Observational feedback refers to feedback in learning settings that requires observations of student actions. Existing feedback models provide limited insight into the way teachers generate such observational feedback. Addressing this gap, this exploratory study investigates how experienced nursing teachers generate observational feedback in high-fidelity simulations. Using the Critical Incident Technique (CIT), we asked teachers to identify relevant incidents during two video-recorded student simulations and how they would generate feedback on these incidents. A qualitative content analysis revealed a structured four-step process: noticing multimodal cues, interpreting student actions, appraising performance against professional standards and devising feedback interventions. Additionally, teachers relied on their domain-specific and pedagogical expertise to contextualise observations, anticipate student reasoning, and determine how to engage students with feedback. These findings extend existing feedback models by unpacking the observational feedback generation process and highlighting how domain-specific and pedagogical expertise shapes the process. This article also has direct implications for simulation-based education. To generate contextually relevant and pedagogically meaningful observational feedback, teachers need to develop the ability to notice relevant student actions, elicit students' reasoning, appraise performance based on both clinical standards and student learning needs, and devise feedback that fosters learning.