That will be the title of one of our chapters, and I'm digging into it now. And I'm finding it a bit surprising.

Unlike other topics we're covering in the book—using search analytics to improve content development, metadata, search results, and so on—there seems to be comparatively little to cover here. While there's nothing wrong with short chapters, I do want to make sure that we're not missing something big. And obvious.

So here's what I've come up with so far, in a rough mix of outline and prose. If you have any experience in using SA to enhance your site's navigation, we'd appreciate your input, not to mention any additions you think we should incorporate to the list below (full credit due, naturally):

The Caveats

• Searching and browsing aren't the same; everything we say here should be taken with a huge grain of salt
• But sometimes what people search can mirror what they browse, and vice versa; so while this chapter is full of conjecture, it's worth considering what benefits search analytics might hold for us as we design means for navigating our sites

Learning from Failures

• Analyzing failed navigation sessions that lead to searches. Use referrer data to see which navigational pages—aside from the main page—originate queries. Analyze the queries associated with each of those pages. If common topics emerge, create ad hoc links to related information from those pages to enable contextual navigation. If types of content emerge, then consider creating a tight link between the type of originating page (e.g., product description pages) and the likely type of destination pages (e.g., product review pages). Now you're building content models that encode contextual navigation into repeatable logic.
• Analyzing failed search sessions to see where users navigated next. This is far more conjectural, and I have no supporting data. But it seems worth examining frequently 0 results queries, see if there are patterns to where the originated (e.g., content types A, B, and C), and see if there are patterns to where users navigated after the failed search (e.g., content types X, Y, and Z). If patterns emerge on either side, it might indicate the need for creating contextual navigation between these types of content (e.g., content types A, B and C should link to X, Y, and Z).

Looking for Opportunities

• Improving on existing navigation. Search analytics will help you improve some existing navigation systems, and less so with others.
  • Site indices: Michigan State University ingeniously uses the same data that backs its Best Bet search results to serve as an A-Z site index. Essentially, frequent queries are used as index entries, and are linked to best bets. The entries aren't as comprehensive as those in a manually developed index, but are broad in scope and cover the most popular queries, and are thus much less expensive to maintain.
  • Site hierarchy (and, by extension, tables of contents): The benefit of SA is less obvious here. Common queries are often navigational in nature; major queries or clusters of queries might suggest category or sub-category labels that should be included in a hierarchy or taxonomy. So at best SA can help validate and perhaps identify gaps in an existing hierarchy.
  • FAQs and Guides: Clearly, FAQs and guides should be developed in response to common queries, both in terms of general topics and specific questions addressed within each.
  • Wizards: Unclear; might their flow/sequence be informed by the corresponding flow/sequence of search sessions?
  • Contextual navigation: Already addressed in the Learning from Failures section.
• Leveraging sessions to better understand navigation. It seems that session analysis—looking at how queries develop and change in the course of a single search session—might unlock some clues about how users seek to navigate. So, for example, we might see patterns like this one: "car model name"=>"that model's pricing"=>"financing the car". Such patterns might suggest models for contextual navigation (and for wizard design). The challenge is detecting these patterns. I think it can be done by searching for some reasonable number of common queries within sessions, and then analyzing those sessions for patterns. It's potentially a lot of work with little payoff, but for certain types of sites with fairly consistent query types—Netflix, for example—it might be cost effective, as patterns would emerge relatively quickly.

Digging into the Long Tail

Everything mentioned above pertains to the natural starting point—the "Short Head". Can the same sorts of analyses be applied to Long Tail data? It seems entirely possible to work with random samples of such data, and we know that queries vary between the two (Short Head queries are typically more navigational, Long Tail more research-oriented). If nothing else, one could analyze Long Tail data to see if content types emerge that aren't reflected in existing content models—essentially, content model gap analysis.

OK, this is what I have; I'd love any feedback on how search analytics can help improve navigation.

—Lou Rosenfeld