Principles of Hierarchical Topics
  • 11 Feb 2022
  • 1 Minute To Read
  • PDF

Principles of Hierarchical Topics

  • PDF

There are a few principles of Hierarchical Topics to keep in mind as you build and use Topics.

Below is a list of 7 principles to keep in mind when creating or editing your Topics.


To learn how Topics are used in Conversations, see Add Topics to a Conversation.

Principles of Good Hierarchies

To get the best out of Topics, we recommend the following rules of thumb.

  • Gladly recommends that you create Topics using the Gladly interface. However, in the event that you have > 50 Topics to import, you can fill out the Topics tab in this template and send it to your Gladly implementations team to help with the import.

Every Conversation should have a Topic

  • Every Conversation should have a Topic that can be applied to it. Agents shouldn’t be left with a Conversation for there’s no Topic to apply to it. This may mean including “Other” or “Misc” Topics throughout your hierarchy.

Watch out for similar Topics

Topics should be mutually exclusive, i.e. non-overlapping. If two Topics could apply to a conversation, agents will be confused about which to use, and your data will be messy.

  • For example, let’s say there’s a Topic of “Return” that includes sub-Topics “Poor Fit” and “Wrong Size”. An item that’s a poor fit may also be the wrong size, so which should the agent apply?
  • That being said, a conversation may be about multiple things. For example, a person may contact you about their current order status and how long shipping takes. In this case, multiple Topics may be applied to the conversation, but that’s because the conversation is about multiple things, not because the Topics themselves overlap.

Avoid creating too many sub-Topics

  • Aim for 2–3 levels of hierarchy. More than that and your Topics will get too granular, making it harder for agents to select the right Topic, and making the data harder to analyze. You’ll get topics that are only applied to a few conversations, which isn’t likely to surface useful insights.
  • If a Topic has too many sub-Topics, agents will have a hard time finding the right one. 5–15 is a good range. More than that the list becomes hard to choose from.
    • Conversely, if a Topic only has 1 or 2 sub-Topics, it may be too narrow. Consider combining it with a sibling, or not having any sub-Topics.



Was this article helpful?