What problem does it solve? Restaurant owners and managers often guess whether a new menu item, price increase, or new concept will resonate with customers, and traditional surveys asking for direct 1-5 ratings collapse toward the middle. This Skill replaces guesswork with Semantic Similarity Rating (SSR): it builds personas from real customer segment data, lets an LLM voice free-text reactions, and maps those reactions onto five-point anchor sets to produce relative rankings. ## Core Features & Use Cases - Persona generation from real data: Pulls customer segments, RFM metrics, and basket analysis from the rocky-regi POS to condition 8-16 synthetic personas, forbidding imaginary personas. - Semantic scoring instead of direct ratings: Elicits free-text reactions and maps them to at least three five-point anchor sets via embedding similarity when a connector is available, or qualitative model estimation with strength labels when it is not. - Measured cross-validation: Cross-checks SSR rankings against actual CSAT survey results, dead-menu items, and menu engineering quadrants before recommending action. - Use Case: A manager asks which of three new menu ideas will land best. The Skill ranks them by segment with strength and distribution shape, flags that one idea resembles an existing dead menu item, and recommends a limited A/B test with a holdout group. ## Quick Start Ask the assistant to compare three new menu ideas using synthetic consumer research based on the store's real customer segments.
