Skip to main content

Mining the zero-results search list

Pro

If full-text search is on, your analytics dashboard has a "Search queries with zero results" panel. It's the most underrated data point on the entire site.

When someone types a query and gets no results, they're telling you precisely what topic should have a page and doesn't. They typed it themselves — no surveys, no interpretation. Just write the page.

How to read the list

Sort by query count (most-typed at top). Three patterns to look for:

1. Direct topic gaps

Queries like "rate limit", "self-hosting", "audit log" — concrete topics where your docs say nothing. These are the easy wins. Write a page; the next visitor who searches for it will find it.

2. Wording mismatches

Queries that should have hit a page, but didn't because the page uses different wording. "Customizing fonts" returns nothing because your page is titled "Typography overrides". Either:

  • Rewrite the page title to match what people search for.
  • Add the alternate phrasings somewhere on the page (in a paragraph, in keywords, in a "see also" line).

Do not just stuff keywords. Search engines and your readers both notice.

3. Things you don't support

"Self-hosting", "on-premise", "Java SDK" — queries for features you don't have. Two responses:

  • If it's a hard no: add a brief page that says so. "Docsio doesn't support self-hosting; here's why and what to do instead." Searchers find clarity, you stop the same question coming up in support.
  • If it's "not yet": add it to your roadmap (we recommend pointing at a public roadmap link). Add a page saying "Java SDK isn't available; here's what works in the meantime."

In both cases, the page that says "no" is more helpful than no page at all.

How often to check

Weekly. Don't do it daily — small queries are noisy, and you'll spend time chasing one-offs.

The format we use:

  1. Open analytics → Search queries → Zero results (last 30 days).
  2. Take the top 5 entries.
  3. For each, decide: write a page, edit an existing page, mark as low-priority.
  4. Schedule the writing for the same week.

Twenty minutes a week of this practice will eliminate 80% of repeat support tickets within a quarter.

A note on noise

Some zero-result queries are typos, gibberish, or one-off oddities. Ignore those. The signal is in the queries that show up repeatedly.

A query with 50 hits in 30 days is a strong signal. A query with 1 hit is noise. Set a threshold (we recommend 5) and ignore everything below it.

When the list goes quiet

If you're acting on this data weekly, after a few months the zero-result list will drop to mostly typos and edge cases. That's a good sign — it means your docs cover what visitors look for.

When the list shrinks to noise, the most useful version of this exercise is to look at all queries (not just zero-result) and ask: "Where do most queries cluster, and is the answer one click or three?" If three, simplify the path to that answer.