Value Measurement vs. Facial Coding

Facial coding infers emotion from facial movement, on the theory that expressions map reliably to internal states. That theory has not held up. A major review concluded that people scowl when angry far less often than the theory predicts, and in 2022 Microsoft withdrew its own emotion-recognition product, citing the absence of scientific consensus. A face tells you what someone is displaying. It does not tell you what they are experiencing.***

***NOTE: Facial Coding has been debunked by numerous studies conducted by Dr. Lisa Feldman Barrett and others. As she noted, "if you’re trying to build software or technology to identify scowls or frowns and pouts and so on and assume that means anger, good luck." Learn more about the challenges to Facial Coding.

WHAT IS FACIAL CODING?

Facial Coding claims to make an educated guess about which emotions are being shown by a person.

Facial Coding is founded on academic research conducted by facial expression expert, Paul Ekman, who purports that all humans share 7 universal emotional expressions: Happiness, Sadness, Surprise, Anger, Fear, Disgust, and Contempt. Ekman adapted a system for coding the presence of these facial expressions, called the facial action coding system (FACS), and he offers a certification program where individuals can be trained to code these expressions by watching for the activation of specific facial muscles. Despite the rich history of facial coding, much of this theory has been disproven, and emotions do not seem to be as universal as claimed.

As a neuromarketing methodology, facial coding involves using consumers’ own webcams to record their facial reactions. These recordings are then processed using automated algorithms designed to detect movement in facial muscles linked to Ekman’s 7 universal emotional expressions.

Facial coding results are typically output in aggregate, with metrics conveying the proportion of consumers who showed each of the emotions at some point during the experience. Output also typically includes a line graph depicting the proportion of consumers who expressed each emotion at varying points of the experience.

What is facial coding good at?

Facial Coding is a relatively affordable option in terms of neuromarketing solutions.

Facial Coding is relatively non-invasive since the cameras are simply looking at participants' faces, instead of requiring an uncomfortable cap of electrodes or awkward finger sensors.

Some software for Facial Coding does not require participants to be in a lab environment.

Where does facial coding fall short?

In recent research, several experts have contested the accuracy of Facial Coding. This research indicates that facial expressions are likely a poor indicator of true emotional experiences and a weak predictor of future behavior.

Facial Coding requires very specific environments to properly work with webcams. Consumers must watch in front of their computer, in a place with good lighting, and sit very still so as not to disturb the initial calibration.

Facial coding does not work well on non-white participants and is not culturally universal, thus it is not a good option for accurate market testing.

Facial Coding can be faked. Research has found that responses can be to some extent controlled by humans and therefore the results might be intentionally or unintentionally falsified.

Due to the challenges of using individual webcams, a good portion of Facial Coding data, especially when used by software, is considered "bad data" and must be thrown out. This requires companies to over-recruit to make up for the loss, and means you pay for data you can't use.

The intensity of facial expressions can vary by culture and societal norms, which can impact results.

Emotion perception is not culturally universal. Studies with populations outside Western samples do not reproduce the universal-expression finding, which means a model trained on one population systematically misreads others.

The commercial systems measure the wrong dimension. When automatic facial coding was benchmarked against psychophysiology, the software's "arousal" output actually tracked valence. That is a measurable failure of a shipping product, not a theoretical objection.

How Value Measurement is Different

Value Measurement is not based on what expressions are made externally while a person is having an experience, but rather what is actually going inside the brain, second-by-second. These results let you predict your audience’s future behavior with over 80% accuracy so you can confidently make impactful decisions.

Value Measurement data is clear, direct, and provides actionable insights. Because you can measure the value delivered by any experience in real-time, in the real world, it ensures the integrity of your results.

Facial Coding is simple, free, and bullish*t. There are over 1,000 open source libraries to add facial coding to your project and it has become very popular for research technology companies. But the core science is bad, and the results are meaningless. Sorry, not sorry.

What are the compliance risks of camera-based emotion detection?

Facial coding and facial recognition are the same technical operation — computer vision applied to faces — and the objections that have driven facial recognition out of public venues apply equally to scoring one. Deploying cameras to read emotion in a venue brings biometric notice and consent obligations with it, and those obligations attach to the venue and the client, not only the vendor. The EU AI Act restricts emotion recognition in workplace and education settings; Illinois BIPA and comparable state laws govern biometric capture; Colorado and California have added neural-data protections.

Value Measurement takes on none of it. No camera, no image, no biometric identifier. Participants opt in to a wristband, data is processed on the event iPad, raw signal is discarded, and the facilitator cannot re-identify any participant.

Dinner party fact for brain nerds

The human face has 43 different facial muscles that work together to create over 10,000 facial expressions. Facial Coding attempts to identify and understand 7 of these as "universal expressions".

(Makes you wonder what the other 9,993 expressions mean...)

Is Facial Coding a Value Measurement?

No. Facial coding measures expression, which is a social signal people control, suppress and perform. Value Measurement reads an autonomic response people cannot manage. One tells you what someone chose to show you; the other tells you what their brain did.

Core Value Measurement Features

FC

Value

NEUROLOGIC

OUTCOME-BASED

IN THE MOMENT

EVERYWHERE

COMPARABLE

MULTIDIMENSIONAL

Other Factors

FC

Value

NO OBSERVATION EFFECT

NO CUSTOM HARDWARE

SCALABLE

NO SPECIAL TRAINING REQUIRED

Feels like magic, because it is good science

Using everyday wearable devices, we can infer the simultaneous release of oxytocin and dopamine throughout the day.

Real-time measure

World anywhere with the devices people already wear.

Proven in science

25 years of peer-reviewed science back up our confidence

Learn more

Join the Value Measurement Movement

What are you waiting for? Value Measurement lets you identify what people authentically care about and will act on, moment-by-moment, removing all bias from the equation. Download our Value Measurement Manifesto and join the movement today!

Download the Manifesto