NeuroMethod 2: Which brain measure is best?
Neuromarketing is currently disrupting the insights industry. By providing both more granular data, novel metrics, and better predictions, neuro is all the rage. But what we also often see, is a complete confusion of what is actually labeled as “neuromarketing methods,” and the industry has traditionally not worked in harmony on aligning on methods or metrics. This section is dedicated to giving a short but valuable overview of the different core methods of neuromarketing, and other methods closely related to it.
What is a neurometric?
This might seem simple to answer, but the truth is a bit more complex. As we look at different pages listing “neuromarketing methods” one could easily get the impression that a lot of methods are used. So let’s start by distinguishing between the terms “neurometric” and “neuromarketing”.
“Neurometric” is a term that should preferably be used for a metric based on brain data. After all, the very name of the metric suggests that it’s based on brain responses. Sure, pupil dilation, heart rate, facial expressions, and other metrics are ultimately based on brain responses. But so are survey responses, choices, and so on — after all, it’s still the brain doing this, but it just makes little sense to call everything a neurometric. So we should draw the line between what is to be considered a “neuro measure” and what is not. So a measure and metric based on measured brain responses should be the minimum criteria for being labeled as a neurometric.
Eye-tracking is not a neurometric
This also means that eye-tracking is not a neurometric. The same goes for galvanic skin response, heart rate, facial coding, or implicit responses. These should be classified as other metrics such as (GSR, and heart rate) and behavioral responses (eye-tracking, facial coding, and implicit responses). One can also argue that eye-tracking can also be seen as a physiology measure since pupil dilation can be used as a measure of physiological responses such as emotional arousal (if you correct for other things that affects pupil dilation, such as brightness and task demands).
What is a good neuromarketing metric?
Before launching ourselves into the discussion about what the best tool is, we should distinguish between two types of research. In my neuromarketing textbook (2nd edition on the way) and in my free Coursera course, I have been stressing that we should distinguish between “neuromarketing” and “consumer neuroscience” — it’s not a semantic exercise, but something we can use to better relate to what we are talking about. What do I mean by the distinction?
- Neuromarketing is the commercial application of neuroscience tools, methods, metrics, and understanding to better understand and affect customer behaviors. Here, neuroscience is the tool and the aim is to better understand customer responses. This type of work is primarily run by commercial companies with a for-profit aim.
- Consumer neuroscience is the academic and basic science study of how consumption behaviors are manifested in the brain. That is, here we are using customer behaviors as a model to understand the brain. This type of work is primarily run by academic institutions with a non-profit aim.
As a consequence of this, we can better determine what is a “good neuromarketing metric.” Since we have already said that neuromarketing is trying to understand consumer responses then only measures that help us reach this goal are good. To put it bluntly, this is how we can define a good neuromarketing metric:
A good neuromarketing metric is one that can produce valid, reliable, and actionable insights into consumer responses that would not otherwise be possible to measure with any other measure, and which provides a better prediction of consumer/market responses.
Here, I will provide my own take at different methods, to better show what I consider to be good neuromarketing methods, which should be better reserved for consumer neuroscience research, and what still needs validation. Let’s take a few examples.
Good measures for the neuromarketing toolbox:
- Eye-tracking, the tracking of eye movements, is probably the most often used method, allowing both stationary and mobile tracking, the latter being that people are allowed to move freely around in a more natural environment, which also is a bit more labor intense for marking the data correctly. Eye-tracking based metrics range from a basic eye-ball count (%seen) to how long time is spent looking at an area (Total Fixation Duration) with a few tens of milliseconds of accuracy. This is a great quantifier of whether something grabs enough attention to produce a response and subsequent memory.
- EEG, or electroencephalography, is the use of electrodes on the surface of the head to measure electrical activity from the brain. In itself, the raw EEG signal needs many layers of processing, including denoising, Fourier transform, and normalization before it can be used as metrics. While many companies are using EEG, the methods and metrics they are calculating are very often treated with a lot of secrecy and with no external validation — which in my opinion is either done wrongly or being outright disingenuous. EEG based metrics should always be using published scientific metrics, and the industry should demonstrate steadfastness on upholding this standard. Some measures, such as valence/motivation and cognitive load are pretty well established in the literature, while other measures such as “frustration” and “mirror neurons” (suggested to be related to mu suppression) require more validation before they can be used in the field.
- Heart rate (HR), which can both be used to measure the rate or the variability in HR, can be a measure of arousal — often, it can be used to support other measures such as EEG. The measure itself can be of interest, but not as a highly reliable measure of second-by-second responses, as HR changes are more subtle, sluggish and slower to detect.
- Implicit association tests are well-established measures that can tap into customers’ gut responses and implicit associations to ads, brands, products, and more. Interestingly, implicit measures are among the few methods that can be run automatically and online. That said, recent studies suggest that there are still outstanding questions, and there is still a need for proper validation of this method.
- Automatic image analysis is a yet underutilized approach in neuromarketing. In academic terms this is called “computational neuroscience” and it typically consists of a method that relies on our knowledge about how the brain works on certain aspects, and then uses this knowledge to predict consumer behaviors. For example, NeuroVision is a tool where you can upload your image /video and have it analyzed on the fly. The algorithm then uses what we know from the visual parts of the brain to produce a heat map of the features in the image/video that are most likely to grab attention. While still in beta, you can set up an account and try out NeuroVision yourself now, for the fraction of the costs of running an eye-tracking study. NeuroVision will soon be launched formally (stay tuned!), and it will also include an AI-based algorithm that has even higher accuracy on predicting consumer attention.
Measures mainly for consumer neuroscience research:
- EEG should be added here also since it works both as a good commercial device as well as a great academic tool to understand how brain activity unfolds over time. Typically, one can distinguish between commercial and academic research in that the latter allows the use of many more electrodes to increase spatial resolution. With more recent advances in signal detection and source reconstruction, high-resolution EEG can be used to detect which areas of the brain that are engaged, and trace activity as it spreads with millisecond accuracy. This contributes to the surge in research on brain dynamics and connectivity.
- fMRI, or functional Magnetic Resonance Imaging, is a measure of the relative oxygenation levels in specific regions of the brain, which is an index of brain activity. This method is so-called BOLD fMRI, and other fMRI methods exist, such as perfusion-based fMRI, but most of the time we talk about BOLD fMRI. This method is extremely good for understanding the parts and intensity of brain region activity related to a task. More recent analyses also focus on the network of activity in the brain. However, in fMRI itself has many issues that make it less suitable for commercial studies, such as a high cost per participant, testing in an artificial environment (loud noise, lying down, repetitive tasks), as well as aspects that make data normalization across studies a challenge (e.g., scanner drift).
- Magnetoencephalography, or MEG, is very similar to EEG in terms of the types of signals that it taps into. However, MEG scanners are typically very large and non-mobile, requiring a lab space with very little electromagnetic noise. As MEG has a very high temporal resolution, itis extremely good for understanding how decision processes unfold over time. But as a commercial tool, it suffers many of the issues we see with fMRI.
- Positron Emission Tomography, or PET scanning, would normally not make it to this list, but as others are putting it on as a neuromarketing measure it is worth adding it where it belongs. This is a brain-scanning method where a radioactive ligand is injected into the bloodstream of a participant (or patient) and it can be used to measure the level of dopamine, serotonin, glucose, or other substances. it is a great measure for understanding how these neurotransmitters and other ligands are at play in driving decision-making, but cannot in any way be considered an ethical, valid or reliable way for commercial use.
Measures that are used but need more validation:
- Galvanic Skin Response (GSR, or skin conductance, SC) — Although GSR has been used for decades, and many neuromarketing companies offer this method as an index of emotional arousal (or the more vague term “engagement”) a recent article suggested that the literature is highly fragmented and inconsistent in how the analyses are performed and reported, leaving us with a lower level of validation of the method for measuring consumer responses.
- Facial coding (FC) — Automated facial coding is all the rage in neuromarketing, so you might be surprised as to why it comes on this part of the list. It turns out that despite all the hype, facial coding is in a lot of problems. First, the largest and deepest meta-analysis to date shows that the theoretical basis of FC is problematic. Furthermore, it has proven hard to replicate many of the original findings that we show a universal and consistent expression of emotions. Finally, as a nail in the coffin, several studies have shown severe problems with automated facial coding solutions, such as how Decode demonstrated that most of the time in a study, the recorded expressions were neutral (hardly an interesting finding), and that the metric revealed facial expressions when they showed a static doll in front of the webcam. Besides this, there are to date no known studies that clearly demonstrate an added value of automated facial coding in understanding customer responses or in predicting subsequent behavior.
- Webcam-based eye-tracking — Some companies offer webcam-based measures of eye-tracking. However, it is a strong claim that webcam technology is sufficiently precise to record eye movements reliably. Indeed, recent studies also suggest that this method is definitely worth pursuing, but that “obviously there is a long road ahead of perfectly reliable and accurate online web technology-based eye tracking” and this paper suggesting that the method is interesting but should not be used for fine-grained metrics (which is what you need for proper neuromarketing analyses).
- Functional Near-Infrared Spectroscopy, or fNIRS — This is indeed a very promising method. The fNIRS very much taps into the same mechanisms as fMRI does, through the measurement of hemodynamic responses that are associated with brain responses. Also, recent advances in this technology suggest that it can be used in mobile settings. That said, although fNIRS is a very promising method, there is still too little experience with it for commercial and scalable use. For now, the method is reserved for the more academic branch of this research, through consumer neuroscience studies. This is, however, hopefully soon to change.
Are we at Neurons doing neuromarketing or consumer neuroscience research? We’re actually doing both. Our for-profit work focuses on providing the best possible metrics for our clients that help them better understand customer responses — this is neuromarketing work, as shown in our cases section. But on the other hand, we also run many basic science studies where we use consumer behaviors as a model to understand the brain, such as understanding pricing, travel preferences, and branding in the brain.
What should the neuromarketing toolbox include?
Based on this, it is pretty straightforward what you should do if you want to run a neuromarketing study.
If you are a company with a research budget, you should do the following:
- Start with NeuroVision! This tool is easy to start with, has a low cost that all can use, and helps you secure customer attention.
- Run a pilot study! Sometimes, it is a good idea to run a small pilot to ensure that you learn the methods, thinking, metrics, and outcome. From this, you and your team can prepare for the next steps.
- Do a full study with eye-tracking and EEG. This is by far the best combination in the field, as it allows you to both understand what customers are paying attention to, and how they respond emotionally and cognitively. These metrics also allow you to reliably compare products/ads and across different customer segments.
- Surveys and memory tests. While eye-tracking and EEG produces a great understanding of the direct and often subconscious responses, surveys and tests allow you to also understand the conscious side of customer responses. Here, interviews and surveys work well together with a detailed memory test at the end of the participant experience. This allows you to understand conscious liking and memories from the ads, products, or other experiences the participants have been exposed to. This also allows you to go back to the neuro data to understand why people differ in their memory and liking.
- Behavioral tracking. Many studies also allow you to combine neuro measures with behavioral tracking. Neurons has solutions for tracking how people, with consent, behave after being exposed to one or more of your assets. These measures can be combined with neuro measures to understand exactly what makes people make desirable behavioral changes.
For academic institutions, Neurons offers a full package solution that includes:
- Equipment acquisition
- Staff selection and training
- Protocol development
- From full turnkey study offers to data processing solutions
- Course curricula and materials
- Student tool offers, such as reduced costs on NeuroVision