Transcript Slide 1
Successful Enterprise Search by Design Be the Hero, not the Goat Agenda Search is a user experience Bad search = low productivity Why configure search Web search is not Enterprise search Configuration Framework –Define Problem space –Define Scope –Content is king, Context is the realm –Build something beautiful –Build something meaningful –Must haves – Nice to Haves Key Takeaways 2 SEARCH IS A USER EXPERIENCE 3 A Frustrating Experience 4 A Bad Experience Searchers do not know “how to search” • 56% constructed poor queries • Proficiency with the machine does not translate into proficiency with the software Searchers get lost in the data • 33% had difficulty navigating/orienting search results • 28% had difficulty maintaining orientation on a website Loss of capacity for discernment • • • • 36% did not go beyond the first 3 search results (not pages…results on page 1) 91% did not go beyond the first page of search results 55% selected irrelevant results 1 or more times 5 A Weakening Experience Searchers do not know “how to search” • 56% constructed poor queries • Proficiency with the machine does not translate into proficiency with the software Searchers get lost in the data • 33% had difficulty navigating/orienting search results • 28% had difficulty maintaining orientation on a website Loss of capacity for discernment • 36% did not go beyond the first 3 search results – not pages…results on page 1 • 91% did not go beyond the first page of search results • 55% selected irrelevant results 1 or more times 6 Difference in Looking for Information • • • • Human Retrieval Contextual Free form Navigational or informational Focused and random – Berrypicking • Constrained – By technology – By biology Machine Retrieval • Literal • Directed • Rigid – One way – Sequential • Constrained – By size of index – By nature of instructions 7 Difference In Relevance Perception Thought-Processing Biped Relevance • Emotional • Environmental Good means pleasing, honest, truthful, operates with integrity Machine Relevance • Literal • Logical Good means fulfills a programmed criteria based on computational mathematics "I shall not today attempt further to define the kinds of material I understand to be embraced within that shorthand description; and perhaps I could never succeed in intelligibly doing so. But I know it when I see it.” Justice Potter Stewart Miller v California (1973) 8 WEB SEARCH IS NOT ENTERPRISE SEARCH 9 Web Has Machine Refinements Semantic • Authority • Contextual relatedness • Phrase indexed based on popular searches: Index, categories, keywords, document-specific data • Similarity estimation: Compares a “sketch” or compact representation for document and uses an established similarity threshold to delete duplicate entries Prediction • Orion Algorithm: Search engine algorithm uses vector space analysis that combines vector positioning with previous user action – Google 2009 – Microsoft Powerset Acquisition 10 Web Relies Heavily on Relevance Inputs Behavior that influences relevance • • • • Query Click through Time on page Path Social Influences • • • • • Likes Comments Recommendations Retweets Click-throughs 11 Bad Search = Low Productivity 12 Many Voices Same Conclusion Enterprise searchers spend longer looking because “they know it is there somewhere” • IDG: 2.5 hours/week/employee • Ford:5-15% of time on non-productive information related activities Coping mechanisms for poor enterprise search • • • • • Recreate Use older assets Interrupt a co-worker Start without info needed Don’t start 13 Configuring Enterprise Search 14 Your Manager Saw the Demos 15 The OOB Experience Won’t Cut It It is not what the vendor used in the demo No matter what “they “ say 16 Governance Does Not Optimize Search 17 Controlled Vocabulary Will Not Optimize Search And neither will a taxonomy 18 CONFIGURATION FRAMEWORK 19 Know What Problem You’re Solving 20 Define Problem Space Objectives • Find out: who is searching • Find out: what they are looking for • Find out: How they are searching (what keywords/phrases, how often they iterate, etc) • Find out: What drives their determination of relevance Tools • Site analytics • Search logs • User/stakeholder interviews 21 Don’t Boil the Ocean 22 Define the Scope Objectives • Define the search requirements • Find out: What to crawl? • Find out: Where the content lives? • Find out: how to index Map internal and external resources • Discover the sacred cows Tools • Discovery workshop • Infrastructure review • Client/stakeholder interviews 23 No…They Don’t 24 Use Content Enrichment Methods Objectives • Have a content Strategy • Reduce the amount of content • De-dupe and Archive • Describe content in effective, machine-readable fashion • Build content relationships (relational content models) • Sustain best practices through education Tools • • • • • User surveys Content Audit Content Creator Workshop Managed properties Define custom entities 25 Content Strategy Intersection of what you have/do with how customers look for what you have/do Use online tools to mine customer search behavior Check to see if you have relevant content and fill gaps Tools Core Metadata Google Insights for Search Site analytics Content audit 26 Relational Content Modeling Tools Guided Tours Produced Views Task List Drop Downs Related Links Best Bets 27 Build Something Meaningful 28 Structure Objectives • Compensate for lack of link relevance o URL Depth: the further from the homepage, the less important it must be o Click Distance: the further from an authority page, the less important it must be • Create Meaningful URLs o Keywords found in URLs are weighted for relevance o Hyphens as separators is best Tools • Flat structure within CMS • Analytics • Cross linking 29 Build Something Desirable 30 Design Relevant User Experience Objectives • Wean users from Google Web search performance expectations • Encourage and enable better query construction through abstraction Tools • • • • Filters Facets Subscription Clustering 31 Provide User Assistance Tools Suggestions as query is entered • At page search box • On search page Augmented Search results • Preview in browser • Contact information Did You Mean (spell check) Best Bets 32 Give the User Some Control Tools • Facets • Filters • More Like This… 33 Post Launch Analytics/Reports are your friends Zero results are the road to Perdition Refine, iterate, tune User feedback is not a one-time affair 34 Evaluate/Review/Refine methods Objectives • Keep ahead of user satisfaction by fixing problems early • Obtain client feedback on performance • Be agile: review and tune accordingly • Benchmark success Tools • • • • Establish benchmarks on what represents success Search logs Power user feedback Periodic company-wide survey feedback 35 Key Takeaways Search is a user experience Users bring outside expectations and behaviors inside the enterprise Enterprise search engines are not smart OOB is not what you paid for 36 End Result Will Be Getting more from less by …making what you have work smarter 37 Thank You [email protected] 38