How to conduct clustering analysis to understand audience segments.

The people who you want your news and information to affect — your intended audience — are not a monolith. Therefore, as you start conducting audience research to understand them, you’ll need ways to recognize their differences and learn about specific groups of people within it.

Audience researchers can understand variations within their intended audiences through utilizing plots and clusters in their research according to common characteristics.

This can enable researchers to understand how different segments of their intended audience interact with information based on their mindsets, behaviors, and preferences, and lead researchers to identify areas for intervention and opportunity.

Jump: HOW to segment | HOW to design a survey | HOW to cluster survey results | HOW to use the patterns revealed

Psychological mindset clustering, which groups people according to similar attitudes and worldviews, has enabled our team to learn about groups of individuals who share similar thinking patterns.

This process is useful when designing targeted news and information reporting products to ensure that your information meets the needs of a range of people within your intended audience, not just one imagined persona or typical person within it.

Like in other steps of the Gazzetta approach, employing this method requires having a clear research goal, which should be informed by prior research, like a research review and other data you have gathered about your intended audience. It should also account for how information needs models apply to your intended audience.

Utilizing these prior methods will help to scope objectives and questions for this phase of your research:

• In a closed information environment, how do people form connections and a sense of belonging?
• How do they see themselves within broader society?
• Are they optimistic or pessimistic about their future?

In essence, what are the factors that determine their mindsets about their own lives, society, and the future?

Understanding different attitudes like these among your intended audience and how their information needs and information seeking behaviors differ can help craft reporting and dissemination methods to reach a broader range of people.

Using mindset clustering, as a data-informed approach, can help you move beyond generalizations and design more effective and inclusive solutions by addressing shared and unique challenges faced by groups within your intended audiences.

Below, we share steps on conducting your own psychological mindset cluster analysis.

Step One: Determine how to segment your audience.

Before conducting this type of analysis, you’ll need to have gathered data on your intended audience, through a survey or other methods, such as interviewing, to create psychological segments. Other types of segmentation, such as demographic and socioeconomic, are also possible for this analysis and may be easier to generate data for, like by utilizing data from prior research reviews on your intended audience.

If you can’t interact with your intended audience to gather data, you could create your own personas to represent them, but this is not a scientific method and would be subject to your own assumptions and biases.

You can segment your data on your intended audience through several lenses:

• Demographic segment (e.g., gender, age, ethnicity): Segmenting based on demographics provides a basic understanding of a population and is often easily accessible. However, it can oversimplify complex behaviors and needs. For example, individuals of the same age do not always share the same worldview, because their socioeconomic conditions may be different.
• Socioeconomic segment (e.g., education, income): Segmenting on socioeconomic factors can help understand things like purchasing power, access to services, and social influences, and provides a more nuanced view than demographics alone. However, individuals are more complex, and understanding the “why” behind their views is important and requires more than data on their economic and social resources.
• Psychological segment (e.g., openness to change): Psychological segment groups offer deep understanding of audience motivations and preferences. They allow for highly targeted messaging and the development of content that resonates on a personal level. Psychological segments cut across demographics and socioeconomics to understand deep ways of thinking, often shaped by life experiences that have formed an individual’s worldview and outlook on life.

Step Two: Design a mindset survey.

After you’ve determined how to segment your audience, it’s now time to design your mindset survey. Designing your survey should take into account your background research and research goals, with the intent to build empathy with your intended audience and understand who they are and what kinds of information they need.

For example, if you’re producing reporting for refugees on job availability, your survey might assess their attitudes about hard work and education, how work relates to their ideas of social class, manual labor compared to desk jobs, their ideas about the future, and other relevant inquiries. However, if your reporting will focus on broader topics, as was the case for our prior Gazzetta project defining more general attitudes may be suitable.

In one of our projects, we designed our survey around binary statements on two major dimensions: Progressive versus Conservative and Communal versus Individualistic.

This was done so that we could cluster respondents into a limited number of groups. Having too many dimensions may create problems when trying to identify clusters or create personas.

Our approach was similar to a personality assessment, but tailored to our specific research objectives. We began by brainstorming a range of attitudes:

• Rootedness vs. mobility
• Tradition vs. modernity
• Self-reliance vs. community support
• Present focus vs. future orientation
• Risk averse vs. opportunity seeking
• Focus on self vs. focus on others
• Active searching vs. passive receiving of information.

After mapping these out, we condensed them all into our two dimensions: Progressive versus Conservative and Communal versus Individualistic.

We did this to ensure that our survey was short and clear enough for respondents to complete, and that our results could be comprehensible to us and able to be clustered. This isn’t hard science; the ranges we used were a choice, and yours could be completely different based on your research goals.

When building our survey, we designed it so that survey participants assessed their alignments with each statement pair within each dimension, and selected the option they felt best represented by.

Here’s how that looked in practice in our mindset survey questions:

Survey instructions: Please score 1-5 according to your self-assessment: 1 means you lean strongly toward A, and a score of 5 means you lean strongly toward B. There is no correct answer.

Conservative versus Progressive:

Which statement do you agree with more? A. Even after living in a new place for a long time, I still don't feel at home. B. I adapt quickly to new places and immediately feel at home.

Which statement do you agree with more? A. I prefer a stable life. B. I prefer adventure and trying different ways of living.

Which statement do you agree with more? A. All situations in life exist for a reason, I should view them with equanimity. B. When encountering unreasonable situations, I should strive for the rights I deserve.

Individualistic versus Communal:

Which statement do you agree with more? A. Life feels quite lonely, I'm just by myself. B. My friends make me feel warm and happy.

Which statement do you agree with more? A. I prefer to solve problems on my own, I don't like seeking help. B. When I encounter problems, I seek help from friends and colleagues.

Which statement do you agree with more? A. I focus more on current problems. B. I prefer planning and preparing for long-term goals.

Which statement do you agree with more? A. I feel lost in life, without goals. B. My life has hope, and I'm moving toward my goals day by day.

Step Three: Conduct a cluster analysis to discover audience segments.

After designing and disseminating your mindset survey, you will have to decipher its results. In our prior project, we did this by feeding our data to a machine learning algorithm to run a KMeans clustering analysis.

For our survey results, participant responses received numerical values corresponding to their answers, leading to vectors such as (3, 4, 5, 1, 2, 3, 4). Similar data points were grouped into clusters, resulting in a basis for segmenting the audience.

Using KMeans allowed us to identify patterns indicative of similar mindsets. We segmented the audience into four distinct persona groups along our two key dimensions:

• Progressive versus Conservative: Reflecting adaptability, openness to change, and growth versus preference for stability and acceptance.
• Communal versus. Individualistic: Reflecting community-connected, future-focused mindsets versus self-reliance and present-focused tendencies.

Below is what our data looked like after the KMeans clustering. You can see that we did not end up with a cluster of “Progressive and Individualistic,” and the ratio of respondents in each cluster was not even. In fact, the plurality of our respondents were clustered as Progressive and in the middle of Communal and Individualistic, and the vast majority are on the Communal end of the dimensions.

Our Conservative and Individualistic cluster was the smallest, but these responses were some of the strongest on our scale based on distance from the center of the chart.

Our mindset survey also asked demographic information and perspectives about information needs. Because we had that data as well, we were able to take these clusters and create demographic profiles for each group, based on factors such as their education, employment and family situation, and apply it to our segment analysis, too.

Although these might not fully represent our intended audience, profiling like this from sample responses provided direction for sizing opportunities and helped frame potential further research.

Here’s what that looked like in the results came of our segment analysis:

Progressive-Communal.

This segment exhibits the strongest sense of belonging and least loneliness across clusters. They report high satisfaction with educational information. They demonstrate high digital literacy, though a significant portion express skepticism about information reliability.

Despite their strong social networks, nearly half the respondents in this group struggle to identify appropriate information sources. Even more striking is that almost two-thirds reported difficulty understanding the information they find, despite their higher education levels. This is a disconnect that requires further investigation.

Their higher education appeared to correlate with a refusal to accept injustice and actively seeking to defend their rights when violated.

Conservative-Communal.

This segment features a high proportion of well-educated and female participants, with the highest representation in white-collar jobs. They generally maintain positive outlooks on life and report low levels of loneliness. They prefer stability, require longer adaptation periods to change, and value long-term planning.

Career advancement is a key concern, with over half citing lack of growth opportunities in their current positions. They struggle with work-life balance, being most likely to report insufficient time for children's education and seeking after-school care information.

They lead all clusters in technology usage for information gathering, though more than half of respondents maintain concerns about information reliability. This group reports the highest overall information satisfaction and shows no significant barriers to information access compared to other clusters.

Similar to the Progressive-Communal segment, their higher education appears to influence their response to rights violations, with many actively seeking remedies. However, they show mixed attitudes toward injustice, with less than one-third viewing it as normal and expressing no clear opinion.

Progressive-Neutral.

As the largest segment, this group shows high representation in manual labor. They demonstrate the strongest ability to adapt to new environments, maintain strong friendship bonds, and display moderate optimism about life. Their orientation toward adventure and long-term planning exceeds that of both Conservative groups.

Their primary concerns center on understanding fundamental rights and school enrollment policies. They demonstrate strong preference for technological channels to access information, while showing the lowest skepticism about information reliability. Among all clusters, they are least likely to rely on word-of-mouth information. This group reports the highest satisfaction across all information categories and demonstrates the greatest confidence in knowing where to seek information when needed.

This group shows a proactive approach to problem-solving, with most actively seeking information to resolve issues when faced with injustice. They generally reject the normalization of unfair treatment, with only a few accepting injustice as a regular part of life.

Conservative-Individualistic.

With moderate education levels (middle/high school) and high representation in manual labor, they strongly prefer stability and focus on immediate rather than future concerns. They report the highest levels of social isolation, with the majority feeling friendless and lonely, struggling to adapt, and expressing hopelessness about their future.

Despite being least likely to report growth limitations in their work, they show minimal interest in upskilling or entrepreneurship information.

They show the strongest preference for word-of-mouth information and lowest use of technological channels for information, while expressing the highest skepticism about information reliability across all groups. This group reports the lowest satisfaction with information across all categories, primarily due to trust issues and difficulty locating information sources.

Their pessimistic outlook appears to influence their response to unfair treatment, as they show the highest tendency to accept injustice as normal and the lowest likelihood to defend their rights when violated.

Step Four: Use patterns across segments reveal insights into information needs.

From these distinct mindset clusters that emerge in your analysis, you can create personas that help you better understand your intended audience, their attitudes toward information, and their information needs.

For example, in our project above, the Conservative-Individualistic persona is a rough-around-the-edges and pessimistic person who is distrustful, but she always finds a way to make it through difficulty on her own. She may not be receptive to information about seeking government support to get her through hard times, but she may want to learn of new job opportunities or available housing.

Referring back to these personas during news and information reporting and dissemination is  useful in meeting the information needs of your intended audience, because you understand their attitudes toward life, information, and society. Clustering and persona creation may lead you to notice certain trends or patterns that become the basis for strategic decisions in your reporting design and dissemination.

In one of our projects, we focused on the findings related to trust, information comprehension, and social isolation. Here are the findings that came from our cluster analysis and personas:

Defining divisions among audience segments: In our analysis, trust mechanisms marked differences between progressive and conservative groups. Progressive segments showed greater willingness to engage with digital platforms. In contrast, Conservative groups display a strong preference for information that flows through established personal networks, with trust built primarily through community validation. This division suggested to us that successful information distribution would require fundamentally different approaches for each segment.

Identifying opportunities for information:  Perhaps surprisingly, information comprehension barriers persisted across all education levels within our data. Even within the Progressive-Communal segment, where most have higher education, two-thirds reported difficulty understanding available information. This challenge appears even more pronounced in other segments, suggesting to us that education level alone does not guarantee someone could effectively consume information we provide. The persistence of these comprehension barriers across our segments pointed us to systemic issues in how information is presented and distributed.

Learning pronounced behavior: Social isolation emerged as a universal factor influencing information-seeking behaviors across all our groups, though its manifestations varied. In the Conservative-Individualistic segment, this isolation reinforces reliance on immediate social circles, with almost all reporting feelings of loneliness and limited social connections. Even among more progressive segments, social isolation shapes how information is sought and validated, with members of our target population often confined to specific information networks that may limit exposure to diverse perspectives and opportunities.

Mindset clustering is only one way to understand nuances within your intended audience

While psychological mindset clustering is a highly effective strategy to define and understand distinct groups within your intended audience, it’s not always possible to conduct this type of analysis for every audience research process.

If you do not have the capacity or resources to conduct a psychological mindset survey and cluster analyze your results, you can still learn from this process. We encourage you to try anyway, and we are here – hello@gazzetta.xyz — if you’d like to get in touch!

Here are our tips on how you can replicate small pieces of this process to better understand distinct audience segments:

1. Intentional interviewing: Instead of conducting a survey, you could rely on well-designed interviews instead. In your interviews, try to get at the diversity within your intended audience, rather than generalizations about the group as a whole. For example, instead of asking an interviewee what the most common need within a group is, ask if they can describe the range of views and competing priorities among people within that group.
2. Rooting surveying in prior research: If you do design a mindset survey, we’ll emphasize this point again: design it based on background research and a clear research goal! This will help ensure that your research can lead to distinct clusters in your intended audience, instead of broad results that don’t lend to insights.
3. Persona development: Whether you have conducted a survey or not, you can also brainstorm different personas on your own. Imagine an intended audience member who is young, one who is old, one with a low level of education, one with a higher level of education, one who is far away from their family members. Although they all belong to the same category defined by your intended audience, their life experiences and ways they came to belong to your defined group are different. Therefore, they likely feel differently about themselves and their place in the community. They have different levels of trust in the world around them. They have different information needs and different ways of encountering and interacting with the information they receive.
4. Daily life mapping: From there, you could map out the typical day of each of the different personas that you create. When do they have free time to receive information? What platforms do they use to receive that information? Who do they communicate with? When and how do they use technology?

Understanding in what ways members of your intended audience are similar and different, and exploring why, will assist you in serving their distinct information needs as you continue with news and information reporting in pursuit of service-oriented journalism.

Join us on our process in the audience research phase and beyond. If you haven’t already, sign up to our newsletter so you don’t miss out.

If you have feedback or questions, don’t hesitate to get in touch at hello@gazzetta.xyz.