Decoding Aesthetics: Machine Learning and Field Experiments on Visual Design Drivers of Demand for Indulgent Products

Products rely heavily on aesthetics to attract consumers, yet firms lack rigorous, scalable guidance on which visual design features increase demand and why. This gap is particularly consequential for indulgent and hedonic products, where purchase decisions often involve self-control conflict, anticipated guilt, and tension between immediate pleasure and longer-term goals. Furthermore, this is an economically important context because consumers’ choices are highly visual and frequently tied to occasions (e.g., celebrations vs. self-treats)that may amplify or attenuate guilt and justification.Partnering with The Cakery (a Hong Kong online-and-offline cake shop) and an interdisciplinary academic team, we will develop, validate, and apply an interpretable machine-learning model to extract aesthetic dimensions from product images and related metadata (e.g., consumer information, product information, marketing messages, and platform context). We will quantify which aesthetic dimensions influence demand using marketplace data and establish causal effects and mechanisms through a field experiment with The Cakery and controlled experiments that measure related psychological factors, such as self-control conflict, guilt, perceived indulgence, and quality inferences. The project will deliver publishable insights on aesthetic drivers of demand and actionable, scalable guidance for firms designing and marketing indulgent products.