How to monitor data anomalies without getting false alarms every day?

The biggest lesson I've learned in data monitoring is that setting thresholds by 'gut feeling' is a mistake. Setting an alert for a 30% increase in a metric might look responsible, but it ends up triggering all weekend, on holidays, and during promotional events, causing the business team to ignore it after a few days. Later, I started layering my metrics: core revenue and payment conversion get…