Automated Key Topics

The donut chart (or sunburst chart) consist of two levels: 

Level 1 - The key topic clusters

These are the overall topics that you customers are calling about. Key topics are always boiled down to one single word e.g.

• Payment

• Delivery

• Import

• Billing

• Internet

• Printer

• etc.

Each cluster contains multiple key topics. 

Level 2 - The detailed key topic clusters

These are the details of a key topic cluster that will elaborate more on the why the customer called with this e.g.

• Questions about billing

• Wrong payment

• Questions about missing delivery

• Import of domestic animals

• Router is not working

• etc.
  

As with key topic clusters, these clusters will hold multiple detailed key topics.  

To investigate a cluster, simply click the tile and you will move to level 2 containing the detailed key topic clusters. 

Click the arrow to move back to level 1. 

Each of the tiles in the donut chart represents a semantic clustering of key topics.

The name of the cluster is defined as the name of the most frequent key topic in that cluster.

So a cluster called "payment" may also include conversations with key topics like "invoice", "billing", etc. 

The tabel shows you more data about the individual cluster: 

• Key topic clusters
  Name of the cluster
  

• Conversations
  Number of conversations in cluster  
  

• Percentage
  The percentage of the conversations in the period

  • On level 2 it shows both percentage of conversations inside the cluster, and percentage of the total amount of conversations in the selected period. 
    

• vs. Expected
  Shows if there are more or less conversations in the cluster compared to expected based on the previous 4 weeks.
  Capturi will mark clusters with a speaker icon 📣 if the difference is large enough to be statistically significant.  

Additional information about the conversations in the period. 

When selecting a cluster, the info boxes will switch to showing info for the conversations inside that specific cluster and a benchmark to the average of all conversations in the period. 

Randomly selected examples of key topic summaries.

Filters can be applied to the key topics page in the same way as you know it from the rest of the platform.

Each time you apply a filter, a new clustering is generated in a few seconds.

This gives you a unique opportunity to get detailed insights on various subsets of data like:

• Queues 

• Conversations with a given NPS interval 

• Conversations with an significantly longer duration 

• etc.