Machine learning as an effective paradigm for persuasive message design
The impact of the development of the internet and new communications channels on the marketing industry pushed practitioners to devise new tools and approaches for influencing consumer attitudes and behaviors towards products and services. This has led to new insights into persuasive message design. In general, persuasive advertisement messaging can be viewed as a combination of context, i.e., a message, and additional affiliated components such as images, video, and special graphics. It is composed out of two main attribute categorizations: (1) textual content such as a product description or affiliated message and (2) sensible content such as product image, color, or even scent. In a competitive market in which consumers are constantly exposed to a hyper-abundance of products that also contain sensible attributes, it is crucial to design persuasive messages that will maximally appeal to desired consumers and evoke their positive response. Yet only a few studies have focused on the effective design of persuasive advertisement messages characterized by two integrated elements. As such, this research focuses on effective persuasive message design with integrated product scent and color attributes. We demonstrate how a machine learning process can be utilized to generate optimal persuasive messages by estimating the contribution of each message attribute to the final class attribute: the purchase intention response. Our results show that several prediction algorithms can enhance consumer response value. In addition, correlations between several attributes affiliated with the message can be derived by graph theory-based estimation. This research thus provides insight into attribute values important for management decisions, with implications for effective persuasive message design. Ultimately, this may lead to higher response rates for marketing practitioners in an increasingly competitive market.