AI can predict what people would buy with 90% of human accuracy. The ETH Zürich and University of Mannheim study used 9,300 real survey responses, without conducting a single human survey.
Instead of asking consumers, “How likely are you to buy this on a 1-5 scale?” The model uses Semantic Similarity Rating (SSR), which interprets open-ended phrases like “I’d definitely buy this” or “Maybe if it’s discounted”. It then translates them into numeric purchase-intent ratings.
This opens the door to synthetic consumer modelling, where AI can mirror human responses accurately. It could greatly reduce reliance on costly surveys, panels, and focus groups, allowing marketers to predict consumer preferences faster and at scale.
Can Semantic Similarity Rating (SSR) make AI better at predicting what you will buy next?
Learn more here: AI Models Might Be Able to Predict What You’ll Buy Better Than You Can