root-cause-investigation

To investigate a metric anomaly, compare current values to historical baselines, visualize trends, drill down by dimensions, test hypotheses, and check correlations to produce a defensible root-cause report.

The best way to identify the primary driver behind metric changes is to perform dimensional drill-downs on segments like geography or channel, validating the findings with hypothesis testing and correlation checks.

Dimensional drill-down works by segmenting aggregate data into specific dimensions like geography, channel, or device to surface the contributing segments that explain a metric anomaly.

No, you do not need external data sources to run this root-cause analysis; the investigation uses the provided data directly to perform baseline comparisons, correlation checks, and hypothesis testing.

A root-cause investigation report includes the primary driver of the anomaly, supporting hypotheses, validated correlations, and clear recommendations based on the analyzed data.

You should use hypothesis testing for reporting anomalies when you need to validate potential explanations for a metric change and quantify the impact of correlated factors.