Can tailoring ads generate considerable customer engagement and brand loyalty?
Maarten Bos, a research scientist at Disney Research, helped conducted studies about tailoring ads that are both appealing and effective.
Personality types vary, which suggests that tailored advertisements may be appealing and effective. In a series of studies, Bos helped investigated both the image preferences of individuals as well as responses to a personality questionnaire.
The investigation consisted of four parts; the first part comprised images rated by participants whose personalities were measured; each image was rated for specific features, such as presence of people in the picture, whether those people looked into the camera, etc.
Subsequently, computational methods were used to extract image features, ranging from contrast and color and content and image composition. After this data was combined, predictions could be made as to whether a new set of participants would prefer certain images.
Bos said the studies he was involved with focused on how Disney could do this, not on how it currently does it.
“Our studies at Disney Research showed that tailoring advertising could be useful,” Bos told Loyalty360. “We did not test advertising effectiveness at Disney (which is our parent company), but we created a model of which personality type might prefer which kind of image features. With that, we can predict above chance level which images will be liked by which kind of personality type. The effects are small, but in a sufficiently large advertising campaign, this could be useful. For our work, you’d need to have access to people’s personality scores, though, and that is a big conditional.”
What were your biggest takeaways from the personality type section of the research?
“We did not type Disney guests/consumers, but a randomly selected group of people from Amazon Mechanical Turk,” Bos explained. “We measured their personality and mapped that onto the features of images they rated. Personality traits are randomly distributed in any sufficiently random sample, so we did not learn too much about personality itself. We did learn which personality type likes which kind of image, but keep in mind, that the effect sizes (how strong this effect is) is small. I can start any sentence on our findings with ‘we found that we can predict above chance level that …’’’
What advice would you give marketers that want to gain maximum exposure through effective, targeted messaging?
“In the end, a marketer would have to weigh several things in their choice to tailor or not,” Bos added. “Some of are: The cost of finding data that allows you to tailor; the quality of the data that allows you to tailor; and the cost of getting tailoring wrong. To give an example we didn’t study: My data may say that my audience is all or largely Spanish speaking. I could then decide to tailor my ads to be all in Spanish. If my data is of low quality, then that will be a costly mistake because, not only will people not feel like the ad is for them, they will potentially not even understand it.”