Building a Faceted Classification IA Summit Redux San

The Project • Global chemical company • Customer-facing web site redesign • Client had completed extensive user research on audience segments, task analysis • Client determined they needed a faceted classification for information access by disparate audience segments © 2005, Contextual. Analysis, LLC

The 3 Librarians Fred Leise Contextual. Analysis, LLC; IA with specialty in metadata and controlled vocabularies, faceted classifications Amy J. Warner lexonomy. com; IA with specialty in metadata and controlled vocabularies; former academic at UM-SI Sarah A. Rice Seneb Consulting; IA with specialty in user research, metadata/CV development. © 2005, Contextual. Analysis, LLC

Project Team Client Team • Key stakeholder as project manager • Stakeholder from IT • Stakeholder from KM group/subject expert Geographically Distributed • UK • Midwest • California © 2005, Contextual. Analysis, LLC

Our Engagement with Client We were asked to: • Build upon AP research • Develop a content model (metadata schema) and appropriate controlled vocabularies (CVs) • Include multiple sessions of user research • Offer solutions for global website to be implemented in four different languages © 2005, Contextual. Analysis, LLC

Methodology: Overview © 2005, Contextual. Analysis, LLC

Mental Model/Audience Analysis What We Found Different user segments used the web site differently, had different information needs • Segment 1 developed new products and looked for new ideas • Segment 2 identified and procured products • Segment 3 used products © 2005, Contextual. Analysis, LLC

In-Depth Interviews What We Wanted • Discover user’s information-seeking behavior – When did they look for information? – Where did they look? – What did they look for? – When did they stop looking? – Did different cultures have different information-seeking behavior? © 2005, Contextual. Analysis, LLC

In-Depth Interviews What We Did • One-hour phone interviews • Two team participants • Taped conversations, made simultaneous transcriptions • Studied analyzed transcripts © 2005, Contextual. Analysis, LLC

In-Depth Interviews What We Found • How users searched (potential facets) • Search terminology (potential vocabulary terms) • No cultural differences • Differing audience segments had different information needs © 2005, Contextual. Analysis, LLC

Content Analysis What We Wanted • What facets are in current content? What We Did • Gathered content types • Performed detailed analysis • Identified possible facets What We Found • Multiple facets having significant overlap with facets from user research © 2005, Contextual. Analysis, LLC

Preliminary Facet List What We Did • Developed exhaustive facet list • Prioritized list • Developed list of facets to be develop, those to be developed later • Identified facet characteristics • Identified appropriate facets for testing © 2005, Contextual. Analysis, LLC

DEFINITION VOCAB. TYPE OPEN/ CLOSED REPEATABLE? REQUIRED? MANUAL/ AUTO. TAGGING? Author The name of the person or persons that produced the document. flat list Open Yes Manual Author Country The places where products are sold and used. Hierarchy Closed Yes Manual Author Document Type Language Product Name Subject Target Audience Document Title FACET NAME © 2005, Contextual. Analysis, LLC SOURCE

Open Card Sorting What We Wanted • Validate preliminary facet list • Identify any

Open Card Sorting What We Did • Client identified and scheduled users • Emailed users instructions and terms 24 hours in advance • Conducted test via conference call and Web. Ex • Used two team members for each call • Analyzed card sort results – Top-down cluster analysis by hand – Bottom-up term co-occurrence analysis by hand – Dendrite diagram analysis using EZSort © 2005, Contextual. Analysis, LLC

Term 1 Term 2 Term 3 Term 4 Term 5 Term 6 Term 7 Term 8 Term 9 Term 10 Term 11 Dendrite Diagram Term 12 Term 13 Term 14 Term 15 Term 16 Term 17 Term 18 Term 19 Term 20 Term 21 Term 22 Term 23 Term 24 Term 25 Term 26 Term 27 Term 28

Open Card Sorting What We Found • Several facets not easily distinguished, needed to be combined • Audience role did not affect results • In general, facet model was on target © 2005, Contextual. Analysis, LLC

Gather Terms What We Wanted • Develop complete CVs for all appropriate facets ©

Gather Terms What We Did • Gathered terms from: – user interviews – client’s existing term lists – client’s website – competitors’ websites • Identified CVs needing client input © 2005, Contextual. Analysis, LLC

Gather Terms What We Found • Many CVs were completed relatively easily • A

Country Africa. Algeria (People’s Democratic Republic of Algeria) (DZA). Angola (Republic of Angola) (AGO). Belize (BLZ). . Americas. Argentina (Argentine Republic) (ARG). Aruba (ABW). Bahamas (Commonwealth of the Bahamas) (BHS). . Asia. Australia (Commonwealth of Australia) (AUS). Azerbaijan (Azerbaijani Republic) (AZE). Bangladesh (People’s Republic of Bangladesh) (BGD). . Europe. Albania (Republic of Albania) (ALB). Andorra (Principality of Andorra) (AND). …. Bosnia-Herzegovina (Bosnia and Herzegovina) (Republic of Bosnia and Herzegovina) (BIH)

Closed Card Sorting What We Wanted • Validate facet labels • Validate assignment of

Closed Card Sorting What We Did • Client identified and scheduled users for testing • Emailed users instructions and term list 24 hours in advance • Conducted test via conference call and Web. Ex • Used two team participants © 2005, Contextual. Analysis, LLC

Closed Card Sorting What We Did • Analyzed results using: – Top-down cluster analysis by hand – Bottom-up term co-occurrence analysis by hand – Dendrite diagrams produced by EZSort • Reviewed preferences for facet labels © 2005, Contextual. Analysis, LLC

Closed Card Sorting What We Found • Users understood facet labels and could sort

Recommendations for Implementation What We Want • Understand how faceted classification will be used in browsing • Understand how faceted classification will be used in searching • Develop maintenance plan for CVs What We Did • Presented findings, CVs and wireframes to various stakeholder groups within company • Developed list of best practices for implementation and maintenance • Suggested high level roadmap for implementation, to start immediately • Provided recommendations for tool selection, including criteria for use and experience with various tools. © 2005, Contextual. Analysis, LLC

Major Findings • Audience type makes a difference in which facets they choose to use to search for information • No cultural differences in classifying and labeling terms, at least in this application • Users’ facet labels differed but meaning of labels tended to be the same © 2005, Contextual. Analysis, LLC

Lessons Learned • Need to begin recruiting for testing well in advance • Pretesting important • User testing took two weeks rather than one because of scheduling complexities • Having multiple team members enhanced overall results • Cooperative client ensured project success © 2005, Contextual. Analysis, LLC

Deliverables List • Initial presentation to client’s core audience (about 40 individuals) • Screeners and questionnaires for all user testing sessions • User research report • User interview transcripts © 2005, Contextual. Analysis, LLC

Deliverables List • Content analysis spreadsheet • Content model testing report • Preliminary and revised facet list • Controlled vocabularies spreadsheet • Final report and presentations • Project tracking spreadsheet © 2005, Contextual. Analysis, LLC