03-11-2025
A forthcoming paper in the Journal of Marketing Research reveals how firms can leverage generative artificial intelligence (AI) to quantify and optimize the visual components of product design. Authored by Ankit Sisodia, assistant professor of marketing at the Daniels School, together with collaborators at Yale University, the work introduces a method to automatically discover the product features that most influence consumer choice — without manual labeling of images.
In “Generative Interpretable Visual Design; Using Disentanglement for Visual Conjoint Analysis,” the authors show how disentanglement algorithms can automatically identify and measure core visual components — such as color, shape and size — from simple product images, eliminating the need for expert labeling. Structured data (e.g., brand or price) guide the AI to learn these factors on its own, and the model’s generative power then enables realistic “what-if” product images by varying each characteristic in isolation. This setup enables visual conjoint analysis, letting marketers test consumer preferences for design changes more rapidly and cost-effectively than traditional methods.
Visual appeal is central to consumer decisions, yet difficult to manage. By extracting a product’s “visual DNA,” companies can refine or create new features that resonate with different market segments. In one example, the researchers applied their method to a large watch image dataset and generated “ideal point” designs — featuring specific dial and bezel colors, strap style and crown size — that garnered more consumer interest than existing models.
The implications of this research stretch beyond watches. From apparel to cars to digital products, any category where visual aesthetics play a key role in consumer choice could benefit from the technique. By combining unstructured image data with standard product information, companies can streamline product innovation processes and run “virtual tests” of design tweaks before incurring manufacturing costs.
Sisodia and his co-authors from Yale joined Brett Gordon, co-editor of the Journal of Marketing Research and the Charles H. Kellstadt Professor of Marketing at Northwestern University’s Kellogg School of Management, on an October 2024 episode of “How I Wrote This.” The interview is available on Apple Podcasts and on Spotify.
Readers may contact Sisodia for more information about the research at asisodia@purdue.edu. Co-authors at Yale are Alex Burnap, assistant professor of marketing, and Vineet Kumar, associate professor of marketing.
