Facial Coding is not a Silver Bullet (and there are no silver bullets)
- October 17, 2017
Michael Sankey and Ken Roberts, in a recent Greenbook Blog (“The Fallacies of Facial Exploring Prophecy Feelings vs Facial Coding”), observe that over a third of Fortune 500 companies are using facial coding for some aspect of ad testing. They note that this news worries them, because “facial coding for ad testing does not predict purchase behavior.” This claim deserves to be examined closely.
Sankey and Roberts go on to identify a number of flaws with facial coding-based approaches, including that emotional expression exists beyond the face, that facial coding should not be used for static content (i.e., print copy), and that facial coding faces technical challenges (individual differences in facial structure, inconsistencies in background lighting, cultural differences in expressiveness). Although these are sound critiques, the field is advancing rapidly in its ability to solve technical challenges, and limiting facial coding to dynamic content is not a fatal problem.
Sankey and Roberts also levy a more fundamental critique: that measures of facial expression at best only reveal the emotions being experienced by a respondent at the time, not anything about how these feelings relate to brand evaluations or purchase behavior. And here, I think, is the problem: facial coding – like many other recent innovations in our field – should not be used in a vacuum. It is a mistake to tout facial coding as a stand-alone skeleton key that unlocks what makes an ad successful. For this reason, facial coding must be used in a sophisticated way, not simply treated as a dependent variable.
Using facial coding as one component of an ad testing approach – along with traditional ad quality metrics and measures of brand and category affinity – offers powerful and holistic insights. With facial coding in the mix, it’s possible to identify the facial expressions that predict ad success in a given category, which can inform creative strategy. Looking more deeply, it’s possible to make finely grained observations like “the people who smiled at that joke at 0:14 are the ones who later want to buy more widgets” that can help refine creative content. Insights like these are valuable.
In sum, Sankey and Roberts have a point, but I’d amend their claim: “facial coding for ad testing [used alone and without integrating other insights] does not predict purchase behavior.” Facial coding is not a stand-alone solution to creative testing – but, despite its imperfections, it is a valuable tool in the creative testing arsenal.