A Smart System for Monitoring Your Data
Anomaly Detection continuously monitors your key metrics, alerting you when something unexpected occurs. It leverages advanced machine learning to learn what “normal” looks like, so you can quickly identify issues, opportunities, or significant trends.
How It Works
Establishing the Normal Range
- Learning Phase: The system analyzes historical data to understand typical patterns and fluctuations for each metric.
Complex Z-Score Calculation
- Deviation Measurement: Using a complex z-score method, the system quantifies how far current data deviates from the established normal range. A higher z-score indicates a larger deviation.
Alert Thresholds and Volume Threshold
- Alert Thresholds:
- Low Threshold: Alerts for even minor changes.
- Medium Threshold: Ignores small changes, alerting for moderate shifts.
- High Threshold: Alerts only for major, critical changes.
- Volume Threshold: In addition to the alert threshold, the volume threshold ensures that only anomalies with a significant volume and impact are flagged. If an anomaly breaches the alert threshold but doesn’t meet the volume requirement, it won’t trigger an alert—helping filter out noise.
Alert Delivery Frequencies
- Hourly Alerts: Continuously monitor data every hour and promptly notify you when anomalies are detected.
- Daily Alerts: Analyze the previous day’s data against a baseline built from around 100 days of historical data, providing a summarized view of anomalies.
What You See on the Dashboard
When you open the Anomaly Detection feature, your dashboard offers a clear and customisable view:
- Hourly and Daily Tabs: Toggle between hourly alerts and a daily alerts for a comprehensive overview.
- Date Range: Easily filter anomalies by selecting a specific time period.
- Metric and Dimension Filters: Narrow down your view to focus on specific metrics or dimensions that matter most.
Frequently Asked Questions (FAQ)
Q: What is Anomaly Detection?
A: It’s a feature that monitors your key metrics using advanced machine learning to spot significant deviations from what’s considered normal, helping you catch issues and opportunities early.
Q: What’s the Difference Between Alert Threshold and Volume Threshold?
A: The alert threshold sets how sensitive the detection is—lower values capture even minor deviations, while higher values focus on critical issues. The volume threshold adds an extra filter, ensuring that only anomalies with significant impact trigger an alert.
Q: How Do Hourly and Daily Alerts Differ?
A: Hourly alerts provide near real-time notifications by analyzing data every hour, while daily alerts give you a summary based on a longer historical baseline (about 100 days), making it easier to spot trends over time.
Q: How is the normal baseline used in alerts?
A: The normal or the baseline is the Z score that the Ai identifies. The Z-score is a statistical measure that indicates how far a data point deviates from the average. It is vital for accurately identifying anomalies in your data.