The implications of Information Foraging Theory on designing user-centered websites have not gone unnoticed. Jakob Nielsen and Jared Spool, among others, have put forth considered recommendations on how to enhance information scent on the web. Most of their guidelines, however, tend to assume that the designer has direct control over the explicit words used in the interface. While this is certainly the case for browse-based websites dependent on site-wide navigation and hyperlinks, it breaks down for search interfaces where both content and navigation are completely dynamic.
While the principles for amplifying information scent in search-based interfaces are complimentary to those of browse-based models, they are yet distinct from them. Understanding how information scent applies to search first requires an understanding of human search behavior and the factors that affect it.
People are just like bears (only less fur)
In her landmark 1989 paper, Marcia Bates outlined search as an evolutionary process. Users often begin with a general query, glean a few nuggets from the initial results, reformulate their query based on that new knowledge, and then repeat this process. Like a bear foraging for food in the forest, knowledge seekers tend to rapidly migrate from one patch of information to the next.
While this iterative behavior is true of virtually everyone using search, there are two key factors that distinguish some users from others: domain expertise and search expertise (though John Ferrara has identified several additional factors). Some websites, for example, may be able to assume that users are highly literate in a specific topic, while other websites may need to design for a range of expertise in a variety of subjects (the case for web search engines). In addition, users experienced at using search interfaces will be more capable of utilizing sophisticated search tools, but less experienced users will demand less complexity.
Though domain and search expertise separate some users from others, a given user may have different goals at different times. The two primary types of goals are recall and exploration. Recall involves a straightforward retrieval of a specific fact or document (for instance, “what it the population of Brazil?”), and can generally be accomplished in a short amount of time. Exploration, on the other hand, is a more subjective process. Choosing where to go on holiday, for example, is a complex question that may take hours or even days to decide.
Carrots and sticks: designing for information scent
Understanding the iterative nature of search and the contexts from which users operate is the foundation for knowing how to effectively harness information scent to improve the usability of search. Above all else, Information Foraging Theory has taught us that users need to feel as if they are always “getting warmer.” As a user searches, information scent must grow increasingly poignant, emanating a feeling of progress to the user. When information scent is strong, users are confident that they’re headed in the right direction. When it’s weak, users may be uncertain of what to next, or they may abandon their search altogether.
When information scent is strong, users are confident that they’re headed in the right direction. When it’s weak, users may be uncertain of what to next, or they may abandon their search altogether.
There are many practical methods for increasing information scent in search. Some of them bear resemblance to Nielsen and Spool’s original recommendations, but are deserving of further elaboration in the context of search interfaces. We will trace the user’s journey from the searchbox, to the list of search results, and end with query refinement using faceted navigation.
In order for a user to have a successful search experience, he must first locate the searchbox and successfully enter a query. These two obvious requirements lead us to our first two design recommendations.
The searchbox should look like a searchbox
Cute attempts to drastically re-style the searchbox usually end in failure. The universal language of the searchbox consists of a border, white background, and a corresponding button that says “search.” In addition to expecting the searchbox to look a certain way, users have also come to expect it in a particular location: the top right corner of the page. The further one deviates from this expected appearance and placement, the more one risks that users will not actually discover the searchbox.
Provide as-you-type query suggestions
Whether the subject is a particular Icelandic volcano or the president of Iran, users are often not sure exactly what to type in order to find what they’re looking for. A little help can go a long way in getting the user off to the right start. As-you-type query suggestions reduce spelling errors and, equally important, give users a sense of confidence that they have entered a dependable query.
Assuming that the user found the searchbox and managed to enter a query, she will then be presented with a set of results matching that query. Consisting of at least a title and description, search results are typically dense with information. The challenge becomes separating the signal from the noise.
Indicate the number of results matching the query
The number of matching results has a significant impact on the user’s confidence in his query. If he sees that a large number of results have been returned, he can safely assume that his query is adequate, whereas only a handful of results may be an indicator that he may have misspelled a word or is simply searching for something that doesn’t exist.
Use descriptive titles
In order for users to detect information scent in search results, the results must be digestible at a glance. Titles are usually the first recipients of the user’s attention, so it’s important that they accurately describe the content that they represent. Avoid using file names as titles, which are often cryptic and usually contain little information scent.
Highlight matching words
In addition to descriptive titles, hit highlighting is one of the most helpful cues on the search results page, making queried words immediately stand out to the user. The user can quickly evaluate the list of results by simply observing the greatest concentration of highlighted words on the page.
Make visited links discernible from unvisited links
A visual indicator of which pages have already been visited provides useful scent to the user. Whether she is trying to re-find a page she found yesterday, or trying to avoid duplicating her efforts, a visited link color is very helpful.
More detail for top results, less detail for the rest
One of Peter Pirolli‘s interesting discoveries is that users tend to prefer more verbose results in some circumstances (when there is no time constraint or when there are few results to choose from), and more concise descriptions at other times (under a deadline or when there are many results). How can these opposing cases be reconciled? An ideal compromise is the best first pattern, in which extensive metadata is presented for the top one to three results, while more concise views are provided for all of the subsequent results.
Avoid zero results
A search result page that has no results is a serious roadblock to users. It will either delay their journey, or cause them to give up completely. It’s important to do everything possible to avoid zero result pages from ever occurring. Two helpful tools are automatic spelling corrections and synonym dictionaries. If the user has obviously misspelled a word in the query that would yield zero results, it’s best to automatically correct the spelling for the user, being careful to notify the user of the modified query.
So the user entered a query and glanced over the first set of results. What now? If the user already found what he was looking for, then job done. But chances are he still has a long way to go. Faceted navigation is the best available tool for facilitating the evolutionary flow of search. It both helps the user understand the shape of the data, and gives him the ability to drill down to a very specific slice of the results.
Show the number of matching results for each filter
Showing result counts for filters helps users understand the overall composition of the results. They provide cues that feed into our decision-making process, influencing how we decide to further slice the data. In addition to providing a numeric count, subtle visual indicators such as horizontal bars can make the distribution of results even more immediately obvious to the user.
Use breadcrumbs to indicate the user’s query and applied filters
In addition to choosing where to go next, users need to know where they are currently and how they got there. Breadcrumbs provide this trail, and also enable users to quickly get back in the event of having taken a wrong turn. Each breadcrumb should be independently removable, while Greg Nudelman has outlined an even more forgiving breadcrumb that allows for the swapping of one filter for a related one.
Make metadata clickable
When filterable metadata is shown for a search result, that metadata should be clickable to allow for organic filtering of the results. For example, when searching a catalogue of books that presents the author alongside each result, clicking on the author’s name should cause that author to be added to the query as a filter.
Find ways to meaningfully visualize facets
Many facets lend themselves to a certain kind of presentation. Whether the facet consists of cities, prices, keywords or categories, there is probably a corresponding visualisation well suited for each, from a map to a slider to plain text. Effective visualisations are ones that make the data tangible and easy to comprehend.
A fairytale ending
Information scent plays a valuable role in making the digital landscape easier to traverse. By applying principles that amplify information scent, we can help facilitate a state of flow that enables users to engage in productive, frictionless, enjoyable search experiences.