Sales forecasting and anomaly detection for restaurant performance

- KFC Vietnam operates 250+ restaurants generating millions of daily transactions across dine-in, kiosk, delivery, and app channels. - Operations and finance teams currently detect sales anomalies reactively, only after a manager manually reviews daily reports, typically 24-48 hours after the fact. - There is no predictive baseline to distinguish a genuine performance issue from expected variance such as weather, promotion, or daypart. This delays corrective action and makes root-cause analysis time-consuming.

- Target daily store-level MAPE of 10% or better. - Reduce anomaly detection lag to under 2 hours. - Reduce manual variance investigation by about 70%.

KFC currently uses PowerBI for BI dashboards and forecasting.

KFC currently uses PowerBI for BI dashboards and forecasting.

- Operations team. - BI team. - Marketing team.

3+ years of POS transaction data at item, store, channel, and daypart level. Channel dimensions. Promotion and campaign calendar.

No specific requirements were submitted for this use case. --- *Track: F&B | Sponsor: KFC | Slug: fnb:sales-anomaly | Source: Agentic AI Build Week 2026 — Production snapshot 2026-08-11*