Solution
We built a Machine Learning system that understands the usual behavior of each KPI at each dimension (country, province, city, hourly, daily, etc) and the correlation with other KPIs.
Internal datasets like promotions, price changes, new system versions were used as contextual information to explain the anomalies.
Automated reports with summaries were set up through email, chat, and other channels to alert the relevant people and give additional in-depth information through custom dashboards.