Finding searches that return nothing

The most useful report in your help centre, and the one almost nobody opens. A list of articles you should write.

The most useful report in your help centre, and the one almost nobody opens. A list of articles you should write.

Build it

1. Open Explore and choose Guide, then Guide - Search.
2. Add the metric Searches.
3. Add Search query as a row.
4. To isolate the fruitless ones, bring in Average number of results and filter it to zero. There is no ready-made "no results" metric, so this is how you get there.
5. Filter to the last month or quarter.
6. Sort by frequency.

What you are looking at

Every row is somebody who came to your help centre, typed what they wanted, and found nothing. In their own words, at the moment they needed it.

No survey and no workshop gives you that quality of information.

Two kinds of row

You have no article

Write one. The search term is your title, near enough.

You have one and it was not found

The more common case, and the more valuable. Your article exists under different words. Add the searched words as labels, or retitle it.

The data can separate them for you, using two of the metrics in that dataset:

• Average number of results is zero. Nothing came back. You have no article.
• Searches happened but Clicks did not. Something came back and nobody thought it was the answer. Your article exists and does not look like what they asked for.

Searching the term yourself still takes seconds and settles any row you are unsure about.

Per language

Where you run more than one, look at each separately. A term searched often in Dutch and never in English tells you precisely what to translate next.

Make it a habit

Once a month, twenty minutes. That single routine keeps a knowledge base aligned with what people ask, which no amount of planning achieves.

See also

• Using labels on articles
• What happens to an untranslated article
• Which knowledge base numbers matter
• Tracking anonymous users