Bookings dropped today — is it a real problem, or just a Monday?
The sample is synthetic: 20 weeks of daily bookings with a weekly pattern, growth, noise and three incidents planted in it.
Most KPI alerts compare today with yesterday, which fires every Monday. Here each day is compared with the median of the same weekday over the previous six weeks, so weekly patterns are built into the "expected" line.
Spread is measured with the median absolute deviation (× 1.4826, the usual scaling to match a standard deviation), with a floor of 6% of the baseline so routine noise isn't flagged. A day is an anomaly when it sits more than the chosen number of these robust standard deviations away. Medians aren't pulled around by the anomalies themselves, unlike a mean.
Each weekday needs at least three earlier weeks of history before it gets scored. This flags candidates to investigate (tracking, site, campaign, pricing); it doesn't tell you the cause.
Built by Keith Chiu with AI-assisted development. Your data stays in your browser — nothing is uploaded.