How to Get the Most Out of Enterprise Knowledge Assets Using Search

The demise of the Google Search Appliance (announced in early 2016) marked the end of the flawed dream of off-the-shelf, one-size-mostly-fits-all enterprise search solutions. In its place is the promise of rich search-based applications that can search, explore and analyze enterprise information.

Download the authoritative guide: Big Data: Mining Data for Revenue

Enterprise.searchmarket2

Knowledge plus data are core assets of any enterprise. A modern search app can utilize that knowledge so employees can find what originated and what it looks like.

You can build a search app far more powerful than Google and focused on your users and their domains, roles and tasks. We'll explain how.

The demise of the Google Search Appliance (announced in early 2016) marked the end of the flawed dream of off-the-shelf, one-size-mostly-fits-all enterprise search solutions. In its place is the promise of rich search-based applications that can search, explore and analyze enterprise information. This isn’t just about search in traditional documents; it’s also about data from databases and facts and relationships in semantic graphs. Bring together all the information from silos, plus context from the world and from your domain, then layer on a set of apps that search, explore and analyze–this is the new world of enterprise search.

To help get you started, eWEEK offers this Data Point article, which consists of 10 things you should know when starting an enterprise search project, based on tips from long-time search aficionado Stephen Buxton. Buxton spent many years as Product Manager for Search at MarkLogic and at Oracle and now runs a consulting business.

Data Point No. 1: Be ambitious: You can do better than Google
Many search teams are intimidated by the success of Google in web search. But you can do better! Google has excellent technology and hardware, an army of smart people  and access to billions of searches every day. But you have all the enterprise’s information – traditional documents, internal web sites, and data on your products and customers. And you have a deep understanding of your users’ domain, their roles and tasks, and you get to talk to them in person. While Google can go broad on the web, your team can go deep in the enterprise.

Data Point No. 2:  Process your searches
When a user types a few words into a text box, it’s a challenge for the app to figure out what that user is actually looking for. Use search techniques such as stemming, synonyms, broader/narrower terms to ensure you find all the right information. Then apply a semantic graph to use real-world context – just like asking a knowledgeable librarian to find a book. Help the user tell you more about what they want with spell-check, auto-complete, and suggestions (“Did you mean … ?”). Use facets to both give an at-a-glance overview of results and to guide the user some useful ways to scope the search.

Data Point No. 3:  Process the data
Annotate your data so it can be easily found. Add metadata: published date, author, key terms, and provenance (where did this come from?). Find and annotate named entities: people, places, company names, credit card numbers, and so on. And mark each document as more or less authoritative, so that my search can return corporate edicts ahead of cat videos.

Data Point No. 4:  Make the results pretty and useful
The search results page doesn’t have to be a list of blue links. Make each result as information-rich as possible: a link, a snippet (a piece of relevant text with highlighted terms), and some metadata. For the most important results, consider using a card with a picture. When cards have a natural grouping, display them together in a carousel. Add information about the overall results set – an info panel that describes the main terms in the search; facets; maybe some graphs and charts.

Data Point No. 5:  Don’t make me do the same search every day
For each user’s favorite set of searches, show a summary of results on his dashboard when he first logs in. If he needs to know about some new information as soon as it comes in – say, whenever there’s new research on the effect of painkillers on tennis elbow – set up an alert to send him an email or a text message as soon as information appears that matches one of his searches.

Data Point No. 6:  Track everything
Track all searches, and track the results of those searches. Are searches succeeding? “Success” means users get results back quickly, click on the top result, read it, then move on. If searches are failing, then tweak the app, the processing of searche, and/or the data. Is there a spike in searches for, say, David Bowie? If so, consider creating a special info panel or landing page for Bowie and re-directing searches there. Make a punchlist each day and prioritize carefully.

Data Point No. 7:  Repeat
Users’ searches change every day. So does the data, and the world that gives your search some context. Improve the search each day and always be on the lookout for ways to delight your users – with more personalization, better info visualization and smarter recommendations.

Data Point No. 8:  Don’t forget about security
As soon as enterprise search goes beyond the basics of searching the open-to-everyone parts of the intranet, it has to respect corporate privacy and security. Ideally, your search app will slot right in to your existing security infrastructure.

Data Point No. 9:  This sounds hard – Google doesn’t do all that!
Google web search doesn’t have to deal with security, but enterprise search does. As for the other tips – yes, Google expands searches; paints a rich results page with information about each result and about the overall results set; and tracks every search and outcome, to make searches better. The one thing Google doesn’t do is annotate the data; instead it documents how website developers should annotate their own data so Google can find it.

Data Point No. 10:  I’m in! What do I do next?
Find a search platform that supports a search application, not just an open search engine. The platform must support all the techniques we discussed, and it must be able to pull together all kinds of information – traditional documents, data and facts/relationships (semantic graphs) – so you can build a search app that uses all the information in your enterprise.

Chris Preimesberger

Chris J. Preimesberger

Chris J. Preimesberger is Editor-in-Chief of eWEEK and responsible for all the publication's coverage. In his 15 years and more than 4,000 articles at eWEEK, he has distinguished himself in reporting...