HomeMachine LearningMachine Learning NewsUsing Machine Learning and Facial Tracking to Help Brands Accurately Reach Diverse...

Using Machine Learning and Facial Tracking to Help Brands Accurately Reach Diverse Audiences

American society is facing seismic cultural transformations. The country is more polarized, divided, and diverse than it has ever been. Multicultural audiences are becoming key drivers of growth across all industries, and younger generations’ shifts in values and consumer behaviors are resulting in significant changes in the consumer marketplace. 

Given this sea change, the level of audience complexity that corporate America is facing is more challenging than ever. Advertisers face unique problems particularly when they are looking to reach people of color and younger generations, and then reduce the risk of defection from others. Most brands struggle to interpret traditional ad testing results to understand not only the appeal of ads across segments, but what resonates and what doesn’t, and why. Brand managers are unsurprisingly anxious their video content will fail and generate harmful backlash. 

What’s more, younger generations are demanding change in how brands perceive and speak to them. They want brands to stand for meaning and purpose. Corporations can no longer get away with traditional messaging that tries to sell through a blanket “GenPop” message — or content that doesn’t reflect the cultural nuances of diverse segments. We know, however, that what appeals to some will offend others.

A major part of the problem lies in the fact that marketing leaders don’t have insight into the deep cultural and emotional influences that inform how consumers from different backgrounds process ads. This perhaps comes about from years and years of treating diverse audiences as extraneous, reflecting little understanding into the nuances of their cultures, which are a function of country of origin, race, ethnicity, language, class, gender, and other factors.

As marketers, we must move on to a reality driven by simple math and understanding of influence. It is not enough to win the Black and Hispanic consumer. Now, we need to understand the cultural influence that these growing segments are having on the vast majority of Americans. This is a new mindset. As marketers enter this new situation, few if any have access to best practices to manage this dynamic, and therefore no way of knowing what aspects of video content will generate groundswell and backlash and for whom. 

The answer lies in the powerful “Better Data,” generated by applying the latest innovative technologies 

Combining machine learning, sentiment analysis, and facial tracking offers one promising way to decipher crucial variations in response to ads across different demographics. For example, my team has been able to use this approach to better predict what generates Groundswell and Backlash, two critical new metrics, and dive into the drivers behind them. Facial tracking involves analyzing consumers’ facial patterns as they are watching ads. This facial tracking is combined with a survey analysis of ad features, stated emotions, and things like virality, brand fit, brand impact, to then understand ad effectiveness. Machine learning ingests the data, detects patterns, and becomes more predictable as the data asset grows.

As defined in our research, Groundswell takes place when viewers switch from a negative view of a brand to a positive one and Backlash happens when the opposite takes place. 

For example, an ad’s biggest risk is that it could provoke unintended audience backlash, as experienced by North Face in its “Wikipedia manipulation” advertising last month. We also saw this in 2017 with an ad for Dove featuring an African American woman removing her brown t-shirt to reveal a White woman underneath with a white t-shirt. The audience reaction was swift and viral, leading Dove to apologize, mentioning that they “missed the mark.”

Indeed when you track emotional responses cross-culturally, you’ll find very distinct differences. Facial tracking analysis of Bertolli’s 2016 “Dance” featuring a biracial couple reveals a remarkable 10-second divergence in emotional response between white and multicultural consumers. Other data has shown us that prestige is key to a positive reaction from White consumers but are not as important for Asian, Hispanic, and African American consumers. 

Analysis of Michelob’s “Pure Experience” 2019 Super Bowl ad featuring actress Zoë Kravitz reveals diametrically opposed responses across segments, especially by race and ethnicity. Asians and Hispanics experienced high Groundswell, but White and African American segments were far less positive. Using machine learning to analyze this and a range of beer commercials, we found that, while the ad over-indexed on ad features like tone and visuals, they don’t actually drive Groundswell. We have also found that Groundswell is highly correlated with Purchase Intent.

We’re discovering that playing to once-safe norms is not enough to develop video content that can speak to multiple audiences. Indeed, another salient insight we have found across all our ads is that many African Americans and Asians respond negatively to themes glorifying the American heartland, a sentiment reflected as well in free text responses. 

Three-part combination of machine learning, facial tracking, and sentiment analysis is helping many of America’s leading brands decipher where and what matters to changes in brand favorability and purchase intent. Whether used as a test for in-process ads or as part of a strategic review of what works in a category, this impressive innovation can distinguish which ads have appeal for whom, which are risky and why — giving marketing professionals a new edge into what drives purchase intent across diverse audiences.

Source link

- Advertisment -

Most Popular

- Advertisment -