Business Problem & Solution

Every hospitality business has a handful of numbers that matter — bookings, revenue, rating, cancellations, support SLA, occupancy — and the failure mode is always the same: a number moves, nobody notices, and it shows up in a board deck two weeks later. Dashboards do not solve this; someone still has to be staring at them.

Anomaly Detection & KPI Alerting watches the series for you. It turns the ecosystem's event roll-ups into trusted KPI series and runs statistical anomaly detection over them, surfacing only the moves a human should actually look at — with the expected value, the observed value, and how many standard deviations apart they are.

It proves that anomaly detection is a deterministic, explainable layer, not a black box: a trailing-window z-score plus a drift test decides what is anomalous, the threshold is an owner-tunable knob, and the optional AI-assist only explains an already-detected anomaly. Two anomalies are deliberately injected into the seed so the board is always alive to demonstrate.

It simplifies the ingestion. The KPI series are generated from the synthetic event history rather than streamed from a live warehouse, and the AI explanation is a simulated, metered narrative. Detection itself costs nothing — it is pure math.

Synthetic KPI series only — six metrics over a 60-day window, with two injected anomalies for the demo. No real customers, no PII. Figures are representative and labelled; the AI-assist explanation is a hypothesis, not a verdict, and says so.