Point of View
However, since then, they have found out that turning cognitive science theory into concrete marketing actions is really difficult. Many attempts to borrow ideas from psychology, such as using metaphorical and projective scales or enhancing so-called "emotional" questions with icons or pictures of faces, are failing to live up to their billing; ultimately, these approaches yield the same results as more traditional measures. And as I have observed before (initially in my November 2006 Point of View "Neuromarketing: Beyond the Buzz"), neuromarketing's hardcore exotica simply don't operate at the scale and price point that are necessary in the day-to-day marketing world.
Thus many practitioners have not seen the value of giving up on the validation, norms, and benchmarks provided by established tools. And yet, having seen the possibility of tapping into consumers' unmediated responses, they can't help but struggle with the nagging thought that they are missing some crucial insights.
Fortunately, the mainstream research community now seems to be taking a new direction. We are seeing an explosion in the use of two approaches, automated facial coding and implicit association techniques, that directly measure unfiltered responses. By integrating these techniques into existing platforms, researchers are obtaining the rounded and holistic view they need to measure all facets of advertising response.
As the only major research agency with a global neuroscience practice dedicated to bringing these tools to market, Millward Brown is observing this shift first-hand. What follows are the reasons why we think these methods are being embraced.
While the coding of facial expressions to measure immediate emotional and cognitive responses was first established in the 1970s, it was not suitable for scale application in advertising research until recently, when companies such as Affectiva developed accurate software to automate the process. In partnership with Affectiva, Millward Brown has integrated this approach into standard Link copy testing surveys. After participants give their consent, a webcam films their reactions as they watch an ad; their responses are then coded and analyzed against more considered measures, yielding a full understanding of viewers' immediate reactions and the consequences for ad effectiveness.
A recent analysis of automotive advertising in the United States was refined and enhanced with the nuance provided by facial coding analysis. One of the ads researched was a spot for the Fiat 500 Abarth titled "Seduction." In the ad, a man walks down the street and is distracted by the sight of supermodel Catrinel Menghia bending over in the road. Spotting him, she angrily remonstrates with him in Italian for looking at her. Then she approaches him and, seductively stirring his coffee with her finger, whispers in his ear. As he leans forward for a kiss, she disappears, and he realizes that he was in fact looking at the Fiat 500 Abarth. The voice-over suggests that you will never forget the first time you see the car, and the ad ends with shots of the car in motion.
Men, who tended to focus on the model and her seductive behavior, found the ad highly engaging and enjoyable, but only moderately motivating and relatively weakly branded. Women were less positive about the spot; some expressed concerns that the idea was demeaning and sexist.
Facial coding helped clarify the conclusions by highlighting two key points. First, while women were less positive about the ad, they did find it funny, and they did like the sequence where the model catches the man looking and denounces him. Facial coding made it clear that concerns about the model's depiction being sexist were secondary and more considered responses.
Second, facial coding suggested that the spot's lack of persuasive power rested with the car itself. Viewers who, according to their survey responses, struggled to "get" the spot frowned and showed other clear expressions of distaste when the car was revealed. In the North American market, which is skewed more heavily toward large vehicles, the small and fiery but largely unknown Abarth appeared to be a disappointment to some viewers—especially men—compared to the idea personified by the model. Hence the ad was more limited in its motivational power than its high levels of engagement might suggest.
This sort of case illustrates the way in which direct measures of viewers' responses can yield a much better understanding of what makes the creative idea work. Clients have built upon this type of insight to evolve and extend their campaigns, refine edits, and make better decisions about creative direction.
In addition, we have found increasing evidence that facial coding can indicate an ad's likely inmarket effectiveness. We have already been able to relate viewers' expressions while watching an ad to the ad's subsequent sales effectiveness (as measured by econometric sales modelling). We have observed this relationship in two different markets in two different countries. In both cases, facial coding data added predictive power to Link. The main learning seems to be that negative expressions can be key indicators; if viewers looked disappointed by an ad, especially on the second viewing, the potential sales effectiveness of the spot was limited. It didn't matter if the disappointment stemmed from a poor reaction to the creative idea, or to the product claim, or to an incongruity between idea and brand—the key seemed to be that disappointing viewers is a really bad idea. Facial coding can effectively highlight that reaction.
Like facial coding, implicit association techniques have gained significant traction in recent years. Using these approaches, we measure the time it takes for people to react and make decisions when faced with particular stimuli. Using these measurements, we make inferences about the strength of associations based on the fact that decision times vary depending on whether someone's automatic ("fast") processing is consistent with or in opposition to their more considered ("slow") processing.
Millward Brown has found two approaches particularly useful, and has used them in hundreds of projects. The first approach, "emotional priming," is an adaptation of the implicit association test developed by Harvard University that measures the strength and direction of people's gut-level emotional responses. The second technique, a response-latency method we refer to as "intuitive association measurement," measures the relative ease with which people associate ideas with brands or ads. Associations that are made automatically are the most intuitive; those that are made only after consideration are less so.
When we apply the intuitive association technique to advertising, we ask people if an ad conveys particular ideas. Then we measure their response time in milliseconds. Using this data, we identify the responses that are faster or slower than we would expect for an individual to determine the most intuitive associations from the ad.
This approach allows us to address a criticism that is often leveled at advertising research: that the measurement of ad communication is overly rational. There is some justification for this concern. For example, in conventional research, to understand what an ad may be able to do for a particular brand, we ask people to think about the ideas conveyed by an ad. This gives us a clear picture of where an ad might be able to move a brand—if people are willing to invest some effort into it.
However, as the critics rightly point out, in real-life situations, people tend to engage with advertising superficially, if at all. Implicit association measurement shows us the associations that are made most readily, enabling us to predict which ideas will be registered by the ad, and which ideas are likely to have a role in real-world decision-making. In the Fiat example cited previously, "sexy" was the dominant association. Car-related ideas such as "sporty" and "stylish" also came through, but were more considered responses.
We also applied implicit association measurement to the recent Google Chrome ad "Dear Hollie," in which a father uses Google Chrome to send his newborn daughter a series of emails throughout her childhood, in anticipation of the time when they can read them together. The technique allowed us to identify "love" and "caring" as the most instinctive associations that people made with the ad. Other more generic brand associations such as "personal" and "innovative" were also made, but with a bit more effort.
Clients have used these approaches extensively in copy testing to get a more realistic understanding of the direction in which their campaigns are likely to take their brands, and in brand tracking work to understand the impact of those campaigns on the brand's instant meaning for consumers.
The essence of the science of "fast" and "slow" thinking is that both processes occur all the time; therefore, it is not realistic to believe that surveys and qualitative research measure only "slow" thinking. People's answers are influenced by fast processing as well, and we now have the scalable and pragmatic tools to tease out these influences. Intuitive association measurement allows us to sift through the full range of considered and less-considered reactions and extract the ideas that are most likely to have meaning when consumers are not motivated to think. Facial coding allows us to interview people with conventional surveys while also measuring their spontaneous responses, moment-by-moment, as the ad unfolds.
Ad research has evolved to encompass the measurement of people's gut reactions while also giving such reactions a practical and realistic role in ad evaluation and development. With thousands of projects being conducted using these approaches, measurement of fast thinking has gone mainstream.