What Zero-Result Searches Tell You About Your Docs
Knowledge's analytics track the exact queries that returned nothing — turning search data from a vague dashboard into a concrete list of content gaps.
Most "search analytics" in documentation tools describe the search box — how many searches happened, maybe the top queries. What they rarely surface is the more useful signal: which searches came up empty. Knowledge tracks that directly, and it changes what search analytics is actually good for.
The specific data point
Knowledge's analytics library tracks zeroResultQueries — a list of the exact queries that returned nothing, each with a count of how many times that empty search happened. This isn't inferred or approximate; it's confirmed by the analytics module's own test suite, which checks for exactly this shape of data: a query like "refund policy" searched four times with zero results, returned as a structured entry rather than buried in a general search log.
Why this is a different, better signal than "search that works"
A search box that returns something for every query can still be failing your readers if what it returns isn't actually what they needed — but that's a harder signal to extract automatically. A zero-result query is unambiguous: someone searched for something specific, using their own words, and your documentation had nothing that matched. There's no interpretation required to know that's a gap.
The industry framing this maps to directly
The advice that circulates in support and documentation content — organize articles around the language customers actually use, not your internal team's vocabulary — is easy to agree with and hard to act on without data. Zero-result queries are the direct, measurable version of exactly that gap: each one is a literal example of a phrase a real reader used that didn't match anything you've published, in their words rather than a guess at what their words might be.
Turning this into actual content work
The practical workflow is straightforward: pull the zero-result list periodically (surfaced through the analytics endpoint), sort by count to prioritize the queries that keep coming up empty, and check each one against your existing content — sometimes the answer already exists but uses different terminology than the reader searched for, which is a quick fix; sometimes there's a genuine content gap, which is a real article to write. Either way, the zero-result list is telling you exactly where to look next, rather than leaving you to guess at what's missing from a knowledge base that, from the inside, always looks reasonably complete.