Conversations in Intercom can be categorized and labelled in three different ways:

Each method has its benefits depending on what’s important to you, but in general we recommend using a combination of all three. Read on to find out about the differences between these methods.

Conversation topics

Conversation topics are the most consistent, automated way for you to categorize every conversation in your workspace.

You define a set of keywords to include (or exclude) and any matching conversation will be automatically added to a topic:

Topics can be applied retroactively to your existing conversations, and require no manual input from your customers or team.

Intercom’s machine learning engine will also suggest new topics for you based on trends in your conversations, opening up potentially missed insights and opportunities.

Conversation topics are purpose built for in-depth reporting too, so you can really dive deep into the subjects your customers chat about:

Read more about conversation topics here.

Conversation data attributes

Conversation data attributes are the most structured way to group your conversations because you can define a specific type like: Text, Numbers, True/false or even a list of predefined values.

They're excellent for categorising all of your conversations for reporting purposes. For example, by type, urgency, product area etc.

Conversation data attributes are perfect for managing your Inbox workflows. For example, if you have a “Priority” attribute it could be set to "high, medium, or low". You can then use these attributes as filters for Inbox views, and conversations will move in and out of views in real time.

Conversation tags and attributes can be applied automatically by bots and Inbox rules, but only conversation attributes can be collected from your customers directly.

This lets you group conversations by values that can’t be discerned from keywords. For example, a customer could choose between “bug report” or “feature request”, or let you know how urgent their request is:

Read more about conversation data attributes here.

Conversation tags

Conversation tags are the most flexible, and specific as they’re applied to individual replies in a conversation.

They’re useful for marking a particular part of the conversation so you can find it later. For example if you’re highlighting beta feedback for your engineers, or need to bookmark conversations affected by a bug or outage.

Tags can also be applied automatically by bots and Inbox rules, but not defined by your customers.

Tags are particularly useful for an individual’s needs, like highlighting a list of conversations to learn from, or collecting a showcase of your team’s top GIFs. 😉

Read more about conversation tags here.

An example conversation

To demonstrate how all these methods can be used in conjunction with one another, let’s take an example conversation:

  1. The teammate has tagged this conversation with ‘“Beta candidate”, something that can only be established during a conversation, by a human. They can now easily pull this conversation up, or share it with the product manager leading the beta.

  2. The customer has defined the conversation attribute: Urgency as “Medium”. The conversation can now be assigned and handled accordingly.

  3. The topic “Task tracking” has been automatically applied based on the keywords the customer used. This conversation can now be explored in reports among others for an overall understanding of how often this topic comes up, and how the team handles it.





  • Mostly automated (can be updated manually if needed).

  • Consistently applied.

  • Applies to historical, and future conversations.

  • Can suggest and track trends you didn’t know about.

  • Ideal for trend reporting.

  • Available in CSV exports.

  • Captures more granular detail that needs a human to understand.

  • Can be provided directly by customers.

  • Perfect for workflow management.

  • Available in CSV exports.

  • Great for marking specific messages in a conversation.

  • Available in CSV exports.

  • Ideal for an individual’s categorization needs.

Did this answer your question?