Why Knowing What Your Customer Is Thinking Beats Knowing Who They Are
Most digital advertising still asks the same question it asked twenty years ago: Who is this person? Tyler Boyd, CEO of Youtility, thinks that is the wrong question. His company asks a different one:
The industry still runs on an assumption that has gone largely unquestioned for two decades: that if you know enough about who someone is (their age, income bracket, browsing history, purchase patterns), you can predict what they want and when they want it. Demographic segmentation, behavioural lookalikes, and predictive models built on historical data: all are variations on the same premise.
Tyler Boyd thinks that premise has a fundamental flaw. “Demographic foundation is additive,“ he says. “It may be interesting, but it’s not defining.“
Boyd is CEO of Youtility, a US and UK based company that has been building what it calls a decision science platform, an AI driven system designed to identify not just who a customer is, but also what emotional state they are in now, in terms of engagement. The distinction sounds subtle, but the commercial implications are not.
The problem with past Behaviour
The standard critique of performance marketing is that it optimises against history. Boyd’s version goes further. Past-behaviour models, he contends, are working from data that may be years out of date — profiles built on PII and credit scores that haven’t kept pace with how someone’s life has actually changed. You got married, had a child, moved house, and changed jobs. Your marketing profile still says single, urban, early career. The segment hasn’t moved; the person has.
“If you’re a hungry guy and you want the two-for-one special, you want a deal seeker. Well, if you just had a steak, you’re no longer a deal seeker. That’s where your mental state is.“
Tyler Boyd, CEO, Youtility
Mental state is dynamic. Someone is not permanently a deal seeker, permanently risk averse, or permanently in the market for a mortgage. They move in and out of those states depending on circumstance, context and recent experience. Advertising that treats people as fixed types will always be misfiring at the margins, and at scale, those margins represent significant wasted spend.
Youtility’s platform attempts to model those shifting states in real time. The system, in which Navier is the model and Stokes is the interface, named after the fluid dynamics equations that describe changing flow, has been built on over 230 million customer profiles. It scores individuals across 24 core emotional dimensions, grouped into eight defined categories, producing what Boyd describes as an individual level emotional profile rather than a segment level approximation. “We do not know anyone else,“ he says, “who can go to the individual level and tell you about the individual emotional drivers for a person.“
Think-alike versus look-alike
The practical output of this is what Youtility calls think-alike audiences: a deliberate contrast with the look-alike models that dominate programmatic advertising. Look-alike modelling finds users who resemble your existing customers in observable demographic and behavioural terms. Think-alike modelling finds users who share the same underlying mind state: the same emotional drivers and motivational profile.
Boyd is emphatic about why these matter in practice. Many organisations, he notes, are trapped by pre-defined demographic segments (“Rising metropolises”, “up and comers”, “young professionals”) that may describe their existing customer base accurately but constrain their view of who their next customer could be. “The 80-year-old grandmothers: a few of them are deal seekers. You haven’t talked to them about a new product because you didn’t think they’d ever want it. Well, actually, maybe they do.“
The commercial implication is that think-alike targeting does not just find more of who you already have; it finds people who think the way your best customers think, regardless of whether they fit the demographic profile you would have specified.
What the data says
The case for targeting emotional states is not purely theoretical. Youtility has campaign evidence across energy, banking and consumer goods, and the performance differential between a first campaign pass and a refined second pass is striking.
Working with a leading UK bank, the company produced a 74% increase in click-through rates and a 17% increase in sales on an energy switching campaign. A more recent campaign for Monese, a European fintech with Barclays and PayPal among its investors, targeting existing customers for micro-loans of between £50 and £200, produced a 91% increase in click-through rates and a 24% increase in sales on the first iteration. When Youtility refined its emotional targeting based on how those customers had behaved (who clicked through but stalled at the FAQ page, who dropped off, who came back), the second round produced a 154% increase in click-through rates and a 285% increase in sales.
A standard A/B test would produce variants of the same message; Youtility’s system identifies that a customer who hesitated has moved from deal seeker to cautious and adjusts the message accordingly. Boyd’s description of what a conventional LLM-based approach misses is pointed: “All English language based models give you a derivation of a different way to say the same thing. We give you: Is this person a deal seeker? Is this person cautious? Has this person changed from a deal seeker to cautious based on the actions they took in response to the first email?“
The activation layer
Emotional profiling only generates commercial value when it connects to campaign execution. That is where Chicago based Fyllo comes in.
James Ramelli at Fyllo describes the partnership in straightforward terms. Youtility defines the mindstate cohort. Fyllo works out how to find those people in the wild and get a message in front of them. “We’re taking that definition of mindset and actually finding the people through their addressable messages,“ he says. In practice, that means translating a cohort of emotionally profiled users into a list of devices (laptops, phones, connected TVs) and identifying which channels and formats will reach them most effectively.
The channel strategy component is not trivial. Ramelli’s point is that knowing a user’s emotional state should inform not just the targeting decision but the medium and timing of delivery. A financially anxious customer might be receptive to a savings product via a mobile push notification at a specific time of day, and less receptive to the same message in a thirty-second CTV slot. Youtility’s model can specify the recommended channel alongside the emotional profile; Fyllo executes across it.
What Boyd is describing is a closed loop: response data from each campaign feeds back into Youtility’s model, sharpening the emotional profile for the next round. “Customers who are more inclined to convert at a higher rate come through,“ he says. “We’ve been talking to them in that tone and tenor since they saw an ad for the first time.“
The publisher case
The use case that should interest publishers most is also the one Boyd is loudest about. His pitch to the industry is direct. “Publishers have been sitting with an unused data source on their balance sheet since day one,“ he says. “It’s been growing, and it’s been considered goodwill, not a revenue generator.“
The argument is that a publisher’s first party engagement data, covering what subscribers read, when, for how long, where they drop off, what they come back to, what they never click, is precisely the kind of behavioural signal that Youtility’s model can interpret as emotional state. That interpretation can then drive more targeted advertising inventory, more effective subscription conversion messaging, and more precise retention campaigns. Youtility claims it needs only a handful of data points to produce an initial customer profile, with accuracy improving as the feedback loop accumulates.
For publishers facing the dual pressure of declining third party cookie signals and eroding referral traffic, the proposition is worth taking seriously. First party data has been the industry’s answer to both problems for several years; Youtility’s case is that, used to model emotional states rather than just behavioural patterns, it is meaningfully more valuable than most publishers currently extract from it.
Where this sits in the market
Youtility is a medium sized company (it has raised $4.2 million to date and is targeting a seed plus round), making large claims in a market full of them. The AI targeting space is noisy, and executives who have been through multiple cycles of targeting innovation are right to be sceptical of anything that sounds like a repackaging of what already exists.
What distinguishes Youtility’s argument is the specificity of the methodology and the granularity of the performance evidence. Think-alike audiences built on emotional profiling, scored at the individual rather than segment level, with a feedback architecture that refines targeting in real time: that is a meaningfully different proposition from Look-alike modelling with an AI wrapper. Whether it scales as claimed across a wider range of verticals and campaign types is still to be tested. But the early numbers are striking, and Youtility is, on this evidence, well ahead of the field it is competing in.










Super interesting and would love to trial a use case here. Certainly in corp and investor communications, the aspect of mental state hasn’t been explored to my knowledge.