GA4 Anomaly Detector

GA4 Anomaly Detector watches your Google Analytics 4 metric series — sessions, users, conversions, or any metric you choose — and flags the days that deviate from a rolling baseline, so a drop is noticed when it happens rather than at the end of the month. Detection is deterministic. It uses a z-score against a rolling baseline by default, or an interquartile method when your data is skewed, and every figure in the output is computed rather than narrated: the observed value, the expected value, the deviation, the score, the severity and a confidence rating. The restraint is deliberate and worth knowing. A language model turns the already-computed anomalies into a readable alert, but it never produces a number and it never asserts a definitive cause. A traffic drop has many possible explanations and an analytics tool that picks one confidently sends people down the wrong path. Without a model available it falls back to a deterministic template, so it always returns something usable. You can run it three ways. On the built-in sample to see the shape of an alert with zero setup. On your own series, pasted or exported from a GA4 report. Or connected to a GA4 property with a read-only service account, in which case it pulls the series itself — a sandbox or test property is recommended for the first run. `detect_anomalies` runs the detection alone over series you already have. `connector_status` tells you what the GA4 connection needs and whether it is configured, without ever revealing a credential value. It is read-only, its network access is restricted to the GA4 API alone, and credentials are masked and never logged. Built for analysts, growth teams and anyone who owns a metric someone will ask about.