Emotional Recognition using Facial Analysis
Do you have a poker face or does your expression give you away? According to psychologists, a lot of emotional information can be drawn from our facial expressions, which is valuable to us as market researchers as we are interested in the emotional responses driving people’s attitudes and behaviours. We recently partnered with a tech company called Affectiva which has analysed over 5 million faces and 24,000 adverts to understand how people respond to digital stimulus such as advertising, websites and apps. Our client wanted to understand physicians’ response to communications materials designed for a disease awareness campaign, so their technology was ideal for the project’s objective. Read the full case study here
Sounds interesting, how does it work?
Affectiva says it measures facial expressions of emotion and provides analytics to help market researchers understand how the emotional content impacts sales lift or purchase intent. Their computer vision algorithms identify key landmarks on the face. Then, machine learning algorithms analyse pixels in those regions to classify facial expressions. Combinations of facial expressions are mapped to 7 key emotions - anger, joy, sadness, surprise, fear, contempt and disgust. Affectiva has a large database of norms against which your survey can be compared including geography, product category and media.
Is it effective?
The analysis did give additional depth to our findings and worked really well with video output. Analysis can be subjective, so we do recommend employing it in combination with qualitative questioning. However, for this project, we were able to identify specific parts of the campaign that were strongly evoking particular emotions of key interest to the client.
What were the challenges?
We found the software easy to work with - all you need is a standard webcam and internet connectivity. However, we did find that respondents are not always willing to consent to their webcam being accessed, even if they said they were fine with it initially. Researchers will need to factor this in, especially if data needs to be analysed at a more granular level, for instance, by market.
Sign up to receive Rapport
Rapport is our e-newsletter and online resource for sharing our expertise and experience in global healthcare market research. Sign up here »