What problem does it solve? When a restaurant manager notices a dish that used to sell well has stopped selling, this Skill prevents knee-jerk reactions like blind discounting. It distinguishes genuine slowdown from a structurally dead item using measured POS data, generates hypotheses about why customers skip it, branches into the right intervention (visual refresh, removal, or promotion), and always returns to verify whether the action actually worked. ## Core Features & Use Cases - Slowdown vs. dead-item classification: Uses store insights, dead-menu detection, menu engineering quadrants, basket analysis, and cancellation reasons to determine whether a dish is genuinely declining or was never viable. - Hypothesis generation via synthetic consumer research (SSR): Builds personas from real customer attributes and maps their reactions to anchors like price, appearance, or lack of demand, clearly labeled as reference hypotheses rather than facts. - Branched intervention with approval gates: Routes to visual refresh (AI-generated imagery with compliance checks), irreversible removal (owner publish approval required), or promotion (LINE coupons with mandatory holdout groups). - Closed-loop effect verification: After 7-14 business days, measures lift, holdout, and confidence via campaign ROI and intervention-effect tools, rolling back if results are not significant. - Use Case: A manager asks why the seasonal pasta stopped selling. The Skill confirms it is a true slowdown (not a dead item), finds pricing headroom, generates SSR hypotheses pointing to weak visual appeal, produces a new photo for approval, relaunches with a holdout coupon, and schedules an ROI check two weeks later. ## Quick Start Ask the assistant why a specific menu item has stopped selling recently and what to do about it, then follow the proposed diagnosis and intervention plan.
