Data
Find deflection opportunities
Which tickets could have been prevented, ranked by volume.
Fill it in
A few months of them.
Not by the tag somebody chose.
Documentation, product change, or nothing.
Your prompt
Find the deflection opportunities. Cluster by CAUSE, from the text, not by the tag. Tags are chosen under time pressure from a dropdown designed a year ago, and the biggest real cluster is always spread across four of them plus Other. For each cluster say what would actually prevent it: a documentation page, a product change, a clearer error message, or nothing. Be willing to say nothing. Most deflection projects fail by writing help articles for problems that needed a fix. Rank by volume times handling time, not by volume. A hundred quick questions cost less than twenty long ones. Say which clusters indicate a bug rather than confusion.
Use Find deflection opportunitiesOpens with everything above already filled in.
Why this works
Tags are chosen under time pressure from a dropdown designed a year ago, so the biggest real cluster is always spread across four of them plus Other. This clusters from the text, ranks by volume times handling time, and is willing to say nothing would prevent a cluster.