Findings are not limited by a number of predefined clusters.
You may find clusters can appear smaller when you run your report with Interconnections segmentation, identifying who is connected within an audience. In this case, as there must be a connection between the affinities, there are always very small clusters that fall under the threshold of 1%, thus are not represented in the cluster.
Our data team found that clustering based on this method creates the most consistent and relevant clusters, as well as identifying more unique characteristics for each cluster.
Once we have identified the clusters/groups we can ask “how do they know each other”, i.e. what are the hidden trends that the platform has revealed, including interests, psychographics and influencers.
Results are not skewed by affinities to major handles who are not a part of the audience (e.g an interest in love by Katy Perry will not create a large single cluster grouping everyone together who is connected to her), as it clusters the audience in terms of their own connection to each other, who follows who.
You can rerun the report and select Affinities segmentation to cluster the audience in terms of similar account they follow. This usually provides fewer clusters and a larger % within each.
For more information about clustering your audience, view this guide.