Anomaly Detection Skill

Analyze data to find anomalies, outliers, and unusual patterns right now. Use this skill when users ask questions like “are there any unusual patterns”, “find anomalies in my data”, “what looks abnormal”, or “detect outliers”. This is for immediate, ad-hoc analysis—not for setting up recurring alerts or notifications.

How It Works

The Anomaly Detection skill analyzes your data to identify statistical anomalies, outliers, and unusual patterns. When activated, the assistant:

1. Uses the Visualization skill to create an appropriate chart or table
2. Executes anomaly detection algorithms on the data
3. Identifies data points that deviate significantly from expected patterns
4. Provides visualizations highlighting the anomalies
5. With data sharing enabled, offers natural language summaries explaining what was found

The skill performs immediate analysis on your current data. It’s designed for ad-hoc investigation, not ongoing monitoring. For recurring anomaly alerts, use the Alert skill instead.

Examples

Finding anomalies in sales:

User: "Are there any anomalies in our sales data this quarter?"
Assistant: [Creates visualization, runs anomaly detection, highlights unusual patterns]

Detecting outliers:

User: "Find outliers in customer acquisition costs"
Assistant: [Analyzes data, identifies statistical outliers, visualizes results]

Identifying unusual patterns:

User: "What looks abnormal in website traffic?"
Assistant: [Runs anomaly detection, shows spikes/drops, explains findings]

Other use cases:

• “Find unusual patterns in customer behavior”
• “Detect outliers in website traffic”
• “Show me any unusual spikes or drops in the data”

Limitations

• Requires the Visualization skill as a dependency
• Provides immediate analysis only—not for ongoing monitoring
• Anomaly detection algorithms work best with sufficient historical data
• Results interpretation requires data sharing to be enabled