There is a well-documented error in psychology called the Fundamental Attribution Error. It describes our instinct to explain people’s behaviour by pointing at who they are: their personality, their traits, their character. Decades of research have shown this instinct is wrong.
The evidence is extensive. Ross and Nisbett’s work on situational attribution demonstrated that observers consistently overestimate the role of personality and underestimate the role of context when explaining behaviour. Milgram’s obedience studies showed that ordinary people would act against their own values under situational pressure. Zimbardo’s Stanford Prison Experiment showed that assigned roles reshaped behaviour within hours. Mischel’s research on personality consistency found that individual traits predict only a small fraction of behavioural variance once situational factors are accounted for, a finding later reinforced by Funder and Ozer’s meta-analyses showing that situational strength routinely overpowers dispositional tendencies. More recently, Meyer et al.’s research on “strong situations” showed that when situational cues are clear (as they are in most professional and organisational settings), individual personality differences are suppressed and behaviour converges around contextual norms.
People’s behaviour is overwhelmingly shaped by their situation: the role they occupy, the company culture around them, the norms of their industry, the regulatory climate, the decision in front of them.
This matters for anyone trying to understand an audience, because almost every audience profiling tool on the market makes this exact mistake. They build a model of the person and assume it will predict behaviour. Psychology says that is the wrong lens.
Ada was built on the opposite principle. It profiles the situation first, then layers individual psychology on top.
The same person, different behaviour
Consider a CFO at a global investment bank. She operates in an environment that selects for risk awareness, rewards caution and runs on institutional consensus. Now imagine the same person moves to a founder-led mid-market firm. Her personality has not changed. Her behaviour will.
The norms, incentive structures and cultural expectations around her are different. Those forces do more to shape how she will respond to your message than any personality trait you could identify.
Or consider an analyst evaluating a deal from Frankfurt versus Abu Dhabi. Same analyst, same deal. Different regulatory environment, different cultural relationship to risk, different institutional priorities. Different behaviour.
This is not a philosophical point. It produces measurably different outputs in the work Ada does for clients every day.
Situation first, personality second
Ada constructs audience profiles across seven layers. The order matters.
Industry context comes first: the structural pressures, regulatory constraints, competitive dynamics and decision-making norms of the sector. Then role and job level, because research shows measurable psychological patterns associated with specific roles (a VP of Engineering processes information differently from a VP of Marketing, not because of who they are personally, but because the role selects for and reinforces certain patterns). Then company personality and culture, because every organisation shapes how its people think and decide. Then geography and culture, drawing on Hofstede’s cultural dimensions and related frameworks. Then buying committee dynamics, because what someone says in front of colleagues is different from what they would say privately. Then situational and temporal context: earnings, leadership changes, regulatory announcements, competitive threats.
Only after all of that does Ada overlay individual psychology: Big Five personality traits, regulatory focus (promotion versus prevention orientation), decision-making styles.
The individual layer modifies the prediction. It does not drive it.
300+ behavioural science models, selected per brief
Ada draws from a curated library of over 300 peer-reviewed behavioural science models, measurement scales and psychological frameworks. They fall into three categories.
- Diagnostic models help Ada understand the psychological machinery behind an audience’s perspective. Dual Process Theory (Kahneman) determines whether an audience is likely processing via fast, intuitive reasoning or slow, deliberate analysis. Regulatory Focus Theory (Higgins) identifies whether someone is motivated by pursuing gains or avoiding losses. Prospect Theory explains why people feel losses roughly twice as powerfully as equivalent gains. Self-Determination Theory identifies which motivational needs (autonomy, competence, relatedness) dominate in a given context.
- Behaviour change models predict what will actually shift decisions. The COM-B Model diagnoses whether a lack of action is a motivation problem, a capability problem or an opportunity problem. The Elaboration Likelihood Model determines whether argument quality or surface cues will carry more weight. The Theory of Planned Behaviour identifies the specific levers (attitude, peer norms, perceived control) that drive intention to act.
- Rhetoric models translate those insights into effective messaging. This includes classical rhetoric foundations, Cialdini’s principles of influence, narrative transportation theory, inoculation theory (building resistance to competitor counter-arguments) and processing fluency (why clear, well-structured writing is literally perceived as more true), alongside 90+ specific rhetorical devices mapped to the psychological mechanisms they activate.
- The set of models Ada uses is not fixed. The behavioural models relevant to a Saudi sovereign wealth fund analyst differ from those relevant to a Silicon Valley CTO. The models for a prevention-focused CFO evaluating risk differ from those for a promotion-focused CMO evaluating opportunity. Ada selects the right combination for each brief.
Why this is different from digital twins
Digital twin platforms (Evidenza is the most prominent) create synthetic representations of people or audience segments, then “ask” those representations how they would respond to messaging.
This approach has four structural problems.
- People change; situations do not. A digital twin is a snapshot. But individual responses are variable: they depend on mood, context, what someone just read, who else is in the room. Situational determinants, role norms and cultural norms are stable. They change slowly and predictably. Building predictions on norms produces more reliable results than building on individual snapshots.
- Pre-built pools cannot capture context. A digital twin pool can tell you how “someone like” your target might respond. It cannot tell you how a specific individual at a specific company, facing specific pressures in a specific cultural context, will respond, because those contextual factors are not part of the twin’s construction. Ada builds every profile from scratch, case by case, incorporating the full situational context. One client went from producing 10 individually versioned pieces of content per quarter to producing 7,200, each calibrated to a specific individual in their specific context. That level of specificity is architecturally impossible with a pre-built persona pool.
- The personality-first trap. Digital twins weight personality and demographic traits as the primary predictors of behaviour. Ada uses personality profiling too, but only after the situational, role-based, cultural and company-level analysis is in place.
- Aggregation masks the insight. When digital twin platforms test messaging against a pool of personas, they aggregate responses. But the value of audience profiling is in the differences: understanding that two organisations in the same industry may have fundamentally different investment philosophies, decision-making structures and cultural orientations. Aggregating across a pool smooths out exactly the distinctions that matter.
Personalise without being personal
Ada follows a governing principle: the recipient should feel like they are discovering something, not being targeted.
This is grounded in the same behavioural science that powers the profiling. Reactance Theory shows that when people feel manipulated, they resist. The Elaboration Likelihood Model shows that sophisticated audiences are alert to persuasion techniques, and visible targeting triggers scepticism. Self-Perception Theory shows that if someone feels they chose to engage with an insight (rather than being targeted), they attribute their engagement to genuine interest, making them more receptive to what follows.
In practice, this means the research stays backstage. The personalisation lives in the relevance of the angle, not in demonstrating knowledge about the person. Content reads as “here is something worth knowing about your world”, never as “we have a dossier on you”.
The depth of personalisation is invisible to the recipient but visible to the seller. That is the point.
Why it works at scale
Ada’s approach has been validated through real-world deployment with clients including one of the largest technology companies in the world. The question prospects ask most often is: why would that company use this when they have their own AI, billions in investment and more data than anyone?
The answer is straightforward. General-purpose AI is designed to produce the most statistically likely output. It reverts to the mean. That is exactly wrong for audience profiling, where the value is in understanding how a specific audience departs from the average.
Ada operates in a specialised niche where the goal is the opposite of reversion to the mean. It uses validated psychological models to reconstruct the specific perspective of a specific audience: the ways they differ from the default, the biases that make them predictably irrational, the cultural and situational norms that constrain their thinking.
A company like that can build a general-purpose language model. What they cannot do, because it requires a different architecture entirely, is the specialised behavioural reconstruction that Ada provides.
There is also a practical legal benefit. Because Ada profiles roles and situations rather than individuals, it sidesteps many of the data protection concerns that affect individual-level profiling. All individual-level inputs are drawn exclusively from publicly available information. The behavioural predictions come from validated psychological models applied to publicly observable contextual factors, not from private data.
Ada works because it is built on what psychology actually tells us about human behaviour, not on the assumption that personality predicts everything. People are contextual chameleons. Their behaviour is shaped more by their situation, their role, their culture and their company than by their individual traits. Profiling the situation first produces predictions that are more accurate, more stable and more actionable than any personality-first approach.
