07-28-2026
Companies introduce tens of thousands of new consumer products each year, yet many fail to generate sustainable sales. Research shows one in four is no longer selling after a year, and as many as 40% are off the market by the end of the second year. These outcomes represent substantial investments in ideation, prototyping, production and testing that never translate into lasting success.
Generative visual conjoint AI offers a different approach. Using disentanglement-based models, companies can automatically identify human-interpretable visual characteristics from product images and test how consumers respond to those characteristics before launch. This shifts learning earlier in the development process, when designs can still be refined at a lower cost.
The opportunity is especially significant in consumer-packaged goods, where roughly 30,000 products are launched annually — enough to fill an average grocery store. Visual conjoint AI can help brands pre-screen concepts, identify more promising designs and focus resources on products that are better aligned with consumer preferences.
Consumers often “hire” products to perform a particular job. By revealing which visual characteristics influence choice, visual conjoint AI can help companies design products that are better suited to that job and reduce the likelihood that new offerings are quickly discontinued.
Today, we revisit key research on generative visual conjoint AI and its potential to reshape product design and testing:
Revisit: Unlocking Product Design Insights with Generative AI
Ankit Sisodia is an assistant professor of marketing at the Daniels School of Business.
