Key Considerations in Maximizing the Value of Cognitive Search

Make no mistake, these are hard numbers to hit. At BA Insight, we focus on making each of these goals achievable. Here’s how:

Goal 1: Searches Should Return Fast Results

BA Insight’s SmartHub software provides the user interface components and query engine that allow for a high-performance search implementation. Its out of the box responsive/mobile-ready design means that regardless of the platform, the user experience is optimized. Its query engine supports integration with multiple search engines including Elasticsearch, Elastic Cloud, Azure Search, SharePoint Online/On premise, and SOLR (coming soon). This capability provides organizations the flexibility to deploy the best search engine for their specific use case or infrastructure.

Goal 2: Users Should Find Relevant Information on the First Page

Relevancy is a direct result of the availability of data, quality of metadata, and intelligence wrapped around the user’s query.

Our ConnectivityHub provides over 70 out of the box connectors to enterprise systems.  Our experience is that most users need to search three or more sources to ensure relevant content is returned.

Our AutoClassifier provides intelligent tagging, metadata generation, and text analytics to make content findable. Content must be tagged with metadata, and metadata creation is not something you can burden end users with.

As previously mentioned, a major component of our SmartHub is its query engine. The most important key to delivering quality relevancy is to personalize results to users. SmartHub uses machine learning to adjust the relevancy of results per user, ensuring relevant information is delivered based on roles, departments, or other attributes. 

In environments where finding the correct information would require users to enter complex searches, they should be able to ask questions of the search system. SmartHub fully supports Natural Language Query and can interpret questions in natural language and return relevant results based on an understanding of what users need.

Goal 3: Users Should Not Have to Click More than Three Times

This goal builds upon the previous one and points to the key aspects of a search UI. Facets, which also require metadata, must be configurable to present filtering options to users.  As an example, think about the search experience in Amazon. Facets are also a key feature of our SmartHub product.

Our Smart Preview tool provides single click access to a mobile-ready instant preview of an entire document regardless of location, further reducing the clicks required to validate the results presented to users.

Goal 4: Users Should Not Have to Actually Perform a Search

This goal introduces the concept of zero search, or the ability of the system to proactively provide access to content that users consider helpful.

Our SmartHub tool allows end users to specify areas of interest and preferences and automatically delivers content to them based on their input or existing profile. For example, if a presentation that was used to train users in a particular job role has changed, then the updated version of that PPT could be proactively pushed to all users in that role.

Goal 5: 95% of Users Should Succeed in their Searches

For users to succeed in a search, they need to find what they are looking for. That much is clear. The approaches reviewed in the above goals all speak to how this overarching goal can be met. But how can this goal be measured? Tracking and reporting on metrics associated with any search system is key to understanding the success of the system. You must be able to report on at least the major goals outlined above, and hopefully on a large range of other attributes to help manage the search system over time.

Our SmartHub application comes with a built-in, fully featured analytics tracking and reporting application, ensuring organizations have access to reports and metrics across the entire search experience. These metrics are then leveraged to increase the chances of users succeeding in their searches. Prior successful search activity, combined with machine learning, automatically recommends potential searches and information to users.  As users type, suggestions are delivered based on successful searches run by others. On search results pages, additional information is recommended based on what other users have found to be useful.