It’s often said that search is a conversation: a dialog between two participants that can be every bit as rich as human conversation. On one side is the user, with an information need articulated in the form of a query, and on the other side is the system, with its response in the form of a set of search results. Like human conversation, the outcome relies on a shared understanding of intent and context. Together, these elements form a crucial part of the search experience, guiding and shaping the dialog in productive directions.
But the conversational metaphor can only take us so far. There are levels of nuance to the linguistic interaction between human beings that go beyond simple bidirectional exchanges, and likewise, there are patterns and sequences of human information seeking behavior that transcend the conversational metaphor. At this level, we need to take a more holistic approach, and view search from the perspective of stages in an information journey. In this post, we consider one such model of the information journey that is valuable for both its simplicity and utility.
Modelling the information journey
The academic discipline of information retrieval (IR) is fortunate in being blessed with a multitude of models of human information seeking, including some that date as far back as the 1970’s (and beyond). Subsequent work has given rise to additional variations and elaborations, each with their own particular focus or perspective. Over time, these models have succeeded in gaining significant attention from the IR research community in the form of empirical studies and scientific papers.
However, despite their popularity among researchers, few of these models have gained any sort of traction among the UX professionals charged with actually designing the search experiences of tomorrow. To a degree, this may be simply a reflection of imperfect channels of communication between the research and design communities. However, it may also reflect a conceptual gap between research principles on one hand and design practice on the other. Either way, it constitutes a missed opportunity for both sides.
But there is one model (or perhaps more accurately, a metaphor) which in my experience does indeed reliably cross the divide: the chess metaphor. In analyzing chess games, it is common to describe them as consisting of an opening game (an initial set of moves in which players develop their pieces), a middle game (the phase in which most exchanges take place) and an end game (when the encounter is brought to a conclusion). This metaphor, although simple, is remarkably useful in analyzing, describing and designing search experiences. Moreover, it addresses the limitations of the conversational metaphor by framing those exchanges as part of a sequence of higher level phases:
- Opening game: the stage in the user journey where the initial query is articulated
- Middle game: the main phase in the search experience, where information needs are clarified, disambiguated, refined, etc.
- End game: the conclusion to the search journey, where content items are consumed or transactions completed
At this point we should recognize that information seeking journeys can of course be highly iterative and nested, and there is a difference between the stages that characterize a single session and those that take place over prolonged periods of time. And of course, in a game of chess the exchanges are essentially adversarial in nature, but in an information seeking dialog the relationship is (or should be) cooperative. But with those caveats in mind, let’s look at some practical examples.
eCommerce organisations, for example, are typically incentivized to move customers along a ‘sales funnel’ as rapidly as possible. For this reason, their opening games tend to present the minimal barriers to progression, often offering little more than simple keyword search (usually augmented by topical promotions and merchandising):
Simple opening games like this often segue directly into a middle game focused on the use of faceted search to support query refinement, clarification and disambiguation:
We see a similar keyword-based opening game at electronics distributor RS components, but this time with a slightly more sophisticated set of search options:
The use of keyword search as an opening game provides the user with a blank canvas to articulate information needs in whatever manner they see fit. But it doesn’t make sense for all search journeys to begin in such an unconstrained, exploratory manner. Car media classifieds site AutoTrader recognizes that buyer behavior can often be motivated by brand or budget, and offers these choices in the opening game via a form-based dialog:
This strategy invites the user to pre-coordinate their search by articulating certain key constraints, but in this case makes these fields optional so that users wishing simply to browse in an exploratory manner are not unduly impeded.
Hotel and travel sites take this strategy one stage further by adopting a more structured opening game to elicit more of the context that is likely to underpin a meaningful search dialog. Kayak, for example, adopts the view that for a flight search query to be considered complete, it must include details of the origin, destination, travel dates and passenger details:
Once these parameters are entered, the journey segues to a conventional faceted search middle game. However, not all information journeys are this simple. For some search tasks, a degree of specialist knowledge is required by the user to make productive navigational choices. Inevitably, this complexity has to be addressed somewhere – if not in the opening game, then it must be delegated to the middle game. We see an example of this at RS Components: if we articulate an ambiguous query such as ‘LED’, how is this to be interpreted? Rather than simply returning thousands of results in an undifferentiated list, a better approach is to invite the user to disambiguate their query in the opening game, using the auto-suggest:
However, not all users will take advantage of this option, in which case the responsibility now falls on the middle game. This it achieves through a comprehensive category selection dialog:
Only once this disambiguation step is complete can the middle game segue into its conventional faceted search experience.
The challenges posed by complex information needs are particularly salient in professional search applications, i.e. those used by knowledge workers (e.g. lawyers, scientists, librarians, etc.) in the course of their duties. These individuals combine expert knowledge with access to specialist subscription databases, and often engage in complex, high-value search journeys that combine many different types of information seeking behavior. In these cases it is particularly important that the balance between the various phases is appropriate, and the transitions between them are as seamless as possible.
Often it is tempting to ‘front load’ the search experience, on the premise that this will ‘get the user to their destination more rapidly’. It’s an appealing hypothesis, but one that is often applied over zealously. We see this frequently in professional search applications and library catalogs, where the convention is typically to expect users to enter pre-coordinated search queries using metadata that must be known in advance and articulated accurately:
In my experience, there is rarely anything special about such metadata which dictates that it can only be effectively applied in a pre-coordinated manner. If the system supports those indexes, they can just as easily be exposed as faceted search refinements as they can fields in a parametric search form. The difference is, of course, that faceted search refinements will enlighten the user and guide them toward productive navigational choices, but a blind form-fill offers no such support.
I appreciate there is a case to be made about the specialist expertise of those individuals for whom such systems were historically designed, and the opportunity this gives for them to demonstrate that expertise. But for the rest of us, think about it: if you were to enter a darkroom to find your keys, would you start searching and then switch the lights on, or would you first switch on the lights and then start searching?
The separation of concerns between opening, middle and end game is a key element in designing the search experience. Knowing, when, where and how to tackle the complexities of clarification, disambiguation and refinement is a design skill that is often under-appreciated. The chess metaphor provides a simple but remarkably useful framework for understanding the choices and articulating the trade-offs. In our next post, we’ll consider some of the principles that govern how best to propagate the user’s navigational state from one phase to the next, and how those transitions shape the search experience.
Chess Image By dbking (Chess Players in Dupont Circle) [CC BY 2.0 (http://creativecommons.org/licenses/by/2.0)], via Wikimedia Commons