Take a look at the submission A statement from Peter become one of the noninees:
To assess ad effectiveness I used facial coding software that tracks 491 superimposed key-points on a 3D artificial facemask. I gathered online 900 recordings (0.7 million frames) of facial reactions to advertisements and analyzed over 4.1 million unique data points. I reduced the data-gathering costs to 700$ by using crowdsourcing. I found that facial expressions explain 24-37% of variance of ad effectiveness  and that exaggerating (i.e. amplifying) one’s expressions leads to enhanced attitudes . One insight for business is that facial coding is reliable, objective and scalable.