From the Lab to Shop Floor: Measuring What Consumers Really Feel
In a shelf study, shoppers took about 50% longer to choose under budget pressure, with little change in their emotional response. Marco Baldocchi explains what researchers gain by measuring attention, emotion and choice together.
The Insight250 spotlights and celebrates 250 of the world’s premier leaders and innovators in market research, consumer insights, and data-driven marketing. The awards have created renewed excitement across the industry whilst strengthening the connectivity of the market research community. Insight250 announced its 2026 winners in October. You can see the full list of Winners at Insight250.com.
With so many exceptional professionals named to the Insight250, we regularly tap into their expertise and unique perspectives across various topics. This regular series does just that: inquiring about the expert perspectives of many of these individuals in a series of short topical features.
As technology advances at an incredible pace and the value of insights grows, I sat down with Insight250 Winner Marco Baldocchi. Marco is a leading expert in consumer neuroscience. Aside from serving as CEO at emotivae, he is also an Adjunct Professor at Florida International University and Vice President & Head of the Research Department at ONCEMS.
Marco Baldocchi, CEO & Founder, emotivae
Crispin: When we last spoke in 2024, your work centred on lab-based tools such as EEG, eye tracking and galvanic skin response. What was the moment you decided to build emotivae, and what problem were you most determined to solve?
MB: I worked on emotivae for years, in complete silence, and only launched it commercially a year ago, once we had reached the results we were looking for. The idea first came to me about fifteen years ago, but the technology didn’t exist yet.
EEG, eye tracking and GSR gave me precise answers, but always inside a lab. People don’t live in labs. They decide in aisles, stations and waiting rooms, and those are exactly the places where a sensor on the head changes the behaviour you are trying to observe.
The turning point was realising that the most valuable signals were already visible. Faces and bodies leak involuntary micro-reactions all the time, and cameras are everywhere. What was missing was a way to read those signals continuously, at scale, without wearables and without asking anyone anything.
The problem I was most determined to solve was timing. Surveys tell you what happened after the fact. I wanted to see a response as it builds, early enough to act on it. In retail, that means understanding a decision while it is being made. In security, it means noticing that a situation is escalating before it becomes an incident. That is why emotivae was built from day one to predict situations, not to judge people.
Crispin: You describe emotivae as reading involuntary signals, the micro-reactions people cannot consciously control, from standard video. In plain terms, which signals do you capture, and how do they become a continuous measure of emotional arousal?
MB: In plain terms, emotivae watches for the small things the body does before the conscious mind has a chance to manage them. Fleeting facial micro- movements, changes in blinking, small shifts in head and posture, the way attention moves. None of these alone means much. Each one is noisy and easy to misread.
The value comes from reading them together and over time. Emotivae combines several of these channels into a single signal, frame by frame, so you don’t get a snapshot or a label but a curve: how intensely a person, or a crowd, is responding from one moment to the next. On top of that curve, emotivae recognises more than 100 distinct emotions, so you see not only how strongly people react, but how.
That curve is read against a baseline. What matters is not whether someone looks tense in absolute terms, but whether their response departs from what is normal for them and for that context. It runs on standard cameras, on-premises, so it can sit inside a store, a room where we are running a focus group, a station, or a clinic without anyone wearing anything.
Crispin: You have spoken before about facial coding being easier to deploy than EEG but less precise. How have you dealt with that trade-off, and how have you validated emotivae’s readings against established neurophysiological measures?
MB: I still believe what I said: video is easier to deploy and less precise than EEG. So we stopped asking video to do EEG's job. Instead of reading emotions from a single frame, which is where facial coding is weakest, emotivae reads them over time and against each person's baseline. Change over time is far more robust than a single label on a single frame.
This is what I was searching for for years: a use of a tool that could be as reliable as an EEG but less invasive. The second answer is fusion. One channel can be fooled or masked. Several involuntary channels moving together are much harder to fake and much less likely to be noise.
On validation, we test against the main benchmark datasets for facial coding and body movement, and against datasets we have built ourselves. We also run concurrent sessions where emotivae is recorded alongside neurophysiological measures, including EEG and fMRI, and check whether the data line up.
Crispin: Arousal tells us how intensely someone is responding, but not necessarily whether the response is positive or negative. How do you help clients interpret the data correctly and avoid drawing the wrong conclusions?
MB: This is the most important discipline we teach. A peak of arousal tells you something mattered. It does not tell you whether it was delight, anxiety or confusion. In psychology, this is close to what Paul Ekman called the Othello error: an innocent person under pressure can look exactly like a guilty one. So we never let the curve speak alone. We help clients read it in three ways.
First, against the moment: what was on screen, on the shelf or in the conversation when the peak happened. Second, against other evidence: behaviour, eye movements, movement speed, blink rate, pupil dilation, what people chose and what they said. Third, against the pattern: one peak in one person is an anecdote; the same peak across a sample is a finding.
We also remind clients that a flat curve is data too. In some categories, people simply don’t feel much when they choose, and knowing that changes how you should communicate with them.
Crispin: In our last interview, you said that asking a consumer what they would pay is like a waiter asking, “How was your dinner?” How does emotivae change the balance between what people say and what they feel, and where does declared data still have a role?
MB: The waiter still asks “How was your dinner?” and we still answer “Lovely,
thank you.” Emotivae lets you watch the dinner instead. You see the moment the main course arrives, the moment the bill lands, and you see it without asking anyone to translate a feeling into words.
But I don’t think declared data loses its role. It changes role. What people feel tells you what happened. What people say tells you how they explain it to themselves, which language they use and what they intend to do next. Brands need both.
The most interesting insight is often the gap itself. In a recent shelf study, shoppers told us they were confident in their brand and felt no strain choosing it. Yet when we asked them to choose again on a tight budget, two-thirds of them switched to the cheaper option. Neither their words nor their emotions warned us. Only putting the two side by side with behaviour did.
Crispin: You are using emotivae in neuromarketing studies of how consumers actually decide at the point of purchase. What have you learned so far that surprised you most?
MB: The biggest surprise came from a study we ran this year for a major household-goods manufacturer, with around 100 shoppers in front of a full-size shelf. We combined emotivae with eye tracking and implicit tests, and asked people to choose twice: once as they normally would, once imagining money was tight.
We expected pressure to make shoppers faster and more targeted. The opposite happened. Under budget pressure, they took about 50% longer, scanned more of the shelf, and moved their eyes more slowly. Attention to the market leader roughly halved, and attention to the private label more than doubled.
The surprise was what emotivae showed alongside that. Emotional response stayed low and almost identical in both conditions. The decision changed dramatically while the feeling barely moved. For a brand, that is sobering: your vulnerability can be invisible to the shopper’s own emotional experience. You only see it when you measure the feeling, the attention and the choice together.
Crispin: In retail, in-store cameras let you see how shoppers respond to products, shelves and service. Could you share an example, anonymised if necessary, where those findings changed what a retailer did?
MB: A chain of perfume stores, around twenty locations, wants to understand why so many customers walk in, try products and leave without buying. The first suspect is e-commerce, where prices are lower. With emotivae running on the cameras already in the store, and without identifying anyone, we read several layers together:
Attention: where heads and gazes turn, and how long people stop at a display or a screen.
Arousal: how intensely they respond, moment by moment, against their own baseline.
Valence and emotions: whether that response leans towards pleasure, curiosity or irritation (we read more than 100 distinct emotions).
Micro-expressions and blink rate: the signs of tension or cognitive load people don’t control.
Movement speed and posture: who browses calmly and who rushes, hesitates or stiffens.
Flows and dwell points: the path through the store and where waiting builds up.
Aggregated demographics: gender and age, always anonymous.
The picture is clear. At the entrance, arousal and valence are high: the window works. At the testers, attention is very close, but when customers wait for an advisor, arousal rises, valence drops, and movements become quick and restless. That is where people walk out. The digital screens at the back of the store, meanwhile, get almost no attention.
The recommendations are practical: move the most popular testers next to the advice stations, add a dedicated advisor in the time slots when tension builds, bring the screens to where people already stop, and adapt their content to the audience that dominates each time of day.
The outcome is a less frustrating experience: shorter waits at the most delicate point of the journey, more people moving from trying to buying, and a store that responds to what customers actually feel, not only to what they say in satisfaction surveys.
Crispin: Reading involuntary reactions through in-store cameras raises clear questions about consent, transparency and privacy. How do you approach this, and what does responsible use look like for researchers working within frameworks such as the ICC/ESOMAR Code and GDPR?
MB: This question should come before every deployment, not after. Our starting principle is simple: emotivae reads situations, not identities. In consumer settings, it does not identify anyone; outputs are pseudonymised and analysed in aggregate, and consumer data is architecturally segregated from security-related applications, ensuring a clear separation between the two industries in which we operate.
Everything runs on the client’s premises, so video doesn’t travel to a cloud. That is a privacy choice, not only a technical one.
For researchers, I see three practical rules:
a. Research participants give informed consent, in line with ICC/ESOMAR: they know what is measured, why, and how long it is kept.
b. In public spaces, transparency is not optional. Clear signage, a lawful basis under GDPR, a DPIA before going live and strict purpose limitation. Data collected to improve a shelf should never be reused to judge a person.
c. Data is anonymised and aggregated. In retail, the data emotivae collects is anonymised and aggregated, and records only gender and age. There is no way to trace it back to an individual. emotivae is also privacy- neutral: it doesn’t change our clients’ privacy rules in any way, and it doesn’t store frames except for security purposes, and even then within the agreed data-retention terms.
The EU AI Act is making these obligations explicit. Responsible teams should treat it as a floor, not a ceiling.
Crispin: Emotivae is also positioned for security and healthcare as well as consumer experience. What has working across such different settings taught you, and how do you keep the approach appropriate to each?
MB: The biology is the same everywhere. A rise in arousal in a supermarket aisle and in a metro station comes from the same nervous system. What changes completely is the cost of being wrong.
In consumer research, a misread peak costs you a weaker insight. In security, a false alarm can cost a person their dignity. In healthcare, a missed signal can cost a patient the right support. So the same signal gets different thresholds, different governance and different people reviewing it.
What this has taught me is that context is part of the measurement, not a layer on top of it. We keep the approach appropriate in three ways: data from different settings stays separate, every output is a prompt for a human rather than a verdict, and we refuse use cases where the technology would be asked to say more than the science allows.
Working in high-stakes environments has also made our consumer work more rigorous. Once you have designed for a security operator, you never present a curve to a marketing team without its caveats.
Crispin: You described the interplay between our rational and irrational systems as a “cognitive cruise.” Now that you can measure emotional response continuously rather than at single points, what does that journey look like in practice?
MB: When I first used the image, it was a metaphor: most of the time the automatic, emotional system is at the helm, and the rational system takes over only at certain moments. Now I can actually see the route. In practice, it looks less like a straight line and more like a sea chart. Long, calm stretches where people move on autopilot. Short bursts where something catches them, a price, a pack claim, a person in front of them.
And moments of hesitation where the rational captain comes on deck and slows the ship down. The shelf study I mentioned shows this clearly. Under budget pressure, the rational system took the wheel: more scanning, slower, longer deliberation.
The emotional system, meanwhile, stayed almost flat. The ship changed course without any storm. That is the real value of continuous measurement. You stop asking where the consumer ended up and start seeing where they changed direction, and why.
Crispin: For an insights team used to surveys, focus groups and CX metrics, how should emotivae sit alongside their existing methods, and what new skills will researchers need to use it well?
MB: Emotivae should not replace surveys, focus groups or CX metrics. It should sit underneath them, as a continuous layer that tells you when to ask and what to ask about. The best design is to measure the moment implicitly, then go back to people with a sharper question.
For example, in focus groups, both online and in person, emotivae is an incredible tool for spotting cognitive biases. In a focus group for an American food brand, some participants rated the food positively, but at the moment of tasting they were actually feeling disgust.
Why weren’t they telling the truth? Mostly the bandwagon effect: they didn’t want to be the odd voice out while everyone else said it was good.
Researchers will need a few new skills:
a. Study design around moments, not just around questionnaires: knowing exactly when the stimulus, the shelf or the interaction happens.
b. Reading time series: baselines, peaks, duration, and the difference between a signal and noise.
c. Triangulation: putting feelings, attention, choices and words on the same timeline.
d. Data ethics: consent, minimisation and knowing what not to claim.
Above all, they will need interpretive humility. The temptation with a new instrument is to over-read it. The researchers who do best are those who treat the curve as a question, not an answer.
Crispin: You have gone from digital marketing in Lucca to consumer neuroscience in Miami, and now to building a behavioural AI company.
Where do you see emotion measurement taking the insights industry over the next five years?
MB: Each step of my journey has been about getting closer to the real moment of decision. In Lucca, digital marketing showed me what people clicked. Consumer neuroscience in Miami showed me what their brains did in a lab.
With emotivae, I can see how people respond in the places where they actually live and choose.
Over the next five years, I expect emotion measurement to move from a special project to an always-on layer, much as web analytics did twenty years ago. Brands will see responses continuously, across stores, screens and service, rather than in one study a year.
I also expect the industry to split in two. On one side, providers that sell certainty: “we know what your customers feel.” On the other, those who measure carefully, explain limits and keep humans accountable. Regulation and public trust will reward the second group. The insights professionals who thrive will be the ones who can tell the difference. Picture a real case. In every mall, and in many shop windows, there are digital screens showing ads. Today, cameras measure how many people walk past, and their gender and age. With emotivae, while respecting people’s privacy, we analyse the level of attention and the emotions generated by age and gender. That makes it possible to target ads on digital screens the way we are used to doing on social media. Not tomorrow: already today.
Hot Topic: The License to Measure
The hot topic is legitimacy. As emotion measurement leaves the lab, the industry’s licence to operate depends on rules that were written for surveys and panels, where people knowingly sign up. In a store or a station, many of the people being measured never chose to take part.
If we get this wrong, we won’t lose a few studies. We will lose public trust in the whole category, and regulators will decide for us. The EU AI Act is already setting obligations for systems that infer emotional states, and other markets will follow.
It matters now because the technology has become cheap enough to deploy everywhere. The norms we set in the next two or three years, on consent, transparency and honest claims, will define whether this field is trusted or feared.
Understanding human beings is more and more our purpose. Doing it where they live and move every day is what makes the difference between something that might work and something that does.
Top Tip: Run it Where You Already Know the Answer
Curiosity is still the starting point, but my practical advice is this: your first emotion study should be one where you already know the answer.
Take a project you trust, a concept test, a shelf test, a CX journey, and add emotion measurement alongside your usual method. Don’t replace anything. Then look at where the two agree and, above all, where they disagree.
Those disagreements are where you will learn what the new measure adds, and also where it can mislead you.
Do it several times before you put emotion data in front of a client or a board. You will build intuition for what a peak means in your category, what a flat line means, and which questions the data simply cannot answer. Start with a question, not with a tool. The technology is the easy part. Knowing what to ask it is the real skill.
Crispin: Marco, thank you. The gap between what people say and what they actually do has been one of the insights industry’s oldest challenges. What stands out about emotivae is that it does not try to close that gap by discarding surveys or focus groups. Instead, it measures the moment itself and then sends us back to people with a sharper question. Your shelf study is a sobering reminder that a brand can lose ground without the shopper ever feeling it, and that only feeling, attention and choice read together reveal the full picture. The discipline you bring matters just as much: reading situations rather than identities, keeping video on the client’s premises, and treating the EU AI Act as a floor rather than a ceiling. As emotion measurement moves from the lab to the shop floor, the industry’s license to measure will depend on providers who explain the limits of their tools as clearly as their promise. Your advice to begin with a study where we already know the answer is something every insights leader should take to heart. Thank you for sharing your expertise, your candour and your vision for where our profession is heading.
Crispin Beale
Chairman at QuMind, CEO at Insight250, Senior Strategic Advisor at mTab, CEO at IDXCrispin Beale is a marketing, data and customer experience expert. Crispin spent over a decade on the Executive Management Board of Chime Communications as Group CEO of leading brands such as Opinion Leader, Brand Democracy, Facts International and Watermelon. Prior to this Crispin held senior marketing and insight roles at BT, Royal Mail Group and Dixons. Crispin originally qualified as a chartered accountant and moved into management consultancy with Coopers & Lybrand (PwC). Crispin has been a Board Director (and Chairman) of the MRS for nearly 20 years and UK ESOMAR Representative for c15 years. As well as being CEO of Insight250, Crispin is currently Worldwide CEO of Digital Communications Solution Agency, IDX. Crispin is also the Senior Strategic Advisor at mTab and the Chairman of QuMind and spent 4 years as Group President of Behaviorally where he was responsibile for the client & commercial teams globally. Crispin is a passionate advocate for blending human intelligence and technology to deliver innovation and leadership across organisations.


