Clusters

What is a cluster

Imagine a large bank with a Mortgage product. Inside mortgage there are sub-topics: family mortgage, IT mortgage, refinancing, new-build mortgage. Each sub-topic is a cluster.

Clusters help quickly manage a group of queries and compute analytics not for the product as a whole, but for a specific direction within it.

How to create clusters

Queries can be clustered in four ways. All work on the same queries, so pick the one that fits your task.

1. While adding queries manually

When you add queries through the input field, the system can sort them into existing or new clusters automatically. Good for short lists of a few dozen queries.

2. When importing from a file

If you upload queries from Excel, clustering happens automatically at import time. Good for large arrays of hundreds or thousands of queries.

3. When generating

In the Query generation dialog you pick a mode: automatic clustering (the model puts queries into existing or new clusters) or manual cluster selection (you specify in advance which clusters generated queries go to).

4. From the queries table

In the Queries section, check the rows you need, open the Actions menu and pick Clusters. The dialog immediately shows which clusters the selected queries already belong to and how many. One click on a cluster adds all selected queries to it, a second click removes them. You can also create a new cluster right here: type a name in the Create new cluster field and press Enter.

5. Via MCP

If you use the system through an MCP integration (for example, a chat with an AI assistant), you can ask: «find these queries and cluster them by this criterion». The assistant will pick the right queries from your base and group them itself.

Cluster metrics

Once a cluster is formed, full analytics are available right away. The cluster page shows three key metrics:

• BMR (brand mention rate). What share of AI responses to queries in this cluster mention your brand.
• SOV (share of voice). How often your brand is mentioned compared to competitors inside this cluster.
• CR (citation rate). How often AI services cite sources linked to your brand.

Cluster positioning

A separate Cluster positioning tab shows all clusters on one chart. The horizontal axis is mention sentiment, the vertical axis is BMR (brand mention rate).

The size of each dot is the number of brand mentions in the cluster. The bigger the dot, the more the brand is heard in that topic. One glance shows your strengths, where the brand is mentioned negatively, and where it goes unnoticed.