Where card sorting and tree testing capture a snapshot from a research session, usage data analytics tracks how real users behave on an existing information space over time. Several metrics are especially useful to information architects specifically: page views and unique page views reveal which pages or categories are popular versus underperforming; bounce rates on category pages often signal poor labeling, since users arrive and immediately leave without engaging further; entrance rates identify which pages people land on first, where a surprisingly low entrance rate on an important page can point to labeling or findability problems; and site search-query volume reveals what people are searching for, which is a direct window into gaps in the navigation.
Analytics data answers what is happening at scale, but not why. Interpreting it well means pairing it with complementary qualitative research, the kind covered elsewhere in this chapter, rather than treating the numbers alone as a complete explanation.
Analytics can tell you that thousands of people bounced off a category page. It can't tell you whether they left confused, uninterested, or satisfied: that's a different kind of research entirely.
A labeling mismatch is the strongest interpretation. The search-query volume for "gloves" and "scarves" tells you customers want this content and know roughly what to call it, but those exact terms are missing from the category's navigation label and subcategory names. The high bounce rate and unusually low entrance rate relative to search traffic both point the same direction: people aren't finding this category by browsing, because the labels don't match their vocabulary, so they either give up or fall back on search entirely.
This is exactly the kind of finding search-query volume is well suited to reveal, as the source material notes: it's a direct window into gaps in the navigation. But analytics alone can only point at the pattern; it can't confirm that labeling is the actual cause rather than, say, unappealing product images or pricing.A strong complementary method here is an open card sort or free-listing exercise with target customers, asking them what they'd call this category and what related items they'd expect to find grouped with it. If customers consistently produce terms like "gloves" and "scarves" rather than "Winter Accessories," that directly confirms the labeling mismatch the analytics data only suggested, and gives the team concrete replacement terminology to test.