Building a tag taxonomy that survives growth

The first tag list a support team builds usually reflects the issues they happened to see in their first few months. That works fine at small scale. At higher volume, those early decisions start to constrain reporting, hide real trends, and pull agents into long debates about which tag belongs where.

What a durable taxonomy looks like

1. It is shallow. Two levels deep, sometimes three, but never the kind of nesting that needs a diagram to explain.
2. It is balanced. No single tag accounts for more than a quarter of total volume on a normal week. If one does, it is hiding several real categories underneath.
3. It is stable. Tag definitions do not change quietly. Any change is treated as a small migration with a clear before and after.
4. It is owned. One person is responsible for the list and approves additions, mergers, and retirements.

Signals that the taxonomy is breaking

• The same conversation gets tagged differently by different agents and nobody is wrong.
• An 'other' or 'general' tag is the largest or second-largest category.
• Reports require ad-hoc grouping every month because the raw tag distribution is too noisy to read.