Information Scent

Information scent describes the cues a person uses to judge whether following a particular path will lead them toward what they are looking for. The concept comes from information foraging theory, developed by Peter Pirolli and Stuart Card, which models information seeking on the behaviour of animals foraging for food: people follow the trail that appears most promising and abandon a path when the cues weaken.

The practical consequence is that navigation succeeds or fails on whether links communicate their destination. A person scanning a page does not read it; they assess a handful of labels and choose the one that seems closest to their goal. If no label appears promising, they leave, regardless of whether the content they wanted exists somewhere on the site. Content that cannot be predicted from its link is effectively absent.

Strong scent requires specific, descriptive labels using the words the user has in mind. Generic labels such as resources, solutions, products, or more information carry no scent because they could lead anywhere. Internal terminology carries no scent to anyone outside the organization. Labels drawn from the vocabulary customers actually use, and specific enough to distinguish one destination from another, are what allow a confident choice.

Scent must also be continuous along a path. A person who follows a promising link and arrives somewhere that does not obviously relate to their goal has lost the trail, and each such break substantially increases the chance of abandonment. Ensuring that a destination page confirms the expectation the link created, using consistent terminology, is as important as the link itself, and mismatch between the two is a common cause of high exit rates on pages that appear otherwise sound.

The theory also explains behaviour that looks irrational in analytics. Users who back out of a section repeatedly are following weakening scent and retreating to a point where the trail was still strong. Users who abandon after two clicks are not impatient; they have concluded the path is unpromising. Pogo-sticking between search results and pages indicates that the results are not conveying which one answers the question.

Diagnosis is possible with straightforward methods. First click testing reveals whether the initial choice is correct, which is where most scent failures occur. Tree testing isolates whether labels alone communicate the structure. Site search queries indicate what people could not find by browsing, and the terms they use are the vocabulary the navigation should have contained. Session recordings show the retreat patterns directly.

Scent is weakened by competition as well as by vagueness. A page presenting twenty equally weighted links gives no basis for choosing between them, and the effort of evaluating each one is itself a reason to leave. Reducing the number of options at each decision point, and differentiating them visually according to likely relevance, strengthens the trail more reliably than improving the wording of every individual label.

Search and browse are complementary rather than alternative, and both depend on scent. Search results are themselves a set of links whose titles and descriptions must convey what each destination contains, and poor result presentation produces the same abandonment as poor navigation. Sites with strong internal search frequently mask navigation weaknesses, until an analysis of search queries reveals the pattern.

Because the remedy is usually labelling and structure rather than visual design, improvement typically requires research into how customers describe what they want. In practice these problems are identified through UX audit work using first click and tree testing, the vocabulary is established through user research, and the resulting information architecture is implemented as part of product design.