Marketers have gotten scarily good at sorting people into segments and predictable behaviours. The problem is, the sharper that profile gets, the easier it is to forget it's still just a profile, not the real person. People don't hold still — their lives, their circumstances, and their reasons for doing things change — so, those neat little boxes they get categorised in go out of date quite quickly.
Melissa McNally sits right in the middle of this. As Head of Research and Analytics at Kagiso Connect's research media house, SoundInsights, she spends her days working on understanding audiences and translating that knowledge into business insights. In this Q&A she talks about the limits of audience segmentation, why human insight still matters even as AI gets better at targeting, and how marketers can keep up with consumers who never stop changing.
How should marketers think about consumers who do not neatly fit into the segments or personas created for them? Is there a risk that segmentation can make audiences easier to target but harder to understand?
Absolutely. Segmentation is a useful tool, but it becomes dangerous when we confuse the model with the human being.
There are many ways to segment an audience. Demographic segmentation groups people according to characteristics such as age, gender, income or life stage, while geographic segmentation focuses on where they live. Psychographic segmentation explores attitudes, identities, values and aspirations.
Behavioural segmentation groups people according to what they buy, watch, listen to or do. Needs-based and occasion-based segmentation considers the problem someone is trying to solve or the context in which a decision is made. Increasingly, predictive segmentation uses multiple data signals to estimate what someone is likely to do next.
Each approach offers value, but each is also a reduction. Demographics are often used as proxies for motivations they do not reliably explain. Behavioural data can tell us what someone did without revealing why they did it. Psychographic research can produce beautifully named personas that sound human but remain static constructions. Predictive models identify probabilities, but probability is not destiny.
Segmentation also frequently strips out culture and context. Two people may look identical in a dataset but attach completely different meanings to the same product, message, or behaviour. Language, community, family structures, social norms, history and place all influence how people interpret brands and make decisions.
Culture is not simply another variable to add to a profile; it is part of the system through which people understand the world.
Context matters just as much. The same person may make very different choices depending on where they are, who they are with, how much time or money they have available and what they need in that particular moment. When segmentation removes people from these circumstances, it can make behaviour appear more stable and predictable than it really is.
People do not experience themselves as demographic categories or fixed personas. The same person can be a parent, professional, caregiver, bargain hunter, aspirational consumer and loyal brand advocate — sometimes within the same day.
Someone classified as price-sensitive may pay a premium in a category that matters deeply to them. A supposedly loyal customer may leave after one poor experience. A digitally confident consumer may still want human reassurance when making a complex or high-risk decision.
This is where the market tests the model — and reality usually wins. People eventually expose the distance between a neat segmentation model and lived experience. If a brand assumes too much, over-personalises, stereotypes people or communicates as though one data point explains their entire identity, the result can feel irrelevant, patronising or intrusive. The targeting may be technically accurate while the human interpretation is completely wrong.
Segmentation can also become self-reinforcing. Once we decide that a group behaves in a particular way, we may show its members a narrower range of products, messages and opportunities. Their subsequent behaviour then appears to confirm the original model. What looks like insight may partly be the result of the choices the model allowed them to see.
Segments help organisations manage complexity, allocate resources and make practical decisions. The danger begins when efficiency replaces curiosity. We may become better at targeting a label while becoming less capable of understanding the person behind it.
The strongest approach is to layer different forms of intelligence. Demographic and geographic data provide context. Behavioural data shows what people do. Needs and occasions help explain what they are trying to achieve. Cultural and qualitative research gives us greater depth around the meanings, relationships and social realities surrounding their choices.
We should also pay attention to the people who do not fit the pattern. They are not necessarily errors in the data; they may be early evidence that needs, behaviours, culture or the market itself are changing.
Segments should therefore be treated as living hypotheses rather than permanent truths. A useful segment helps us ask better questions about people. It should never give us permission to stop asking questions altogether.
With AI and automation making audience targeting ever more sophisticated, what role do you think human insight should play in an industry becoming more data-driven?
As machines become better at identifying patterns, human insight becomes more important — not less.
AI can process information at extraordinary speed, identify correlations and help marketers personalise at scale. What it cannot automatically provide is a complete understanding of culture, meaning, lived experience or the emotional consequences of a decision. A pattern may tell us what is happening, but it does not necessarily explain why it is happening or whether acting on it is appropriate.
Human insight provides context and critical judgment. It helps us question whether the data is representative, whether an apparent pattern reinforces an existing bias and whether a technically efficient intervention will feel relevant, respectful or intrusive to the person receiving it.
There is also a risk that increasingly precise targeting creates the illusion of increasingly precise understanding. Knowing that someone is likely to click does not mean we understand what they need, value or fear.
AI should help us see more, test faster and identify questions we may not previously have considered. Human intelligence must still interpret what those patterns mean.
The future of insight is not human versus machine; it is using technology to extend our capabilities while retaining curiosity, empathy and accountability.
How should marketers approach consumers whose needs and motivations are changing faster than the researcher used to define them?
Marketers need to move from occasional audience definition to continuous audience learning.
A large annual study can provide a valuable strategic foundation, but it cannot be treated as a permanent description of the market. Economic pressure, technology, culture and personal circumstances can change behaviour quickly. Research must therefore become a living system that combines deeper strategic studies with regular pulse checks, behavioural data, cultural observation and direct audience or customer feedback.
It is equally important to track tensions, not only trends. What people say they value and what they are able to do may be very different. A consumer may want convenience but be constrained by cost, or value sustainability while prioritising affordability. These contradictions often contain the most useful insight.
Brands should build faster feedback loops and remain willing to revise what they believe. That means testing assumptions, listening to unexpected responses and paying particular attention to people whose behaviour does not fit the established narrative.
The researcher's role is not to prove the persona or story we have already created. It is to remain curious enough to recognise when people have changed — and humble enough to change our understanding with them.
Did you find this Q&A insightful? Let us know in the comments section below.
*Image courtesy of contributor and Canva