Incorporating Clicks, Attention, and Satisfaction into a Search Engine Result Page Evaluation Model
Topics: SERP Serving (Tangram/Glue), User Signals
The research paper by Google introduces a new evaluation model called the CAS model (Clicks, Attention, and Satisfaction), which aims to better predict user satisfaction on SERPs by incorporating not just clicks but also user attention and perceived satisfaction. The authors argue that traditional models, which rely solely on clicks, fail to account for “good abandonments” (cases where users find what they need without clicking) and more complex attention patterns due to the non-linear layout of modern SERPs. They propose an evaluation metric based on this CAS model, which better aligns with user-reported satisfaction than conventional metrics like DCG (Discounted Cumulative Gain).