
area will I find the information I need?”


Impacts on the enterprise
Potential of “small steps” around which to build
more centralized enterprise efforts
•  Site map and site index creation and maintenance
•  Guide and topic page creation and maintenance
•  Large editorial role, minimal technical
requirements for both
May be preferable to tackle more ambitious
areas much later
•  Developing and maintaining top-level taxonomy
•  Connecting high-level and low-level taxonomies


Top-Down Navigation Roadmap
Main page

Site hierarchy

Site map

Site index

Selective navigation


Top-Down Navigation Takeaways

Main pages and portals: Bypass for now, add guides
over time
Site hierarchy/taxonomy: Start shallow,
"simple" (e.g., products); add progressively harder
taxonomies (work toward faceted approach)
Site map/ToC: Use as a staging ground for a more
topical approach
Site index: Move from generalized to specialized
around a single topic, or augment with frequent
search queries/best bets work
Guides: Start with a handful, then expand and rotate
based on seasonality or other criteria of relevance


Bottom-Up
Navigation


Bottom-Up Navigation Roadmap

Content modeling

Metadata development

Metadata tagging


Bottom-Up Navigation:
The basics
Focuses on extracting answers from
content
•  How do I find my way through this content?
•  Where can I go from here?
Goals
•  Answers “rise to the surface”
•  Leverage CMS for reuse and syndication of
content across sites and platforms
•  Improve contextual navigation
•  Increase the effectiveness of search

Content Modeling:
The heart of bottom-up navigation
Content models
•  Used to convey meaning within select,
high-value content areas
•  Accommodate inter-connectedness
Same as data or object modeling?
Absolutely not!
•  Many distinctions between data and semi-
structured text
•  Text makes up majority of enterprise sites


Content Modeling:
The basics
Based on patterns revealed during
content inventory and analysis
What makes up a content model?
1.  Content objects
2.  Metadata (attributes and values)
3.  Contextual links
Applies to multiple levels of granularity
•  Content objects
•  Individual documents


Content Modeling:
We’re already doing it at page level
album page = title/artist/release + tracks + cover image


Content Modeling:
Content analysis reveals patterns
album pages artist bios

artist descriptions
album reviews


Content Modeling:
Answer some questions
What contextual navigation
should exist between
album pages artist bios these content objects?
(see Instone’s
“Navigation Stress
Test”--http://user-
artist descriptions album reviews experience.org/uefiles/
navstress/ )
Are there missing content
objects?
Can we connect objects
automatically?


Content Modeling:
Fleshing out the model concert calendar

album pages artist descriptions
TV listings

album reviews discography artist bios


Content Modeling:
Connecting with metadata, rules
Content …link to other Content Objects… …by leveraging common
Objects… Metadata Attributes
album page album review, discography, artist Album Name, Artist Name,
Label, Release Date…
album review album page Album Name, Artist Name,
Review Author, Source,
Pub Date…
discography album review, artist description Artist Name, Album Name,
Release Date…
artist artist bio, discography, concert Artist Name, Desc Author,
description calendar, TV listing Desc Date…
artist bio artist description Artist Name, Individual
Artist Name…
concert artist description Artist Name, Tour, Venue,
calendar Date, Time…
TV listing artist description Artist Name, Channel,
Date, Time…


Content Modeling:
Problematic borders concert calendar

album pages artist descriptions
TV listings

album reviews discography artist bios


Content Modeling:
When to use
Use only for high value content
High value content attributes based on users,
content, context, including
•  High volume
•  Highly dynamic
•  Consistent structure
•  Available metadata
•  Available content management infrastructure
•  Willing content owners
Much content can and will remain outside
formal content models

Content Modeling:
Steps for developing a model
1.  Determine key audiences (who’s using it?)
2.  Perform content inventory and analysis
(what do we have?)
3.  Determine document and object types (what
are the objects?)
4.  Determine metadata classes (what are the
objects about?)
5.  Determine contextual linking rules (where do
the objects lead us to next?)


Content Modeling:
Content object types 1/2
List known object types
For each audience:
•  Are there types that don’t fit?
•  Examples: company executive bios, Q&A
columns
•  Venue reviews may be part of a separate
content model


Content Modeling:
Content object types 2/2
For each audience (continued):
•  Gap analysis: are there types missing that
users might expect?
•  Examples: Gig reviews, Buy the CD, Links to
music in the same genre
•  Which types are most important to each
audience?
•  Fans of the band: Interviews with the band
members
•  Casual listener: Samples of the CD tracks


Content Modeling:
Metadata 1/2
Determine which objects would benefit
from metadata
Develop three types of metadata
•  Descriptive
•  Intrinsic
•  Administrative


Content Modeling:
Metadata 2/2
Aim to balance utility and cost
•  Answer most important questions: who,
what, where, why, when, how?
•  Cost-benefit analysis
•  Development and maintenance costs of
controlled vocabularies/thesauri
•  Ability of in-house staff to apply properly


Content Modeling:
Contextual linking rules
Are there specific objects for which these
questions arise again and again?
•  Where would I go from here?
•  What would I want to do next?
•  How would I learn more?
You have a rule if
•  The questions apply consistently
•  The answers work consistently
•  Metadata can be leveraged to connect questions
and answers
Unidirectional links or bidirectional?


Content Modeling:
Content models are a means for tying together
content across business unit boundaries
Content modeling is modular; over time, content
models can be connected across the
enterprise
Major benefits to users who get beyond main
page
Can help justify CMS investments
Not all content areas and owners are
appropriate to work with


Content Modeling:
Putting it all together


CMS Selection:
EIA needs
Support metadata management
(Interwoven)
Support shared metadata workflow
•  Author creation/submission/tagging
(distributed)
•  Editorial tagging (centralized)
•  Editorial review (centralized)
Ability to support contextual linking logic


Metadata:
What is metadata?
Data about data
Information which describes a document,
a file or a CD
Common metadata
•  CD information: title, composer, artist, date
•  MS Word document properties: time last
saved, company, author


Metadata:
Three types
1.  Intrinsic: metadata that an object holds
about itself (e.g., file name or size)
2.  Descriptive: metadata that describes
the object (e.g., subject, title, or
audience)
3.  Administrative: metadata used to
manage the object (e.g., time last
saved, review date, owner)


Metadata:
Common sources
Vocabularies from other parts of your
organization (e.g., research library)
Competitors
Commercial sources (see
www.taxonomywarehouse.com)
Your site’s users
•  Search analytics
•  Folksonomies
•  User studies (e.g., free listing, card sorting)


Metadata:
Value for the Enterprise 1/2
Search: cluster or filter the search by
metadata, like title or keyword
Browse: create topical indexes by
aggregating pages with the same
metadata
Personalization and customization: show
content to an employee based on their
role or position in the company, e.g.
engineer or manager


Metadata:
Value for the Enterprise 2/2
Contextual linking: create relationships
between individual or classes of content
objects (e.g., cross-marketing on
llbean.com)

The purpose is to connect
•  Content to content
•  Users to content
To provide value, metadata requires
consistency (structural and semantic)

Metadata:
Enterprise
big picture


Metadata: Scaling problems
Barriers to
enterprise
metadata
development:
•  Volume of
metadata
vocabs./silos
•  Complexity
of semantic
relationships
(beyond
synonyms)


Metadata attributes:
Easy to difficult 1/2
Level of Metadata Comments
Difficulty Attribute
Easy Business unit These are typically already
names available and standardized
Easy to Chronology Variations in formats (e.g.,
Moderate 12/31/07 versus 31/12/07) usually
can be addressed by software
Moderate to Place names Although many standards exist
Difficult (e.g., state abbreviations and
postal codes), many enterprises
(and their business units) use
custom terms for regions (such as
sales territories)


Moderate to Product Product granularity can vary
Difficult names greatly; marketing may think in
terms of product families; sales in
terms of items with SKU numbers,
and support in terms of product
parts that can be sold individually
Difficult Audiences Audiences, such as customers or
types of employees, vary widely
from unit to unit
Difficult Topics The most ambiguous type of
metadata; difficult for individuals,
much less business units, to come
to agreement on topical metadata

Metadata:
Structural consistency
Standard formats and approaches enable
interoperability, which enables sharing of
metadata.
Examples
•  RDF (Resource Description Format)
•  Topic Maps
•  Dublin Core
•  OAI (Open Archives Initiative)
Sources
•  Academia/scholarly publishing world
•  Little from data management world


Metadata:
RDF (Resource Description Format)
A syntax for expressing semantic
relationships
Basic components
1.  Resource 3. Value
2.  Property type 4. Property

2
1
3
4
From Andy Powell: http://www.ukoln.ac.uk/metadata/presentations/ukolug98/paper/intro.html

Metadata:
Topic Maps
Potential syntax for content modeling, semantic webs
Most simply, made up of associations
topics (e.g., “Lucca”,
“Italy”), occurrences
(e.g., “map”, “book”),
and associations (e.g., topics
“…is in…”, “…written
by…”)
Source: Tao of Topic
Maps, Steve Pepper
(http://www.ontopia.net/
topicmaps/materials/
tao.html)
occurrences

Metadata:
The Dublin Core
A schema for expressing semantic relationships
Can use HTML or RDF syntax
Useful tool (or model) for creating document
surrogates (e.g., Best Bet records)
A standard, but not a religious one
•  Selecting fewer attributes may be a necessity in
enterprise environment
•  Attribute review can be useful as an enterprise-
wide exercise


Metadata:
Dublin Core elements 1/2
Title: A name given to the resource
Creator: An entity primarily responsible for making the
content of the resource
Subject: A topic of the content of the resource
Description: An account of the content of the resource
Publisher: An entity responsible for making the resource
available
Contributor: An entity responsible for making
contributions to the content of the resource
Date: A date of an event in the lifecycle of the resource


Metadata:
Dublin Core elements 2/2
Type: The nature or genre of the content of the resource
Format: The physical or digital manifestation of the resource
Identifier: An unambiguous reference to the resource within a
given context
Source: A Reference to a resource from which the present
resource is derived
Language: A language of the intellectual content of the
resource
Relation: A reference to a related resource
Coverage: The extent or scope of the content of the resource
Rights: Information about rights held in and over the resource

Metadata:
Dublin Core in HTML

Dublin Core elements identified with “DC”
prefix


Metadata:
Dublin Core and RDF

Syntax and schema combination is useful
But where are the metadata values?


Metadata:
OAI and metadata harvesting
OAI: Open Archives Initiative
•  Comes from academic publishing world
•  Provides means for central registration of
“confederate repositories”
•  Repositories use Dublin Core; requests between
service and data providers via HTTP; replies
(results) encoded in XML
Metadata harvesting
•  Enables improved searching across compliant
distributed repositories
•  Does not address semantic merging of metadata
(i.e., vocabulary control)


Metadata:
Semantic consistency 1/2
Provided through controlled vocabularies.
What is a controlled vocabulary?
•  A list of preferred and variant terms
•  A subset of natural language
Why control vocabulary?
•  Language is Ambiguous
•  Synonyms, homonyms, antonyms,
contronyms, etc. (e.g., truck, lorry, semi,
pickup, UTE)


Metadata:
Semantic consistency 2/2

Users Example
Personal Digital Assistant

Synonyms
Handheld Computer
"Alternate" Spellings
Persenal Digitel Asistent
Abbreviations / Acronyms
PDA
Communication Chasm Broader Terms
Wireless, Computers
Narrower Terms
PalmPilot, PocketPC
Related Terms
WindowsCE, Cell Phones

Documents and Applications
Control vocabulary…so your
users don’t have to!

Metadata:
Semantic relationships
Three types
1.  Equivalence: Variant terms with same
meaning (e.g., abbreviations and
synonyms)
2.  Hierarchical: Broader term, narrower
term relationships
3.  Associative: Related terms that are
related to each other


Metadata:
Levels of control

(Vocabularies)

Synonym Authority Classification
Thesauri
Rings Files Schemes

Simple Complex

Equivalence Hierarchical Associative
(Relationships)


Metadata semantic relationships:
Hard to hardest

Level of Type of Examples
Difficulty Relationship
Hard Synonymous Synonym rings and
authority lists
Harder Hierarchical Classification
schemes
Hardest Associative Thesauri


Metadata:
Synonym rings
Used in many search engines to expand
the number of results
Words that are similar to each other are
linked together
Example for a multinational company
•  Annual leave (Australia), the holidays (US),
public holidays (Australia, US), vacation
(US), bank holidays (UK), holiday
(Australia and UK), personal leave (all)


Metadata:
Authority files
Pick list of the authorized words to use in
a field
Can have some equivalence relationships
Example using authors
•  Poe, Edgar Allan--USE FOR Poe, E.A.
•  Poe, E.A.--USE Poe, Edgar Allan


Metadata:
Classification schemes
Classification
•  Systematic arrangement of knowledge, usually
hierarchical
•  Placement of objects into a scheme which makes
sense to the user and relates them to other
objects
Two types of classification schemes
•  Enumerative classification: hierarchical
organization into which objects are placed
•  Faceted classification: organization by facets or
attributes that describe the object


Metadata:
Enumerative classification
Really good to classify small numbers of objects
or objects that can live in only one place
Provides good browsing structure
Can be polyhierarchical, where objects live in
many places
Best known: the taxonomy of life, Dewey
Decimal Classification, Library of Congress
Classification
Most familiar on the Web: Yahoo!, Open
Directory


Metadata:
Enumerative classification example


Metadata:
Faceted classification 1/2
Describes the object with numerous
facets or attributes
Each facet could have a separate
controlled vocabulary of its own
Can mix and match the facets to create a
browsing structure
Easier to manage the controlled
vocabularies


Metadata:
Faceted classification 2/2
Facets for a roast chicken recipe
•  Preparation: Roast / bake
•  Main ingredient: Chicken
•  Course: Main dish
Drawbacks of faceted classification
•  Too many facets attached to an object can
make indexing hard to do
•  Browsing facets may not be as clear as
browsing a hierarchy; many paths to the
same object

Metadata:
Faceted classification example


Metadata:
What is a thesaurus?
Traditional use
•  Dictionary of synonyms (Roget’s)
•  From one word to many words
Information retrieval context
•  A controlled vocabulary in which
equivalence, hierarchical, and associative
relationships are identified for purposes of
improved retrieval
•  From many words to one word


Metadata:
Thesaurus entry example


Enterprise Metadata:
Challenges
Two barriers to enterprise metadata
1.  Interoperability (structural)
2.  Merging enables controlled vocabularies to
work as a whole (semantic)
Interoperability must come before
merging (merging requires knowledge
of which vocabularies to merge)
Few standards in use


Structural approaches
If directly marking up documents, this
approach is probably impractical in the
enterprise
Better uses:
•  Limited high value documents (e.g.,
content models)
•  Document surrogates (e.g., Best Bet
records)


Merging vocabularies
Extremely difficult, and currently rare
Mostly found in libraries, academia,
scholarly publishing, and other
resource-poor environments
Examples, hard to hardest
•  Cross-walking vocabularies
•  Switching vocabularies
•  Meta-thesaurus
•  Single thesaurus


Merging Vocabularies:
Vocabulary cross-walking
Map terms peer-to-peer between
individual vocabularies
•  Primarily handles synonyms, not
relationships
•  Can be handled manually or through
automated means (pattern-matching)
Doesn’t scale well beyond two or three
vocabularies


Switching vocabulary
A single vocabulary that maps to existing
vocabularies (primarily synonyms)
Similar to cross-walking, but better at
handling translation when there are
more than two or three vocabularies to
connect


Meta-thesaurus
A switching vocabulary which also
includes thesaural relationships
(essentially a thesaurus of thesauri)
Example: National Library of Medicine’s
UMLS (Unified Medical Language
System)
•  Merges over 100 vocabularies
•  Describes fairly homogeneous domain
(medical literature) for fairly homogeneous
audience (health science professionals)

Single unified thesaurus
Highly impractical in enterprise context


Metadata:
What is metadata?
Data about data
Information which describes a document,
a file or a CD
Common metadata
•  CD information: title, composer, artist, date
•  MS Word document properties: time last
saved, company, author


Metadata:
Three types
1.  Intrinsic: metadata that an object holds
about itself (e.g., file name or size)
2.  Descriptive: metadata that describes
the object (e.g., subject, title, or
audience)
3.  Administrative: metadata used to
manage the object (e.g., time last
saved, review date, owner)


Metadata:
Common sources
Vocabularies from other parts of your
organization (e.g., research library)
Competitors
Commercial sources (see
www.taxonomywarehouse.com)
Your site’s users
•  Search analytics
•  Folksonomies
•  User studies (e.g., free listing, card sorting)


Metadata:
Big org,
big picture


Metadata: Scaling problems
Barriers to
enterprise
metadata
development:
•  Volume of
metadata
vocabs./silos
•  Complexity
of semantic
relationships
(beyond
synonyms)


Metadata in the Large Org:
Challenges
Two barriers to enterprise metadata
1.  Interoperability (structural)
2.  Merging enables controlled vocabularies to
work as a whole (semantic)
Interoperability must come before
merging (which requires knowledge of
which vocabularies to merge)
Few standards in use


Easy Business unit These are typically already
names available and standardized
Easy to Chronology Variations in formats (e.g.,
Moderate 12/31/07 versus 31/12/07) usually
can be addressed by software
Moderate to Place names Although many standards exist
Difficult (e.g., state abbreviations and
postal codes), many enterprises
(and their business units) use
custom terms for regions (such as
sales territories)


Moderate to Product Product granularity can vary
Difficult names greatly; marketing may think in
terms of product families; sales in
terms of items with SKU numbers,
and support in terms of product
parts that can be sold individually
Difficult Audiences Audiences, such as customers or
types of employees, vary widely
from unit to unit
Difficult Topics The most ambiguous type of
metadata; difficult for individuals,
much less business units, to come
to agreement on topical metadata

Metadata:
Levels of control

(Vocabularies)

Synonym Authority Classification
Thesauri
Rings Files Schemes

Simple Complex

Equivalence Hierarchical Associative
(Relationships)


Metadata semantic relationships:
Hard to hardest

Level of Type of Examples
Difficulty Relationship
Hard Synonymous Synonym rings and
authority lists
Harder Hierarchical Classification
schemes
Hardest Associative Thesauri


Metadata:
Strategy for large orgs 1/2
Coordinate to ensure:
•  Structural interoperability from the start
•  Semantic mergability over time
•  Vocabulary control and maintenance
through both manual and automated
means
•  A workflow model and policies to support:
•  Decentralized tagging and vocabulary updating
(through suggestions of new terms)
•  Centralized review and maintenance

Strategy for large orgs 2/2
“Serious metadata” is beyond the means of
most enterprises
•  Encourage local (e.g., departmental) vocabulary
development
•  Provides organizational learning and local
benefit
•  Enterprise-wide, start with “easier” vocabularies;
work your way to harder ones over time;
suggested sequence:
1.  Business functions
2.  Products
3.  Topics

Bottom-Up Navigation Roadmap

Content modeling


Metadata tagging


Bottom-Up Navigation Takeaways
1/3
Content models
•  Use to support contextual navigation
•  Apply only to homogenous, high-value
content
•  Won't transfer easily across silos and will
require significant metadata development


2/3
•  Distinguish attributes (and structural
interoperability) from values (and
semantic merging)
•  Costs and value both increase as these
increase:
•  Complexity of relationships between terms
(equivalence=>hierarchical=>associative)
•  Level of control (synonym rings=>authority
files=>classification schemes=>thesauri)
•  Think small: facets instead of a single
taxonomy

3/3
Metadata tagging
•  Make choices based on actual needs
(e.g., content models) rather than
exhaustive indexing
•  Consider costs of application and upkeep
•  Need for professional expertise
•  Metadata is a moving target that matches
other moving targets (users and content)


EIA and Search


EIA and Search

Search systems are a natural enterprise IA tool
•  Automated
•  Crawls what you tell it to
•  Doesn’t care about politics
Problems with shrink-wrapped search tools
•  Default settings, IT ownership minimize
customization to fit the enterprise’s needs
•  Results often not relevant, poorly presented
Customization is the answer
•  Within the realm of your team’s abilities
•  … and if IT will allow it!


EIA and Search:
Visually


Enterprise Search Design:
Potential improvements

Our focus:
1.  Clear interface
2.  Enhanced
queries
3.  Improved
results
(relevance &
presentation)

Basic search system components


Enterprise Search Roadmap

Search interface

Search queries

Search results


Search Interface Design:
The “Box”
The “Box” unifies
IBM.com
Consistent:
•  Placement
•  Design
•  Labeling
•  Functionality


Combine interfaces when possible

Will users understand?

Two boxes bad, one box good, usually…


The role of “advanced search” 1/2

Not a likely
starting point
for users who
are searching

Continued…

The role of “advanced search” 2/2
Suggestions
•  Use for
specialized
interfaces
•  Reposition as
“Revise
Search”
•  Don’t bother


Contextualizing Search Help:
Ebay example


Search Interface and Queries:
Functionality and visibility
Hide functionality? Consider the “Google
Effect,” human nature and the LCD
Don’t hide it?
•  Not if users expect it
•  Legacy experience (e.g., Lexis-Nexis users)
•  Specialization (e.g., patent searchers)
•  Not if content allows/requires it
•  Specialized content and applications (e.g., staff
directory)


The Query:
Query language considerations
Natural language
•  Usually don’t show up in search logs
•  Low priority, but nice to support
Operators (Booleans, proximity, wild cards)
•  Booleans: use default “AND” for multi-term queries
•  Less forgiving than treating as phrase, more selective than
“OR”
•  Most retrieval algorithms will find results for just one term
•  Rely on other approaches (e.g., filtering, clustering, Best
Bets) to reduce search results overload
•  Low priority: Proximity operators (e.g., “enterprise
(W3) architecture”), wild cards (e.g., “wom*n”)


The Query:
Query building considerations
Large potential benefits to improving
“intelligence” behind search queries
•  Adding semantic richness to queries allows for
stronger searches without “touching” content
•  Overrides “enterprise bias” embedded in content
•  A centralized (enterprise-wide) process
Query building approaches
•  Spell checking: can be automated
•  Stemming: can be automated
•  Concept searching: requires manual effort
•  Synonyms (via thesaurus): requires manual effort,
but no need to be comprehensive

Spell Checker:
Sur La Table example

A la Google…


Stemming:
IBM example

IBM uses
Fast Search


Concept Searching:
Social Security Admin. example

SSA uses
Convera

Thesaural Search:
ERIC example


Enterprise Search Interface:
Guidelines
Hide functionality on initial enterprise-wide
search
Cast the net widely: rely on query builders
to generate larger, higher quality result
sets
Use filtering/clustering to narrow
Use Best Bets to ensure strong initial
results


Individual Search Results:
Goals
Enable users to quickly understand something
about each document represented
That “something”: confirm that a known-item
has been found, or distinguish from other
results
Align to searching behaviors (determined
through user testing, persona/scenario
analysis, site search analytics)
•  Known-item
•  Open-ended/exploratory
•  Comprehensive research


Approaches
Basic approaches
•  Document titling
•  Displaying appropriate elements for each
result
These approaches have value in any
context, but especially useful in
enterprise setting


Document Titling:
DaimlerChrysler example

What do these document titles tell you?
And what do they tell you about DaimlerChrysler?

Document Titling:
Ford example
Descriptive document titles provide clear value


Displaying Appropriate Elements:
1) Determine common elements
Develop table of available elements (including
metadata) for disparate documents and records
•  Comes after content inventory and analysis
Develop table of common elements
•  Collapse similar elements (e.g., creator derived from author,
artist, source…)
•  Consider Dublin Core as model
•  Include bare minimum elements (e.g., title and description)


Displaying Appropriate Elements:
2) Select appropriate elements
Choose common elements which match most common
searching behaviors
•  Known-item
•  Open-ended
•  Comprehensive research
•  Etc.
Considerations
•  Which components are decision or action based?
•  Which components are of informational value only?
Display these elements for each search result


Step #1: common content elements
Step #2: select elements to display
Step #1 Title Description Creator Topic Date
Tech. Report Y Y Y Y Y

Policy Y N Y Y Y
Product Y Y N Y N
Sheet
FAQ Y N N Y N

Step #2 Title Description Creator Topic Date
Known-Item Y N Y N Y

Open-Ended Y Y N Y Y


Columbia University example
Long display for open- …shorter display for
ended searchers… known-item searchers


What happens next?
Augment with “next step” actions per result
•  Open in separate
window
•  Get more like this
•  Print
•  Save
•  Email
Determine next steps
through contextual
inquiry


Presenting Search Result Groups:
Ranked results
Difficulties with relevance ranking
•  Depends on consistent elements across
documents
•  Term frequency-dependent approaches create an
“apples and oranges effect” on ranking
•  Google effect: benefits of popularity make less
sense in enterprise context than in open web
Consider alternatives
•  Clustering and filtering
•  Manually-derived results (aka “Best Bets”)


Clustering & filtering
“Our user studies show that all Category interfaces were more effective
than List interfaces even when lists were augmented with category names
for each result” —Dumais, Cutrell & Chen

list results
clustered results
Consider using
clustered results
rather than list
results


Methods of clustering and filtering
Use existing metadata and other distinctions
(easier)
•  Document type (via file format or CMS)
•  Source (author, publisher, and business unit)
•  Date (creation date? publication date? last
update?)
•  Security setting (via login, cookies)
Use explicit metadata (harder)
•  Language
•  Product
•  Audience
•  Subject/topic

Clustering by Topic:
LL Bean example

Category
matches
displayed
rather than
individual
results


Filtering by Source:
BBC example

Selecting a
tab filters
results


Clustering by Content Type:
c|net example
Mention content modeling

Results clustered in multiple
content types


Clustering by Language Example:
PeopleSoft Netherlands
Result clusters
for Dutch and
English


Mixed Presentation of
Search Results


“Best Bets”:
By popular demand
Recommended links
•  Ensure useful results for top X (50? 100?)
most popular search queries
•  Useful resources for each popular query
are manually determined (guided by
documented logic)
•  Useful resources manually linked to
popular queries; automatically displayed in
result page


“Best Bets” Example:
BBC
Logic for BBC
Best Bets
•  Is query a
country name?
(yes)
•  Then do we
have a country
profile? (yes)
•  Then do we
have a
language
service? (yes)


“Best Bets”:
In the enterprise context
Who does the work?
•  Difficult to “assign” queries to different business units (e.g.,
“computing” means different things to different business
units)
•  Can serve as impetus for centralized effort
Operational requirements
•  Logic based on users’ needs (e.g., queries) and business
rules
•  Policy that assigns responsibilities, negotiates conflicts (e.g.,
who owns “computing”)
Opportunity to align Best Bets to user-centric divisions
(e.g., by audience: a “computing” best bet for
researchers, another for IT staff)


Enterprise Search:
Designs
•  Simple query builders (spell checker, stemming)
•  Search-enhancing thesaurus
Policies
•  Best Bets design and selection
•  Style guide (result titling, search interface implementation)
Staffing needs
•  Content inventory and analysis
•  Interface design
•  Work with IT on spidering, configuration issues
•  Ongoing site search analytics
•  Editorial (e.g., Best Bets creation)


Search Tool Selection:
EIA needs 1/2
To basic evaluation criteria (from
SearchTools.com)…
•  Price
•  Platform
•  Capacity
•  Ease of installation
•  Maintenance


Search Tool Selection:
EIA needs 2/2
…add:
•  Ability to crawl deep/invisible web
•  Ability to crawl multiple file formats
•  Ability to crawl secure content
•  API for customizing search results
•  Work with CMS
•  Duplicate result detection/removal
•  Ability to tweak algorithms for results retrieval and
presentation
•  Federated search (merge results from multiple
search engines/data sources)


Enterprise Search Roadmap

Search interface

Search queries

Search results


Enterprise Search Takeaways

Search interface and queries
•  Consistent location and behavior
•  Keep as simple as possible
•  Use "refine search" interface instead of "advanced
search"
•  Soup up users’ queries (e.g., spell checking)
Search results
•  Feature appropriate elements for individual results
•  Consider clustered results, especially if explicit,
topical metadata are available
•  Best bets results for top X common queries


EIA Research
Methods


EIA Research Methods:
Learn about these three areas

Content, users and
context drive:
•  IA research
•  IA design
•  IA staffing
•  IA education
•  …and everything else


Sampling challenges
How do you achieve representative
samples in the face of these difficulties?
•  Awareness: Who and what are out there?
•  Volume: How much is there? Can we
cover it all?
•  Costs: Can we afford to investigate at this
order of magnitude?
•  Politics: Who will work with us? And who
will try to get in the way?


Reliance on alternative techniques
Standard techniques may not work in
enterprise settings
Alternatives often incorporate traditional
methods and new technologies
•  Web-based surveys (e.g., SurveyMonkey)
•  Remote contextual inquiry and task
analysis (via WebEx)
•  Web-based “card” sorting (e.g., WebSort)
•  Log analysis tools (e.g., WebTrends)


A closer look
Content-oriented methods
•  Content inventories
•  Content value tiers
Context-oriented methods
•  Sampling stakeholders
•  Departmental scorecard
User-oriented methods
•  2-D scorecard
•  Automated metadata development
•  Freelisting
•  Site search analytics


Content Inventory:
Enterprise context
Issues
•  Even greater sampling challenges
•  Content research is even more critical:
serves as a cross-departmental exercise
Approaches
•  Balancing breadth and depth
•  Talking to the right people
•  Value-driven


Multidimensional Inventory:
Incomplete yet rich
EIA requires balanced, iterative sampling (where CMS
implementation may require exhaustive inventory)
Balance scope (breadth) with granularity (depth)
Extend inventory to all discernible areas of content,
functionality:
•  Portals and subsites
•  Application (including search systems)
•  Supplemental navigation (site maps, indices, guides)
•  Major taxonomies
•  Structured databases
•  Existing content models
•  Stakeholders


Content Migration Strategy:
Value Tier Approach
Determine value tiers of content quality that
make sense given your users/content/context
•  Answer “what content is important to the
enterprise?”
•  Help determine what to add, maintain, delete
How to do it?
1. Prioritize and weight quality criteria
2. Rate content areas
3. Cluster into tiers
4. Score content areas while performing content
analysis


Value Tier Approach:
Potential quality criteria
Select appropriate criteria for your
business context, users, and content
•  Authority
•  Strategic value
•  Currency
•  Usability
•  Popularity/usage
•  Feasibility (i.e., “enlightened” content
owners)
•  Presence of quality existing metadata

Weighting and scoring


Prioritization


Assessing Stakeholders:
What to learn from them
Strategic
•  Understanding of business mission and goals, and
fit with larger enterprise mission and goals
•  Theory
•  Practice
•  Culture: tilt toward centralization or autonomy
•  Political entanglements
Practical
•  Staff: IT, IA, design, authoring, editorial, usability,
other UX (user experience)
•  Resources: budget, content, captive audiences
•  Technologies: search, portal, CMS


Stakeholder Interviews:
Triangulate your sample
Org chart: business unit representatives
•  Will provide strategic overview of content and
whom it serves
•  May have some knowledge of content
•  More importantly, they know people who do in
their units
•  Additionally, political value in talking with unit reps
Functional/audience-centered
•  Subject Matter Experts (SMEs): represent power
users; valuable for pointing out content that
addresses major information needs
•  Audience advocates (e.g., switchboard operators):
can describe content with high volume usage

Finding the low-hanging fruit
Assessment should reveal degree of
“enlightenment”
•  Early adopters
•  Successful track records visible within the
enterprise
•  Understand/have experience with enterprise-wide
initiatives
•  Willingness to benefit the enterprise as a whole
•  They just plain “get it”
You’ve got to play to win: lack of interest and
availability mean loss of influence


Indicators of enlightenment
Technology assessment: who has/uses the
“classic 3”?
•  Portal
•  Search engine
•  CMS
Staff review: who has relevant skills/expertise
on their staff?
IA review: what areas of enterprise site have
strong architectures?
These areas may indicate redundant costs,
targets for centralization

Involving Stakeholders:
Departmental Report Card

Information Architecture Dept. Dept. Dept.
Heuristic 1 2 3
Supports orientation B- B B
Supports known-item searching A C+ C
Supports associative learning B C C
Supports comprehensive research A B+ B
Passes “navigation stress test” C F C+
… … … …


“Safe” User Sampling:
The 2D Scorecard
Combines alternative, apolitical methods
for determining segments to sample,
e.g.:
•  Role-based segmentation
•  Demographic segmentation
Distracts stakeholders from “org chart-
itis,” to purify sampling
Enables evaluation methods (e.g., task
analysis, card sorting)


The 2D Scorecard:
Role-based segmentation
Roles cut across political boundaries
•  Profile core enterprise-wide business
functions
•  Why does the enterprise exist?
•  Examples: Sell products, B2B or B2C
activities, manufacture products, inform
opinion, etc.
•  Determine major “actors” in each process


The 2D Scorecard:
Demographic segmentation
Standard, familiar measure; also cuts
across political boundaries
•  Gender
•  Geography
•  Age
•  Income level
•  Education level
Your marketing department probably has
this data already


The 2D Scorecard:
Combining roles & demographics
TEST Demo. Demo. Demo. Demo.
SAMPLE Profile Profile Profile Profile TOTAL
SIZE A B C D
Role 1 1 3 3 2 9
Role 2 2 2 1 1 6
Role 3 3 4 2 1 10
Role 4 0 3 4 0 7
TOTAL 6 12 10 4 32

The 2D Scorecard:
Incorporating contextual bias
Role/demographic “scorecard” is pure
•  Serves as a structure that doesn’t have to
change substantially
•  But how to incorporate stakeholder bias?
Stakeholder bias can be accommodated
•  Poll/interview stakeholders to determine
how cell values should change
•  Axes and totals stay mostly the same
•  Distraction is our friend

The 2D Scorecard:
After stakeholder input
TEST Demo. Demo. Demo. Demo.
SAMPLE Profile Profile Profile Profile TOTAL
SIZE A B C D
Role 1 1 2 5 1 9
Role 2 1 1 3 1 6
Role 3 3 4 2 1 10
Role 4 0 3 3 1 7
TOTAL 5 10 13 4 32

Maintaining a User Pool:
Build your own for fun and power
Through automated surveys, lower level
information architect built an enterprise-
wide pool of 1,500 users
•  Prescreened by demographics and skills
•  Provided him with substantial leverage with
others who wanted access to users
•  He just got there first and did the obvious
More information: http://louisrosenfeld.com/
home/bloug_archive/000408.html


Metadata Development:
Conventional techniques
Techniques
•  Open card-sorting to gather terms
•  Closed card-sorting to validate terms
•  Can be difficult to carry out in enterprise environment
(scope of vocabulary, subject sampling)
Modifications for enterprise setting
•  Use remote tools (e.g. IBM’s EZsort)
•  Apply in “stepped” mode: test subsections of taxonomy
separately
•  Drawback: lack of physical cards may diminish value of
data


Metadata Development:
Classification scheme analysis
Review existing schemes, looking for:
•  Duplication of domain
•  Overlapping domains
•  Consistency or lack thereof
Can some vocabularies be reused?
Improved? Eliminated?


Automated Metadata
Development:
Two classes of tools
Auto-categorization tools
•  Can leverage pattern-matching and cluster-
analysis algorithms to automatically generate
categories (e.g., Autonomy, Interwoven)
•  Can also use rules (i.e., concepts) to generate
categories (e.g., Inktomi, Verity, Entrieva/Semio)
Auto-classification tools
•  Apply indexing to existing categories
•  Require controlled vocabularies (generally
manually-created) to index content


Automated Metadata
Development:
Pros and cons
Benefits
•  Apolitical applications that disregard org
chart
•  May be a necessary evil in a large
Drawbacks
•  Limited value in heterogeneous, multi-
domain environment
•  Perform better with rich text, not so good
with database records and other brief
documents

Automated Metadata
Development:
Semio example
“E-commerce”:
A human would
collapse many
of these
categories

At best, an 80% solution; none truly “automated”
•  Significant manual proofing of the 80% of content indexed
•  Significant manual indexing of the 20% not indexed


Finding Metadata:
Free listing
Simple technique:
•  “List all of the terms you associate with ______”
•  Perform pair analysis (co-occurrence) on results
Benefits
•  Harvests terms associated with a concept or
domain
•  Can be done in survey form with many subjects,
multiple audiences
•  Supports card sorting
•  Less useful for structuring relationships between
terms
•  Possible alternative to site search analytics

The Zipf Curve:
Consistent and telling

Zipf distribution
from Michigan
State University
search logs
(derived from site
search analytics)

From http://netfact.com/rww/write/searcher/rww-searcher-msukeywords-searchdist-apr-jul2002.gif

Common Search Queries:
What they tell us


Site Search Analytics:
What does this data tell us?
Keywords: focis; 0; 11/26/01 12:57 PM; XXX.XXX.XXX.2
Keywords: focus; 167; 11/26/01 12:59 PM; XXX.XXX.XXX.2
Keywords: focus pricing; 12; 11/26/01 1:02 PM; XXX.XXX.XXX.2
Keywords: discounts for college students; 0; 11/26/01 3:35 PM;
XXX.XXX.XXX.59
Keywords: student discounts; 3; 11/26/01 3:35 PM; XXX.XXX.XXX.59
Keywords: ford or mercury; 500; 11/26/01 3:35 PM; XXX.XXX.XXX.
126
Keywords: (ford or mercury) and dealers; 73; 11/26/01 3:36 PM;
XXX.XXX.XXX.126
Keywords: lorry; 0; 11/26/01 3:36 PM; XXX.XXX.XXX.36
Keywords: “safety ratings”; 3; 11/26/01 3:36 PM; XXX.XXX.XXX.55
Keywords: safety; 389; 11/26/01 3:36 PM; XXX.XXX.XXX.55
Keywords: seatbelts; 2; 11/26/01 3:37 PM; XXX.XXX.XXX.55
Keywords: seat belts; 33; 11/26/01 3:37 PM; XXX.XXX.XXX.55


Instructions
Sort and count queries
Identify and group similar queries (e.g., “cell
phones” and “mobile phones”)
Understand users’ query syntax (e.g., use of
single or multiple terms, Boolean operators)
and semantics (e.g., use of lay or
professional terms)
Determine most common queries
•  Identify content gaps through 0 result queries
•  Build “Best Bets” for common queries
•  Map common queries to audiences through IP or
login analysis

Benefits for interface development
Identifies “dead end” points (e.g., 0 hits, 2000
hits) where assistance could be added (e.g.,
revise search, browsing alternative)
Syntax of queries informs selection of search
features to expose (e.g., use of Boolean
operators, fielded searching)

…OR…


Benefits for metadata development
Provides a source of terms for the creation of
vocabularies
Provides a sense of how needs are expressed
•  Jargon (e.g., “lorry” vs. “truck”)
•  Syntax (e.g., Boolean, natural language, keyword)
Informs decisions on which vocabularies to
develop/implement (e.g., thesaurus, spell-
checker)


Site Search Analytics: Benefits for
content analysis
Identifies content
that can’t be
found
Identifies content
gaps
Creation of “Best
Bets” to address
common queries


Pros and cons
Benefits
•  Data is real, comprehensive, available (usually)
•  High volume
•  Can track sessions
•  Non-intrusive
Drawbacks
•  Lack of good commercial analysis tools
•  Lack of standards makes it difficult to merge
multiple search logs (not to mention server logs)
•  More difficult to merge with other logs (e.g. server)
•  Doesn’t tell you why users did what they did


Enterprise context
Makes case for EIA; usually demonstrates
that users are requesting things that
aren’t tied to departmental divisions
(e.g., policies, products)
Informs “Best Bets”
Informs synonym creation
Limited value if not analyzing merged logs


EIA Research Methods Takeaways

Challenges
•  Many traditional methods can be adapted to the
•  But sampling, geography, volume and politics
force a less scientific, more pragmatic approach
•  Also force greater reliance on automated tools
We need new methods
•  Focus on minimizing politics and geographic
distribution
•  Most are untested
•  Information architects need to be willing to
experiment, innovate, and live with mistakes


EIA Framework


EIA and the Enterprise:
Phased, modular model
Phasing is not just about roll-out and
timing
Should be overarching philosophy for EIA
initiatives
•  We can phase in whom we work with
•  We can phase in whom we hire to do EIA
work
•  We can modularize what types of EIA we
do
•  We can phase in what degree of
centralization we can support

Why a Phased Model?
Because mandates don’t work
“Just do it!”…
•  …all (e.g., all subsites)
•  …now (e.g., in 3-6 months)
•  …with few resources and people (e.g., one sad
webmaster)
•  …in a way that minimizes organizational learning
(e.g., hire an outside consultant or agency)
Results of the mandated “solution”: completely
cosmetic, top-down information architecture


The EIA Framework
Seven issues
1.  EIA governance: how the work and staff are
structured
2.  EIA services: how work gets done in an enterprise
environment
3.  EIA staffing: who handles strategic and tactical
efforts
4.  EIA funding model: how it gets paid for
5.  EIA marketing and communications: how it gets
adopted by the enterprise
6.  EIA workflow: how it gets maintained
7.  EIA design and timing: what gets created and
when


The EIA Framework
Critical goals
Re-balance the enterprise’s in-house IA
expertise to support an appropriate
degree of centralization
Enable slow, scaleable, sustainable
growth of internal EIA expertise
Create ownership/maintenance
mechanism for enterprise-wide
aspects of IA (currently orphaned)
Ensure institutional knowledge is retained


EIA Governance:
Questions
What sort of individuals or group should
be responsible for the EIA?
Where should they be located within the
organization? How should they address
strategic issues? Tactical issues?
Can they get their work done with carrots,
sticks, or both as they try to work with
somewhat autonomous business units?


EIA Governance:
A separate business unit 1/2
Logical outgrowth of
•  Web or portal team
•  Design or branding group
•  E-services, e-business or e-commerce unit
Goals
•  Ensure that IA is primary goal of the unit
•  Retain organizational learning
•  Avoid political baggage
•  Maintain independence


EIA Governance:
A separate business unit 2/2
Ambitious, fool-hardy, unrealistic?
Necessary!
•  Models of successful new organizational
efforts often start as separate entities
•  Alternatives (none especially attractive)
•  Be a part of IT or information services
•  Be a part of marketing and
communications
•  Be a part of each business unit


EIA Governance:
Balancing strategic and tactical
Strategic: Model on Board of Directors
•  Represent key constituencies
•  Track record with successes, mistakes with
organization’s prior centralization efforts
•  Mix of visionaries, people who understand
money
Tactical: Start with staff who “do stuff”
•  Extend as necessary by outsourcing
•  Enables logical planning of hiring and use
of consultants and contractors


EIA Governance:
Board of directors 1/2
Goals
•  Understand the strategic role of information
architecture within the enterprise
•  Promote information architecture services as a
permanent part of the enterprise’s infrastructure
•  Align the group and its services with those goals
•  Ensure the group’s financial and political viability
•  Help develop the group’s policies
•  Support the group’s management
Makeup
1. Draw first from effective leaders
2. Then from major units that would be strategic
partners

EIA Governance:
Board of directors 2/2
Qualities
•  Experience and duration in the enterprise
•  Wide visibility and extensive network
•  Can draw on institutional memories and experiences
•  Track record of involvement with successful initiatives
•  Entrepreneurial (can read and write a business plan)
•  Experienced with centralization efforts
•  Does not shy away from political situations
•  Can “sell” a new concept and find internal funding
•  Is like the people you need to “sell” to
•  Has experience with consulting operations
•  Has experience negotiating with vendors


EIA Governance:
Caterpillar’s boards
Strategic board (quarterly; @10 members)
•  “Owners” of enterprise site
•  Decide on major policies
•  Settle conflicts
Stakeholder board (monthly; 15-20)
•  Ensure broad participation
•  Ensure two-way communication
•  Make recommendations re: policy to strategic
board
User advocacy board (meets as needed; 5-10)
•  Represent major user groups
•  Maintain pool of sample users

EIA Services:
Questions
What should a team responsible for EIA
actually do?
How do their “services” fit with work that
happens within business units? Or with
outside contractors and consultants?
What kind of people should manage these
efforts?
How do IA generalists and specialists fit
together?


EIA Services:
Modular service plan
Avoid “monolithic” approach: “Hi, we’re the EIA
team and we’re here to help… and we’re going
to centralize all of your information…”
Break IA and CM into digestible, non-threatening
tasks and sell those
•  Allows you to divide and conquer clients…
•  …and helps you understand IA challenges better
(e.g., applying metadata in a centralized
environment)


EIA Services:
Potential service offerings 1/3
Client workflow-oriented (map to content
publication process)
•  Content authoring and acquisition
•  Metadata development
•  Content titling
•  Content tagging
•  Content review (voice, accuracy, etc.)
•  Content formatting
•  Formatting review
•  Optimization for search engine optimization
•  Publication

EIA Services:
User-oriented
•  Persona and scenario development
•  User testing and task analysis
•  Search and server log analysis
Content-oriented
•  Content inventory and analysis
•  Content evaluation and assessment
•  Content model design
•  Content development policy (creation, maintenance)
•  Content weeding, ROT removal, and archiving
•  Content management tool (acquisition, maintenance)
•  Metadata maintenance
•  Manual tagging
•  Automated categorization and classification

EIA Services:
Context-oriented
•  Business metrics development and analysis
•  Internal marketing strategy and implementation
•  Stakeholder and decision-maker interviews
•  Business rules development (for best bets, content models, etc.)
Production/Maintenance
•  Template design and application
•  Training
•  Policy/procedure/standards development and acceptance
•  Publicity of new/changed content
•  Tool analysis/acquisition (CMS, search, portal)
•  Quality control and editing
•  Link checking
•  HTML validation
•  Liaison with visual design staff, IT staff, vendors


EIA Services:
Assessing departmental IA needs


EIA Services:
Basic & premium levels

Free services can lead to fee services

EIA Services:
Phased demand for IA services


EIA Staffing:
Questions
Who should be involved: in-house,
consultant, contractor? What type of
specialization should the staff have?
Should they be centralized or located
within business units or both?


EIA Staffing:
Tactical team basics 1/2

Goals
•  Delivers IA services to the enterprise in
content, users, and context areas
•  Implements the strategic team’s policies
•  Works directly with clients to understand their
needs and develop new services to meet
those needs


EIA Staffing:
Tactical team basics 2/2
Make-up driven by “market demand,” existing
resources
“Vertical” IA generalists: split between EIA project
enterprise business units
“Horizontal” IA specialists: “consultants” for both
groups of generalists
•  Tools (e.g., search, portal, CMS)
•  Metrics
•  Evaluation
•  XML and other markup languages

EIA Staffing:
Tactical team qualities

Entrepreneurial mindset
Ability to consult (i.e., do work and justify IA
and navigate difficult political
environments)
Willingness to acknowledge ignorance and
seek help
Ability to communicate with people from
other fields
Sensitivity to users’ needs
…and know about IA and related fields

EIA Staffing:
Tactical team backgrounds/skills
• Human Computer
interaction • Librarianship (tech.
• Cognitive services)
Psychology • Information Science
• Librarianship • Journalism
(reference) • Technical
• Marketing Communication
• Branding • Computer Science
• Merchandising • Graphic design

• Organizational Psychology
• Business Management
• Operations Engineering
• Social Network Analysis
• Ethnography
• Economics

EIA Staffing:
Shoot for this org chart


EIA Staffing & Governance


EIA Funding Model:
Questions
How should this group be funded?
How should other expenses (e.g.,
software licenses) be covered?
Charge-back fees for individual
services? Flat “tax” paid by business
units? Covered by general
administration's tab? Some hybrid
thereof?
Should certain services be performed
gratis, while others require payment?

EIA Funding Model:
Looking for inspiration
Study the successes/failures of the enterprise’s
other centrally funded services
Possible plan
•  Initially: “tax” on business units and/or “seed
capital” from senior management
•  Ultimately: self-funding (models: IT, HR, special
projects)
Key: funding should be from central group (e.g.,
senior management) or self-funded; else too
much dependency on business units


EIA Funding Model:
Ensuring independence
Potential models already in existence in the
enterprise
•  Charge-back
•  Tax on business units
•  Money from general fund
•  Hybrids
Charge-back model is attractive
•  Increasing perceived value of IA by charging fees
•  Compares well with duplicated expenses incurred
by business units


EIA Funding Model:
Diversify revenue streams


EIA Marketing & Communications:
Questions
How to position this work and the group that
supports it: IA? User Experience? Web
Design? How do these terms affect the
scope of the work/charter of the group?
How does a plan like this get “sold,” and to
whom?
Whose support is needed, and what tactics are
useful in convincing them to support EIA
work?
How to prioritize which business units around
the enterprise to work with?


Positioning the EIA initiative
Approaching “clients”
•  No carrot or stick
•  Offer services and consulting that save
money, reduce tedium
Branding: choose the term that is
•  Hottest
•  Has least baggage
•  Steps on fewest toes


Selling IA
Concrete
•  We can make work easier and save money
for individual business units
•  We can improve the user experience and
build brand loyalty among customers,
organizational loyalty among employees
•  We can minimize the enterprise’s habit of
purchasing redundant licenses and
services


One unit at a time
Start with low-hanging fruit
•  Killer content
•  Plentiful or influential users
•  Strategic value (business context)
Determine current status of the “client”
•  What are they doing now?
•  What expertise is in-house?
•  What relevant tools do they own (extend
licenses)?
•  Are they enlightened?

Illustrating the concept
Select an initial model for centralized
approach that’s familiar, accessible
Staff directory often the best
•  Serves all enterprise users
•  Useful, highly structured content which
may have significant metadata, searching
and browsing capabilities
•  Has high value in context of the
enterprise’s daily operations


EIA Design/Timing:
Questions
An EIA design is an overwhelmingly large
undertaking; how might it be broken into
more digestible pieces?
How should they be sequence: what
makes sense to take on now, later, or
perhaps not at all?


EIA Design/Timing:
Modular, phased


EIA Design/Timing:
3-6 years, not months
Use early successes as models
Anticipate greater centralization among
and within business units over time
Support different levels of centralization
concurrently (Neanderthals coexist with
Space Agers)


EIA Workflow:
Questions
How does the content authoring and
publishing process work now?
Who and how many are involved?
How can the group support that work, and
determine the best mix of centralized
and autonomous responsibilities within
that workflow?


EIA Workflow:
Supporting variation, evolution
Build around business units’ demand
Use as driver for CMS selection


EIA Framework Takeaways

Be entrepreneurial
•  Market and sell services to internal clients
•  Become self-sustaining by diversifying revenue
streams
Offer modular services
•  Specific services, not full package
•  Logical migration path accommodates all stages of
evolution along centralization/autonomy axis for
customers
Do what can be done in baby steps
•  Start with projects that are low hanging fruit
•  Selective roll-out


Discussion


Contact Information

Louis Rosenfeld, LLC
457 Third Street, #4R
Brooklyn, NY 11215 USA

lou@louisrosenfeld.com
www.louisrosenfeld.com
@louisrosenfeld

+1.718.306.9396 voice
+1.734.661.1655 fax


Enterprise Information Architecture: Because users don't care about your org chart

More Related Content

What's hot (20)

Viewers also liked (8)

Similar to Enterprise Information Architecture: Because users don't care about your org chart (20)

More from Louis Rosenfeld (20)

Recently uploaded (20)

Enterprise Information Architecture: Because users don't care about your org chart