Utilizes machine learning to establish baselines and flag deviations in core metrics.
Automated anomaly detection continuously analyzes historical data trends to establish normal performance baselines. When a statistically significant spike or drop in key metrics occurs, such as an unexpected surge in organic traffic or a sudden collapse in e-commerce revenue, the system flags the event in the Insights dashboard. This proactive monitoring helps teams quickly identify issues like broken tracking or viral content without daily manual checks. However, the system only highlights anomalies; analysts must investigate the underlying dimensions and events to determine the root cause.
The detection engine uses proprietary statistical modeling to automatically identify and flag significant deviations in data trends across both standard and custom metrics.
Built natively into Analysis Workspace, the anomaly detection engine continuously evaluates historical data using advanced statistical algorithms (like Holt-Winters) to establish expected performance bands. When a metric breaches these predictive bands—whether it is a sudden spike in traffic or an unexpected drop in custom event conversions—the system highlights the anomaly directly within the trend charts. Crucially, this feature integrates seamlessly with the "Contribution Analysis" tool, which uses machine learning to automatically scan hundreds of dimensions to identify the potential root cause of the anomaly. This combination significantly accelerates troubleshooting for enterprise data teams. However, accurately tuning the statistical sensitivity requires historical data volume, and highly volatile seasonal traffic can occasionally trigger false positives.
Includes automated anomaly detection on metric charts, highlighting unexpected spikes or drops.
To quickly identify technical issues or shifts in user behavior, the platform features native anomaly detection. It automatically establishes expected confidence intervals based on historical trend data. If a metric or event volume deviates significantly from this baseline, such as a sudden drop in checkouts or a spike in error events, the system highlights the anomaly on the Insights chart. While effective for spotting irregularities during daily monitoring, it primarily serves as a visual aid within reports rather than a standalone alerting infrastructure.
Anomaly Detection automatically highlights statistically significant deviations in event volumes or conversion rates within standard trend charts.
To help teams proactively spot tracking issues or viral product adoption, the platform includes automated Anomaly Detection. By applying machine learning models (like Prophet) to historical event data, the system draws expected confidence bands on trend charts. When a metric spikes or drops outside this expected range, it is visually flagged for the analyst. This is highly useful for catching silent deployment bugs where a specific event stops firing, or for identifying a sudden surge in usage of a specific feature. However, it operates primarily as a visual aid on charts; it requires analysts to actively monitor their dashboards rather than serving as a completely separate, automated alerting system.
Anomaly detection automatically highlights statistical outliers in event trends and correlates them directly with qualitative session recordings.
The platform features native anomaly detection that automatically analyzes historical event volumes to establish expected trend baselines. When a metric deviates significantly—such as a sudden collapse in successful checkouts or a massive spike in API errors—the system visually highlights the anomaly on the chart. What sets this apart from competitors is the immediate next step: when an anomaly is flagged, analysts can click the outlier point to instantly watch the session recordings of the users affected during that specific timeframe. This drastically reduces the time it takes to diagnose whether a data spike is a tracking error, a backend bug, or genuine viral traffic.
