Continuous monitoring of sales, tracking and infrastructure

Key metrics — sales, payments, ad tracking, site performance — were checked manually once a day. Problems that started at night or on weekends were found late, and slow degradation went unnoticed.

1. Connections to the store platform, payment data, analytics and server monitoring
2. Baselines by hour and weekday learned from historical data, plus explicit thresholds for critical metrics
3. Anomaly alerts in the team's messenger with context: what changed, by how much, since when
4. A weekly trend report

Incidents such as broken conversion tracking were flagged within the monitoring interval instead of being discovered the next morning. Gradual changes — for example, slowly growing server response time — showed up in the weekly report before they turned into outages. Nobody needed to open dashboards just to check that everything was fine.