What problem does it solve? When a store manager or HQ notices a sudden sales drop at one location, it is easy to overreact to noise or launch untargeted promotions. This Skill first verifies whether the drop is a genuine anomaly against an expected range, then attributes the cause using in-store data, identifies churning customer segments via RFM, and connects the diagnosis to a holdout-measured win-back coupon campaign. ## Core Features & Use Cases - False-alarm gate: Compares actual sales against the expected range (mean±std) from similar business days and stops the workflow when the drop is within range or the sample size is insufficient. - Cause attribution: Cross-references cancellations, hourly heatmaps, menu engineering quadrants, and survey feedback to determine whether the drop is operational or demand-driven. - Churn-to-winback chain: Extracts real churning segments via RFM, generates SSR hypotheses about why they left, prepares approved visuals, and sends LINE coupons with mandatory holdout groups for ROI measurement. - Use Case: A manager asks "Did Store A's sales drop yesterday?" The Skill confirms the drop is outside the expected range, finds dinner-hour Star items collapsed, identifies 200 churning dinner regulars, and prepares a holdout-measured win-back coupon pending owner approval. ## Quick Start Ask the assistant to check whether yesterday's sales drop at a specific store is a real anomaly and, if so, diagnose the cause and prepare a win-back campaign for churning customers.
