What problem does it solve? Restaurant owners often run Instagram or Meta ads on gut feeling, with no way to know whether the ads actually brought customers in. This Skill closes that loop: it uses rocky-regi POS data (visit sources, customer segments, menu performance) to decide who to target and what to promote, designs the Meta ad campaign, and then measures actual in-store visits and lift against a holdout group to judge whether the ad worked. ## Core Features & Use Cases - Opportunity Detection from POS Data: Identifies the right campaign goal (new customer acquisition, churned-customer winback, boosting star menu items, rescuing slowing items) using customer segments, LTV, and menu engineering quadrants. - Ad Design with Approval Gate: Builds audience, creative, budget, and KPI proposals for Meta Ads MCP, but never launches without explicit owner approval; if Meta MCP is not connected, it outputs a manual execution spec instead. - Closed-Loop Measurement: Measures meta/instagram-sourced visits and sales, then judges effectiveness via holdout-based ROI (get_campaign_roi) or pre/post intervention analysis, honestly reporting "worked / didn't work / inconclusive". - Use Case: A manager asks "did that Instagram ad from last month actually work?" The Skill pulls visit-source sales, compares lift versus the holdout segment, and recommends either scaling the budget gradually or stopping the ad with a single root-cause hypothesis. ## Quick Start Ask the assistant to design an Instagram ad campaign to win back churned customers using the store's POS data and measure whether it brings them back.
