Knowledge Architecture MSWeb case study and beyond IA

Knowledge Architecture : MSWeb case study and beyond IA 2000 Vivian Bliss Microsoft Information Services vbliss@microsoft. com Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved.

Where are we going today? MSWeb case study • 1995 to 1997 (prototype to v 3. 0) • our content culture • the re-design challenge • 1998 to the present (v 4. 5) Beyond MSWeb…extending the architecture 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 2

MSWeb case study 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved.

MSWeb (prototype) 10/27/2000 August 1995 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights

MSWeb 1. 0 1996 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights

MSWeb 1. 5 1996 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights

MSWeb 2. 0 1996 (news integration) 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp.

MSWeb 3. 0 Front 10/27/2000 1997 Copyright 2000 Vivian Bliss and Microsoft Corp. All

MSWeb 3. 0 Search 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights

Our content culture…from the beginning and nothing has changed… • number of pages on the intranet grows every minute – (currently 2. 5 million) • we do not create the content – created internally by employees – purchase from third party vendors, i. e. news • we have no control over the creation or publication of the content • any and every employee in the company can be an author (currently >39, 000) • content is stored on servers across the company, • servers storing the content can be located anywhere in the world where an MS employee is located (currently >65 countries) • no one is obligated to tell us of the existence of content or its location 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 10

What we knew • MSWeb is the enterprise portal for Microsoft employees worldwide-used by 94% of the company • customer satisfaction surveys show that most users come to MSWeb to: – search for information on the Intranet – find other Microsoft sites 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 11

User concerns • although users rated the portal very highly in many areas, specific concerns included: – number of clicks to get to desired results – ease of site navigation • recommended improvements included: – intelligent search (better results at top) – more sharply defined categories (navigation structure) 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 12

Other factors • MSWeb also informally provided search services to many sub-portals in the company – most sub-portals didn’t have staff or resources to support their own search and browse services – with a small team, it was difficult to provide custom services or support updates to code – lack of consistent metadata made it difficult to improve search results – most sub-portals didn’t understand the complexities of navigation and search 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 13

The re-design challenges • how can we identify the important content for exposure on MSWeb? • how could we improve enterprise search and navigation for users, and provide these resources to sub-portals, without large resource commitments? • all of this with the goal to save Microsoft employees time, improve their productivity, and enhance their decision-making 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 14

MSWeb Re-design Process Start Portal Reviews MSWeb unsolicited feedback Communications Review User Tasks Survey MSWeb v. 3. 0 Benchmark Key Content Publisher Interviews January 1999 Vision Project Charter Scope & Requirements January 1999 Keep Going Key stakeholder buy-in January 1999 Review December 1998 Revisions Content enrichment Knowledge architecture Development Functional testing Technical Specs February 1999 Prototype February 1999 Usability Testing Solicited feedback March 1999 SP 1 June 1999 Usability Testing July 1999 Version 4. 1 September 1999 10/27/2000 April 1999 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. Launch May 3, 1999 December 1999 15

Portal architecture 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 16

What we did. . created. . lost sleep over… • decided on the content pieces for the front page-stock ticker, events, news, knowledge centers, key sites • found, crawled and indexed even more of the intranet, approx. 2. 5 million pages. • cataloged the intranet sites with the best information for the most often asked questions thru research/reference requests and MSWeb Search queries -- MSWeb Best Bets, plus “I Need To’s” in natural expression – developed cataloging database, tool and process • created a set of controlled vocabularies (search taxonomies) – in cooperation w/ groups across the company – for cataloging/tagging Best Bets and other content – for enhancing Search • developed database and tool for storing vocabs complete with hierarchical and associative relationships, plus tool to create and manage 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 17

What we did. . created. . lost sleep over cont’d • developed “Categories” for a browse navigation based on users and tasks, not organization structure (navigation taxonomies) • used the Categories as a classification system for the Best Bets • ran a metadata/tag audit of the intranet to understand benchmark what was happening out there • developed and recommended metadata schema for web publishing across the company • created a metadata registry to capture schemas and crosswalk between conceptually similar tags • changed the user interface 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 18

MSWeb 4. 0 Front (top) May 1999 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft

MSWeb 4. 0 Front (lower) May 1999 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft

MSWeb 4. 0 Search “access” 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All

Were we successful? Measure Task Success Rate Time on Task Number of Clicks 10/27/2000 v. 3. 0 Average v. 4. 01 Average 68. 30% 79% +10. 7% 3 min: 26 sec. 3 min: 10 sec. -16 sec. 13 5 -8 clicks Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. Change 22

MSWeb Re-design Process, revisited Start Portal Reviews MSWeb unsolicited feedback Communications Review MSWeb v. 3. 0 User Tasks MSWeb v. 3. 0 Benchmark Key Content Publisher Interviews January 1999 Vision Project Charter Scope & Requirements January 1999 Keep Going Key stakeholder buy-in January 1999 Review December 1998 Revisions Content enrichment Knowledge architecture Development Functional testing Technical Specs February 1999 Prototype February 1999 Usability Testing Solicited feedback March 1999 SP 1 June 1999 Usability Testing July 1999 Version 4. 1 September 1999 10/27/2000 April 1999 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. Launch May 3, 1999 December 1999 23

Portal architecture 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 24

What we did. . created. . lost sleep over…part 2 streamlined the UI tweaked the Categories re-worked Best Bets eliminated exposure of “Related Terms” in Search leveraged vocabularies in Search automatic query expansion term disambiguation added five additional content stores to Search, including familiar content stores such as the Library Catalog 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 25

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 26

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 27

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 28

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 29

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 30

Beyond MSWeb…extending the architecture to other portals 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft

Portal architecture 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 32

10, 000 lines of code and 14 humans images mr MSWeb Search box news Categories Best Bets I Need To’s Search Intranet News MR Library… content crawl BB library intranet content crawl Site Server catalogs other portals tool BB crawl tool Query logs tool 10/27/2000 Best Bets Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. Vocabs schemas 33

670 lines of code and 14 humans Search box MSWeb Categories Search XML Search Request dosearch content crawl Best Bets Intranet News MR Library… tools consolidated in a web interface 10/27/2000 collection library XML files XML delivery of results with XSL content crawl Site Server catalogs BB crawl create & manage Best Bets view query logs view search metrics create & mange vocabs & schemas Query logs Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. Best Bets Vocabs schemas 34

Extending the architecture to other portals at MS • Search Query and Results – crawl and index their specific content – choose one or more of our established collections to expose in their portal search in addition to their specific content – provide search results back to their own portal search in XML • design their own Search UI to display the results, or • take our Search UI in XSL along with the content in XML • URL Cataloging Service – share the tools we created and teach them how to catalog an intranet site leveraging the controlled vocabularies and a common metadata schema • Category Labels – walk them through the process of creating a browse navigation based on user behavior, including contextual inquiry and analysis • Vocabularies – use the controlled vocabularies we created in their own publishing systems for metadata consistency • Consulting Consulting 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 35

670 lines of code, 14 humans and other portals Other. Portals Categories Search box Search XML Search Request dosearch Best Bets I Need To’s Customer content Intranet News … collection library XML files XML delivery of results with or w/o XSL content crawl Site Server catalogs BB crawl IS SAS Console Welcome Customer! 10/27/2000 content crawl create & manage Best Bets view query logs view search metrics view vocabs & schemas Query logs Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. Best Bets Vocab. Man Vocabs schemas 36

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 37

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 38

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 39

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 40

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 41

HRWeb • • used all SAS services identified all content for SAS to crawl and index created Categories for browse navigation created Best Bets set just for their content using SAS vocabs for certain tag values classified their Best Bets using their Categories. HR Best Bets displayed through browse navigation and search results new UI we consulted and assisted every step of the way 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 42

Windows UA – Search Query & Results • identify their specific content • sort options for their content • choose from other available content collections • number of results per page • language for “no results found” • XML feed only? need our XSL? • if XSL, choose a color and tab name for their content 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 43

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 44

10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 45

Please insert two pieces of code… This will send the users’ query to our Search box telling the “dosearch” function where it is coming from, the collections to target and what color to make the tab with the results display. http: //issas/host/default. asp? hostdef=winsuadef host/. xml&rcolor=gold This string places our “Powered by ISSAS” branding on the customer’s search page. href="http: //issas/"> <img src="images/poweredby. SASlogo. gif" width="108" height="26" alt="Powered by ISSAS" border="0"></a></td> 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 46

Are we successful? • in the first three months of offering SAS, 9 sub-portals have implemented search on their sites • 2 of those have created site-specific categories for their navigation; a third is in progress • all have leveraged the MSWeb Best Bets results in their custom search • no increase in staff at KNAC • equivalent to a cost savings of 45 person years in avoided work 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 47

Questions? Vivian Bliss Microsoft Information Services vbliss@microsoft. com Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved.

Appendix Additional information 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved.

The volume and complexity of information and knowledge available to individuals from both within an organization and from outside is continually expanding. Successful access to and integration of these resources depends on a wellstructured knowledge architecture. The interaction of this architectural layer with other components of a portal such as search, user interfaces and business processes can provide a rich set of services and content, if the design is based on user's needs. Examination of what makes content relevant for users, and the context that surrounds their use of knowledge and information are important in this approach. Issues such as privacy, user education, and information channeling must also be considered in the overall design. ~Mike Crandall, October 2000 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 50

What are Taxonomies? • Taxonomy: a classification of elements within a domain – Domain: a sphere of knowledge, influence, or activity – Classification: the operation of grouping elements and establishing relationships between them (or the product of that operation) – Relationships: a defined linkage between two elements – Element: an object or concept 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 51

Search As a Service • a project to formalize and productize MSWeb offerings in the enterprise search arena • includes search, metadata support, search metrics, and optional UI • also includes search and vocabulary management tools, formal documentation and processes for customer support and change control • the first step toward an object-based portal 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved. 52

Screenshots of the UCS Tool Best Bets tagging tool using a defined metadata schema and pulling from controlled vocabularies 10/27/2000 Copyright 2000 Vivian Bliss and Microsoft Corp. All Rights Reserved.