The Personalisation Paradox: Why the More You Target, the Less You Understand the Human You're Targeting

The $50 billion marketing technology industry promised brands they would finally understand their consumers. It delivered something far more dangerous: the illusion that they already do.

The Room Where It Happens

Picture a scene that plays out in marketing departments across the world, every quarter, with remarkable consistency.

A brand's analytics team presents its latest audience segmentation work. Hundreds of signals. Thousands of data points per individual. Propensity scores. Lookalike models. Predictive intent layers. The slides are immaculate. The confidence intervals are tight. The audience segments have names - "Ambitious Achievers," "Mindful Millennials," "Value-Conscious Families" - that make real human beings sound like characters in a brand's own internal mythology.

The CMO nods. The agency nods. The media team begins building activation strategies against each segment. The personalisation engine is briefed. The campaign launches.

And somewhere, entirely outside the model, a forty-three-year-old woman who was categorised as a "Mindful Millennial Value Shopper" makes a completely unpredictable purchase decision - driven by a memory, a mood, a conversation with a friend, a moment of aspiration she couldn't have explained to herself, let alone to a survey - that no propensity score saw coming.

She is not an anomaly.

She is the rule.

And the industry has built a $50 billion technology infrastructure on the premise that she is the exception.

The Seductive Promise

To understand how we arrived here, it is necessary to revisit the promise that drove the personalisation revolution in the first place.

The pitch was elegant and, on its surface, entirely reasonable. Mass advertising was wasteful. Showing the same message to everyone was inefficient. Brands were spending enormous resources reaching people who would never buy their products, with messages that didn't resonate, in contexts that were irrelevant. Technology could fix this. Data could fix this. By understanding each consumer as an individual - their behaviour, their preferences, their context, their intent - brands could eliminate waste, improve relevance, and build deeper, more valuable relationships with the people most likely to buy.

It was the right diagnosis. Undifferentiated mass communication does have real limitations. Relevance does matter. Context does influence reception.

But the prescription - increasingly granular data collection, increasingly precise algorithmic targeting, increasingly individualised message delivery - carried within it a philosophical assumption that nobody examined carefully enough at the time, and that the industry is only now beginning to confront:

The assumption that human beings can be sufficiently understood through their behavioural data.

That assumption is wrong. Not partially wrong. Fundamentally wrong. And the consequences of building a global marketing infrastructure on a false premise are now working their way through brand equity metrics, customer lifetime value analyses, and boardroom conversations that nobody quite knows how to frame.

What Data Actually Captures

Let us be precise about what behavioural data captures, because precision matters here.

Behavioural data captures what a person has done, in specific digital contexts, within the tracking parameters of the platforms recording those actions, filtered through the particular moments in time when those actions occurred.

It captures the click. It does not capture why. It captures the purchase. It does not capture the meaning the purchase held. It captures the browse. It does not capture the aspiration, the anxiety, or the memory that initiated it. It captures the pattern. It does not capture the contradiction.

The same person buys premium extra-virgin olive oil and store-brand cereal in the same shopping basket. Researches luxury cars at midnight and takes public transport to work. Is fiercely brand-loyal to a coffee they associate with a late parent and entirely promiscuous in their choices of everything else. Makes seventy percent of purchase decisions based on something in the moment - a conversation, a mood, an aesthetic impulse - and then constructs rational narratives about value and quality afterwards.

Human decision-making, as two generations of behavioural economics research has established beyond serious dispute, is non-linear, context-dependent, emotionally driven, and routinely post-rationalised. Daniel Kahneman's dual-process theory, Rory Sutherland's work on the logic of irrationality, and Dr. Robert Cialdini 's research on influence all converge on the same uncomfortable truth for marketing technologists:

People do not behave like their data predicts they will behave, with a regularity that should fundamentally disturb anyone whose business model depends on predicting behaviour.

What personalisation engines actually model is not the person. It is the person's recent behavioural residue in trackable digital environments. This is not a small distinction. It is the difference between knowing someone and knowing their browsing history - and the industry has been consistently, commercially motivated to treat these as equivalent.

The Aperture Problem

There is a phenomenon in visual perception called the aperture problem. When you observe a moving object through a small opening - an aperture - you can perceive its movement within that aperture clearly, but you systematically misperceive its actual direction of travel, because you are seeing only a portion of the full motion.

Personalisation at scale creates an aperture problem of extraordinary commercial consequence.

The more data points you accumulate on an individual, the more precisely you can see their behaviour within the aperture of your data collection environment. The CTR model improves. The propensity score tightens. The predicted next purchase becomes more accurately predictable.

But the aperture itself - the slice of a person's experience that is captured in trackable digital behaviour - remains a fraction of who they are and what drives them.

And here is the structural danger: as the precision within the aperture improves, confidence in the model grows. The analytics team becomes more certain. The segments become more granular. The personalisation becomes more specific. The entire organisation develops an increasing conviction that it understands its consumer.

While the consumer continues to make decisions for reasons that live entirely outside the model.

The greater the analytical confidence, the smaller the intellectual humility. And in marketing, the erosion of intellectual humility about consumer complexity is typically followed, with a delay of one to three years, by a brand equity problem that nobody saw coming.

The Brand Dissolution Problem

There is a second dimension to the personalisation paradox that receives far less attention in industry discourse, and that carries arguably more serious long-term consequences for brand value.

Brands are, at their functional and economic core, shared meanings.

A brand works - creates pricing power, generates preference in moments of parity, earns the loyalty that smooths the commercial cycle - because a large number of people hold a consistent set of associations, emotions, and meanings about it in common. The brand is a cultural object. It operates in the shared space between people, not just in the individual minds of segmented audiences.

Nike's "Just Do It" did not build one of the most valuable brand identities in commercial history by telling different things to different people. It told one story - about human potential, about the courage to begin, about athletic aspiration - to everyone. And because it told that story consistently, at scale, across contexts, it became a piece of shared cultural meaning that transcended demographics, segments, and behavioural cohorts.

When a brand's message is algorithmically optimised for each individual - when the headline, the image, the offer, the emotional register, and the narrative are all dynamically assembled based on each person's predicted preference state - the brand is no longer telling one story. It is telling millions of micro-stories that share a logo and a colour palette.

And logos and colour palettes, on their own, do not constitute brand meaning.

The consequence, already visible in brand tracking data for the most aggressively personalised digital advertisers, is a gradual hollowing of brand salience and distinctiveness. Performance metrics look strong. CTRs are healthy. ROAS models look efficient. Meanwhile, the shared cultural meaning that made the brand worth advertising in the first place is slowly being liquidated.

Professor Byron Sharp 's empirical work at the Ehrenberg-Bass Institute established with considerable rigour that brand growth is driven primarily by mental availability - the breadth and ease with which a brand comes to mind in purchase contexts - and that this mental availability is built through consistent, broad-reach communications that create and refresh memory structures across the full buying population, not just the predicted high-propensity segment.

And when the harvest runs low - when the high-propensity segments are exhausted, when performance starts declining, when the brand equity that was quietly not being invested in begins to show in the numbers - the organisation discovers, too late, that efficiency ate the seed corn.

The AI Escalation

Everything described thus far operates at the level of current marketing practice. What is coming next makes the stakes significantly higher.

Generative AI is entering the creative production layer of personalisation. It is no longer simply that targeting is algorithmically determined. The creative itself - the message, the visual language, the tone, the narrative architecture - is increasingly being dynamically assembled by AI systems, individualised at a granularity that human creative teams could not achieve at any scale.

This development has been received in marketing circles with considerable enthusiasm, framed predominantly as an efficiency and relevance story: more messages, more individually tailored, at lower cost. Unilever, Coca-Cola, and dozens of major advertisers have publicly committed to AI-driven creative personalisation as a core capability investment.

The question the industry is not asking with sufficient seriousness is this: when AI generates individualised creative at scale, who is responsible for what the brand means?

Creative development, historically, has been the primary mechanism by which brands exercise deliberate judgment about meaning. The brief, the creative territory, the tonal guidelines, the decision about what to say and how to say it - these are acts of intentional meaning-making. They are how brands exercise agency over their own cultural presence.

When a creative is dynamically assembled by an AI system optimising against short-term response signals, the brand's meaning becomes, in a real sense, an emergent property of the optimisation algorithm - not a deliberate strategic choice.

That is a profound transfer of brand sovereignty. And it is happening largely below the level of board-level scrutiny, embedded in vendor contracts and technology platform capabilities, described in language of efficiency and personalisation rather than the language of brand governance it actually warrants.

The Measurement Trap That Locks the System In Place

Why, given the growing evidence of these dynamics, does the industry not course-correct?

The answer lies in the measurement architecture that has become the dominant framework for marketing accountability - and that is itself a product of the same digital infrastructure that created the personalisation machine.

Last-click attribution and its various multi-touch descendants are, structurally, optimised to reward the final moments of the purchase journey - the precisely targeted, high-intent, lower-funnel interactions that personalisation engines are built to deliver. They are, equally structurally, incapable of measuring the slow-building, broad-reach, brand-meaning-creating work that happens in the upper funnel and that drives the consumer to the lower funnel in the first place.

CFOs see the ROAS numbers. The brand investment looks unmeasured and therefore unjustifiable. The budget migrates toward performance. The personalisation machine grows. The brand equity softens. And three years later, the organisation discovers that it has very efficiently harvested a brand it forgot to grow.

Marketing Mix Modelling and incrementality testing, properly constructed, can partially correct for this. But they require investment, patience, and a willingness to surface findings that challenge the short-term performance narrative - qualities that are unevenly distributed across marketing organisations facing quarterly pressure.

A Different Philosophy of Consumer Understanding

This article has argued, at some length, the case against the dominant paradigm. It is only fair to articulate what a more philosophically sound approach to consumer understanding might look like.

The most effective consumer understanding is not the most granular. It is the most generative - the understanding that opens up creative and strategic possibilities rather than narrowing them to the already-predicted.

This means investing in research methodologies that give consumers space to be contradictory, irrational, and surprising - ethnographic approaches, long-form qualitative work, cultural analysis, semiotics - alongside the quantitative behavioural signals. It means treating behavioural data as one input among many, rather than the primary truth layer.

It means protecting brand consistency as a strategic asset in an era that treats it as an inefficiency. Understanding that the consistency of a brand's communication - its distinctiveness maintained over time - is not a failure to personalise. It is the mechanism by which shared meaning is built and maintained.

It means developing measurement frameworks that can see what performance dashboards cannot: the slow accumulation of mental availability across the broad buying population, the brand equity that makes performance activity work better than it would otherwise, the cultural presence that takes years to build and can be quietly eroded in a single cycle of over-indexed performance investment.

And it means - perhaps most importantly - recovering intellectual humility about the complexity of human motivation. Recognising that the most commercially valuable insight about a consumer is frequently the thing that doesn't appear in their data, because it hasn't happened yet: the aspiration they're beginning to form, the identity shift they're navigating, the life stage transition that is about to make them a completely different buyer.

The consumer you understand through their data is the consumer they were.

The consumer you need to reach is the consumer they are becoming.

The Wisdom That Technology Cannot Provide

There is nothing inherently wrong with personalisation. Context-sensitive communication is, at a basic level, simply good communication. Relevance matters. Timing matters. Recognising that different people have different needs is not a philosophical error - it is a prerequisite for effective marketing.

The error is in the scaling: in taking a sensible principle and industrialising it to the point where it inverts its own logic. In allowing the pursuit of individual relevance to erode the shared meaning that makes brands commercially valuable. In mistaking behavioural precision for human understanding.

Between the data exhaust and the fire.

Between the signal cluster and the human being.

The brands that will define the next decade of marketing are not the ones that personalise most aggressively. They are the ones that are wise enough to know what personalisation cannot see - and brave enough to invest in it anyway.

That is not a technology problem.

It is a leadership one.

Authored By

Anil Pandit

*Disclaimer: This post is for informational purposes only and does not endorse or disapprove of any specific tools, platforms, or technologies. The views and opinions expressed in this article are those of the author and do not reflect the official policy or position of the company where he is employed.