Delivering Successful Search Within the Enterprise

There is a strange phenomenon surrounding enterprise search. Unlike the practical factors of managing enterprise content, search is shrouded by the expectations of magic.  It is the only component of a document management or content management platform that is expected to work flawlessly out of the box, just point-and-index for perfect search results every time. After all, enterprise search engines are just like Web search engines only smaller, right? Wrong and the purpose of the article is to examine why enterprise search is not only different from Web search but has the potential to exceed the discovery success of Web search by more than we can imagine.

Blame it on Google

Most of the time, users are handicapped by not knowing what they need to know to find what they need to know. Consequently, the user gives the search engine overly vague or meaningless queries in order to find a direction for their search. Unfortunately, technology is “face value” and so does not know how to interpret our queries. It does not understand that we can have a single word mean multiple things (order a meal, put things in order) or multiple terms mean the same thing (star: celestial entity, celebrity). Where search engines are methodical, users are emotional. Where search engines are logical, users are heavily influenced by their environment and context of need.

As a result, Web search engines must employ a collection of calculation and programming mechanisms to produce a search results set. These mechanism span user behavior patterns, social media inputs, content change patterns, link qualities both in and out, and much, much more. Google reportedly has over 200 algorithms to accommodate a vast publishing base as well as a user base that span multiple disciplines, needs, assumptions and capabilities.

Perhaps the most fundamental difference between Web search and Enterprise search is reliance on the influence of thought-processing bipeds. Google’s paradigm shifting PageRank algorithm was the first introduction of overt user input with its focus on the number of hyperlinks pointing to a page as a dominant element in relevance calculation. Over the last 10 years, PageRank has been refined to include other user-related factors, i.e. calculation of a shared context between the linking pages to the most recent Panda updates that incorporate click-through, time on page, social sharing and other human action indicators.

The ultra-sophistication of Web search has produced a cadre of lazy information seekers that are no longer able to construct useful queries, do not explore beyond the first few results, are distracted by the myriad of choices presented and carry this distraction to their destinations where they afford precious little time evaluating the landscape before jumping back to another search result or another search all together. (1)

Searchers carry their Web search expectations inside their workplace. Inside the enterprise, they expect what they get from Google. However, their enterprise search engine does not have the same level of care, configuration or data from which to construct meaningful results from meaningless queries. This all too often leads to the purchase of more bright shiny software that also fails to deliver the “Google experience” that is only to be had with the same attention to detail on a smaller scale.

Why Bother?

The successful configuration of enterprise search is significant because searchers within the enterprise spend more time on failed searches due to feeling certain that the document “is out there somewhere.” An Association for Information and Image Management (AIIM) study found that Ford Motor Company knowledge workers spend 5-15% of their time on non-productive information related activities (2).

Content management and document management software often contain search appliances that can be as effective, if not more so, than their Web counterparts. If enterprise search is so important and often listed as a critical factor in intranet design (3), why is it often left under-done or undone completely? The user experience that is the foundation of extremely successful Web search is nowhere to be found within the enterprise and that is its undoing.

In the same AIIM Study, an IT Manager from a Fortune 500 company communications firm estimates that by improving search and retrieval systems for just the firm’s 4000 engineers would recover this cost within a month and would contribute $2 million monthly productivity gain thereafter. And if that is not enough, giving the IT staff the ability to walk the halls freely without hearing “search sucks” should be added incentive.

The first step towards successful, cost-saving search within the enterprise is to treat it as a user experience because that is where it starts, with the users, then the technology.

Enterprise Search and the User Experience

Successful enterprise search is not achieved through spontaneous assembly of features and programming. It results from a clear understanding of search behavior inside the fire wall so that the right search features will be mapped to specific needs and behaviors. This is accompanied by an assessment of the types of content, content quality and machine readable elements to produce significant results.

Achieving productive enterprise search involves more than search analytics, corporate taxonomy or managed metadata. It is a collaborative effort between information architect, developer and designer to create a search experience where colleagues can find what they need, when they need it. Below is a three part approach to ensure enterprise search success.
Part 1: Define Problem Space: Existing search analytics is a great starting point IF the appliance is configured to produce them and IF the server logs can be obtain. Often, this is not the case within the enterprise. And, noted usability expert Jared Spool rightly points out that website analytics tell us what happened and not why.

However, the system users are readily available through online survey, in-person interviews or focus groups and, not surprising, willing to pour forth on their problems finding the information that they need through search. The resulting direct, antidotal evidence of the problem space is the foundation for measuring the return on investment.

Part 2: Shared understanding of the technical and content landscapes: Enterprise search engines come with a broad array of features not all of which are suitable for the enterprise culture, content or behavior. Most enterprises try to boil the ocean by using every tool in the box or merely point the crawler to the content, build the index and be done with it. Both are suboptimal to bringing a user-centered search solution approach that will ensure:

  • Content repositories are structured in a discoverable manner
  • Content is described in a meaningful, machine readable way
  • Content strategy to build content relationship models that facilitate discovery and exploration
  • Users that have an opportunity to configure results through features such as: self-tagging (folksonomy), Best Bets, results filtering and facets, federated search across structured repositories, unstructured content and the Web
  • Practical governance that includes content creator education
  • Security settings that are enterprise-specific to protect sensitive content

Part 3: User-Centered Design for maximum feedback: Internal users tend to look longer because their search is more specific and they have a higher confidence that what they are looking for exists out there, somewhere. This makes it more essential to “design a search experience” that gives enterprise searchers the tools they need to maximize engagement and query refinement opportunities while minimizing floundering around and encountering multiple dead ends. A zero results search page is a road to the perdition of information need abandonment, interrupting colleagues for direction, using bad or outdated resources and clogged Inboxes from emailing large documents.

Available tools such as user controls (i.e. filters, facets and clustering) and user assistance (subscription alerts, “Did You Mean” spelling correction, editorialized Best Bets results, search suggest (popular successful related queries) and augmented search results afford opportunities for the searcher to directly engage with the search engine in refining their query for success or discovering additional information.

Conclusion

Regardless of the search appliance in question, taking a user-centered approach to enterprise search configuration and management will produce cost-savings and increased productivity. Including a user-experience professional in planning and deployment is low cost investment that produces big results.

  1. Using the Internet: Skill Related Problems in User Online Behavior; van Deursen & van Dijk; 2009
  2. http://ejitime.com/materials/IDC%20on%20The%20High%20Cost%20Of%20Not%20Finding%20Information.pdf
  3. http://www.useit.com/alertbox/intranet_design.html
About Marianne Sweeny
Marianne Sweeny

Marianne Sweeny started out as an information architect in 1997. She transitioned to optimizing search systems after studying information retrieval at the University of Washington. Through her company Daedalus Information Systems, she offers strategic search consulting for organic, paid and enterprise search systems. She can be reached at msweeny@speakeasy.net.