Goal
Find what customers look for in the help centre and do not find, and decide what to fix first.
Inputs
- Failed or zero-result searches: [paste search terms with counts for the period]
- Current article titles: [paste the list of help centre article titles]
- Top ticket reasons for the same period: [if available]
- Product terms customers use versus internal terms: [e.g. "bill" vs "invoice", or none]
Steps
- Normalise the searches: fix spelling, merge plurals and synonyms, keep the original terms as examples.
- Group them into topics and add up the search counts per topic.
- For each topic, check whether an existing article already covers it under a different title.
- Decide the action: write a new article, retitle an existing one, add synonyms or keywords, or expand a section.
Output format
- | Topic | Example searches | Total searches | Matching article (or none) | Action |
- Top ten actions ranked by search count, then by ticket volume if given.
- For each "write new" action, a suggested title in the customer's own words and a five-point outline.
Check before answering
- Search counts are copied from the inputs, never estimated.
- An article is only marked as matching when its title clearly covers the topic.
- Titles use the words customers typed, not internal product names.
- Searches that look like spam or random characters are listed separately and excluded.






