When you open a new report, you will notice clusters listed along the left-hand side as “Cluster 1,” “Cluster 2,” and so on. Cluster labeling is one of the few manual components to Affinio’s platform, and the process gives you the opportunity to discover what truly makes an audience tick. There’s no set formula for naming clusters, but these five steps will help you get started. As you get familiar with the Affinio platform, you will quickly develop a system that works for you.
Having meaningful cluster names is key because they help set the stage for your strategy formulation.
1. Review the Cluster Summaries
Find a summary for each cluster at the bottom of the main Overview section. These summaries will help you establish the lay of the land for your audience, and give you an idea of what makes each segment unique. These summaries draw the three most nuanced traits to the surface for each cluster. This also makes it a lot easier to decide which cluster you want to name first. Quickly compare the clusters, and name the ones that are easiest for you to identify. Based on your own personal knowledge, some clusters will be faster to name than others.
You can choose a name from one of these top resonating traits, create a hybrid of them all, or simply use these insights as a reference point for your own research. Depending on the cluster, sometimes the summaries alone will be too broad to pick a name from. You’ll also notice these traits listed when you select a cluster from the left menu and click the pencil icon; this is ultimately where you will go to name the cluster.
To demonstrate the process, we ran a Twitter analysis on the Canadian followers of @justinbieber. Within the Cluster Summaries, we noticed that 4/15 clusters were likely to self-describe with the word “justin” in their bios. This is one of those situations where you need to take a closer look at each cluster to find the right name.
Clusters with a higher concentration of shared interests are typically more niche and easier to name. For this reason, I decided to take a look at the Interest Relevance for each cluster and determine which of these “justin” clusters to start with. Interest Relevance is calculated by averaging the Relevance scores of the top 100 interests for each cluster. You'll notice differences between these averages from network to network because some networks are more broadly interconnected than others.
Cluster 1 has a very low Interest Relevance (3.06), and is also the largest cluster in the audience, making up 11.4% of its population. I’m going to label this cluster, General Bieber Fans. Generally speaking, an Interest Relevance score below 10 represents a broad audience that doesn’t share many common interests. Let’s move on to the next “justin” cluster, Cluster 5.
Affinio has identified that the three most distinguishing traits for this cluster are: justin, bieber, and selena. These descriptors alone don’t speak to their personalities, interests, or distinguish them from the other “justin” clusters. Beside the list of traits, you’ll see a mosaic of interests that are unique to this cluster (not shared by the other clusters). The images that stand out to me here are selfies and cartoons, indicating that perhaps this is a younger crowd. But remember, these are the unique interests only and may not necessarily carry the most relevance. Let’s look a little deeper, so we can validate our assumptions. Click on the cluster you want to name from the left menu. Then you’ll be able to view the insights that are specific to this group of people.
2. Explore Cluster Overviews
The first thing I look at in the Cluster Overview is the High Relevance Accounts. Here you will find a mosaic of the top 200 accounts that are most contextually relevant to your cluster. I’ve screen captured just 50 of those interests for Cluster 5 below. You can view more of the cluster’s interests, including categorization, on the Interest page itself.
Followed closely behind Justin Bieber, we see Disney stars like Selena Gomez, Demi Lovato, the Jonas Brothers, Miley Cyrus, and more tenured Disney alumni, such as Britney Spears and Justin Timberlake.
In the Overview section, you can also see what words appear most in the bios of these cluster members, their locations, their top used hashtags, and their top shared links. These insights are expanded upon in the Members and Content sections.
I’m starting to develop a sense of whom these people are, but I still want to prove out what distinguishes them from the other clusters.
3. Compare Similar Clusters
The Compare section allows you to cross-analyze two clusters for their common and differentiating traits. Particularly in situations where multiple clusters seem alike, this can help you distinguish the two quickly. I chose to look at Cluster 5 and Cluster 10.
You can pull a lot of insight from these traits, which you can read about later on in the User Guide, but here I’m going to focus on the contrasting interests. You’ll find these insights when you scroll down a bit further in the Compare section.
While both clusters are following Jazmyn Bieber, Justin Bieber’s 8-year old sister, only Cluster 10 goes as far in Bieber devotion to follow his even younger sibling, Jaxon, as well as the unofficial Global Purpose account. We can also see that Cluster 5 is following Brandi Cyrus, Miley’s older sister, and Shane Dawson, a Youtuber well known for Disney parodies.
4. Round Out Personas
Finally, we look at the ‘Members’ section on the Affinio platform. We leave this as the last step because people’s descriptions of themselves will often differ from their interests. For example, not everyone that fits into a “WWE Fans” cluster will self-describe as a wrestling fan.
In the Members section, you will find information about the cluster members, including their bio keywords, age-groups, genders, locations, rate of engagement, etc. These details can help round out your personas, and validate any assumptions you may have had. It is important to note that depending on the network used to run the report, the features in the ‘Members’ section may differ.
The screenshot below shows demographic profiles which can be used to perfect your cluster personas. You can choose to include these details in your cluster name, or use them to inform other marketing decisions (for example, media buys).
5. Decide on a Name
Now that you've taken some time to get to know this cluster, it's time to name it! Don’t worry, names aren’t permanent. If you decide to rename the cluster at a later date, it’s easy to do. If you're sharing the report and want someone to know your thought process, feel free to leave them a note.
To name a cluster, simply click on the pencil icon beside your selected cluster, which can be found in the column on the left-hand side of your browser.
Using the insights gathered in our example, we know that Cluster 5 is influenced by Disney pop-stars, uses hashtags, and self-identifies with “justin” and “selena.” The majority of this cluster is made up of young women who like Justin Beiber, but they aren’t the most intense Beliebers in this audience. With these insights in mind, I decided to name this cluster, “Mickey Mouse Club.”
Remember, there is no perfect way to name a cluster, and sometimes you will come across audiences that are a little trickier to name than others. Cluster 15 had a relevance score of 13.64 which puts it in the low-medium range for shared interests, making it somewhat difficult to name. Our advice would be to pick a couple of key insights from a cluster (i.e. traits from the summary, tops interests, and so on) and use those as a base for naming a cluster.