At the 2013 ESOMAR Annual Congress in Istanbul, Nick Langeveld and Rana El Kaliouby of Affectiva discussed their research with Mars into using facial coding to understand the relation between emotional ads and sales effectiveness. “Good ads solicit strong emotion: happiness, surprise, even crying a little,” said Nick. “But does that emotion translate into a sales event?”
To study this, Affectiva and Mars asked over 1,000 American, British, French, and German respondents to watch a number of web videos, from a library of 115 commercials from the categories of chocolate, gum, pet care, and instant foods. In turn, the webcam watched the respondents, using Affectiva software to track and code their facial responses throughout the experience of watching each commercial.
Facial coding captured expressions ranging from joy and surprise to confusion and disgust. How sensitive is the technology for picking up expressions? It does not just pick up the prototypical expressions: that wouldn’t be sufficient for ad testing, since most of the times ads don’t produce big emotional expressions. It can also capture subtle smiles and fleeting expressions lasting for just 100 milliseconds.
Research participants also provided self-reported data. Before and after they watched each ad, they answered questions about purchase intent and brand likeability. Affective discovered that commercials would change reported purchase intent but had little effect on brand likeability.
Affectiva used sales volumetric data for the four weeks before each ad aired and the four weeks during which it aired. Based on this sales performance, they classified 36 commercials as “bad ads” (causing a decline in sales), 25 as “neutral ads” (no effect on sales), and 54 as “good ads” (causing an increase in sales).
Chocolate commercials provided the most positive emotions, while food ads – being more informative – produced the least amount of positive emotions. Viewers in the US showed the most emotion, and the most positive emotion, while viewers in the UK showed the least positive emotion.
Affectiva then built a model to predict short-term sales from facial codes, using sales volumetric data for the four weeks before each ad aired and the four weeks during which it aired. They found that self-reported data on the quality of an ad was 69% accurate in identifying the potential effect on sales, while facial coding was 75% accurate. Combining the two measures produced even greater accuracy.
Self-reports are not always in sync with the facial coding. “Both converge in the M&M ads,” said Rana, “but where they diverge is in the self-report.” Nick pointed out that “moderator biases and societal norm biases” can reduce the quality of self-reported data. As an example, Rana said, “For a brand body lotion in India, the commercial showed a young woman, newly married, exposing her bare midriff, with her husband touching her midriff. Very provocative! Every single women who saw the ad smiled at this instance, but in the debrief most women didn’t report it and some women said they didn’t like it – they had to be politically correct.”
Rana said, “In our research, the more emotion you can get out of people, the more engagement you will get, the more sharing behavior, the more recall. You can definitely measure that from the face. Most of our research we do online, watching the subject, which is more valid than having them in a lab, so we are step closer to the real experience, but we are not quite there, where it is the average consumer sitting in front of a TV.”
In the future, Affectiva expects to integrate heartbeat monitoring, which is a sign of arousal (engagement). “We are looking at measuring heart-rate responses from webcams,” said Rana.
Affectiva concluded its presentation with a testimonial from Laurent Larguinat of Mars, who said, “Combining Affectiva’s technology and Mars’ marketing expertise, we’ve significantly increased our understanding of the role of emotions in advertising. This knowledge will shorten the odds of producing advertising that improves our sales performance and perhaps even help us achieve ‘lightning in a bottle’.”
Author Notes:
Jeffrey Henning
