It’s happening…  slowly … but it’s happening. 

Attention in search is finally shifting from a focus on low-level features and relevancy models to looking at the whole user experience for information access. I, for one, am very glad to see this trend. Of all the enterprise technologies out there, few are planted so squarely at the interface of humans and machine as search. And yet, for many users, the search input box and a list of blue links is still the pinnacle of a search user experience – a user interface model that hasn’t changed appreciably in over 10 years. There is room for improvement.

So what exactly is happening?  I’ll point out three things:

1)      New Books on Search User Interfaces and User Experiences

Three recently published books have put some focus on human-computer interaction (HCI) and search. The first book by Ryen White, from Microsoft Research, and Resa Roth, published earlier this year, covers the topic of exploratory search. From the abstract:

Exploratory search has gained prominence in recent years. There is an increased interest from the information retrieval, information science, and human-computer interaction communities in moving beyond the traditional turn-taking interaction model supported by major Web search engines, and toward support for human intelligence amplification and information use.

The second book, Daniel Tunkelang’s on faceted search, looks at a particular interaction pattern that is now a mainstay of most commercial search platforms. Daniel, co-founder and Chief Scientist at Endeca, can speak with some authority on the topic of faceted search since his company was essentially built on the idea.

The third book, and perhaps the most ambitious, is from Marti Hearst, a respected researcher in information retrieval and text mining at UC Berkeley, who has recently released for online reading (print version coming in September) a comprehensive review of search user interface research.

A general theme, with the first two books especially, is on user models for search that are interactive and iterative. They address, in part, the fact that users are not very precise in communicating their information needs in an ad hoc query. While there is some evidence that keyword queries are getting longer, the oft-referenced 2.3 term average query length still demands user experiences that don’t just try to deliver the best possible results on the first attempt, but that can help the user ask a better question through contextual navigation, iterative feedback and refinement options.

2)      New and Evolving Examples Online

Beyond the three more academic works above, there is also evidence that commercial search applications are focusing more on search-based user experience. In a post last month, I referenced a couple Microsoft/FAST customers, Oodle and Globrix, who have put a particular emphasis on user experiences built completely around search. Other sites, like Getty Images’ Catalyst search take advantage of the uniqueness of the domain (image search) to create rich and engaging experiences built on search.

On the wider Web, Microsoft launched the Bing “decision engine” in May with query disambiguation features built in. Even Google has relaxed its keep-it-simple position by adding search options to enrich the user experience. Compared to domain-specific enterprise search applications, the Web search engines are just beginning to dip their toes in the water, but otherwise the same theme exists:  search user experiences that are more interactive, iterative, and conversational.

3)      Search UI Design Patterns

Finally, the past couple years have seen efforts to formalize UI design patterns for search. Peter Morville has championed this idea and posted a nice compendium of discrete search patterns with example screen shots on Flickr (also see his wiki). The idea of cataloging UI patterns for search is so that the good patterns - those that have been proven to work well and to result in a positive user experience - can be promoted and reused. There is also the concept of “anti-patterns” or patterns that have been shown to have a neutral or negative impact on user experience.   (As an aside, Peter’s catalog of patterns focuses on GUI patterns – many of which will be familiar even to non-practitioners. In my post on search and Natural User Interfaces , or NUIs, I mentioned that these new “touch and gesture” UIs do not have established patterns yet for search. It is truly a greenfield and it will be interesting to see what patterns emerge.)

 

As said, I’m a fan of this focus on user experience in search and also of the formalization of best practice design patterns. I’d like to see it all go a little further, though. Having a set of discrete and generic patterns is helpful, but even better will be having best practice patterns that are oriented toward specific business processes where search is used. Understanding these meta patterns in enterprise search is especially important in order to understand user experience differences between search for Research, search for eCommerce, search for Customer Service, search for eDiscovery, etc... Some of these differences are in the search features themselves, others are in how search interfaces with other non-search features and workflow (e.g. shopping carts in eCommerce or communication tools for collaborative research).  For example, the “product comparison” view is something common in eCommerce applications and, while not obviously a search UI element, its rendering is dependent on search results.

In time, I expect these meta patterns to evolve into user experience and UI templates (customizable) that will help organizations quickly stand up search front-ends that take into consideration not just how people search (functional patterns), but why people search (process patterns).

Nate