Ir al contenido
Back to Anomaly Detection
Anomaly Detection Anomaly Settings

Configuring Anomaly Sensitivity

Fine-tune detection thresholds for your environment

Accessing Sensitivity Settings

Navigate to Settings → Anomaly Settings in the main navigation to configure how sensitive the detection system should be. Proper configuration helps balance catching real issues versus generating false positives.

Using Presets

For quick configuration, choose from our recommended presets based on your monitoring needs:

Sensitivity Presets
🔇
Quiet
Only critical issues. Best for stable environments.
⚖️
Balanced
Good mix of coverage and noise. Recommended.
🔔
Sensitive
Catch subtle issues. Good for critical systems.
⚙️
Custom
Fine-tune each setting individually.

Custom Configuration

Select the "Custom" preset to fine-tune individual settings for your specific needs.

Detection Thresholds
Warning Threshold (σ) 3.0σ
1.5σ (Sensitive) 4.5σ (Quiet)
Standard deviations from baseline to trigger a warning alert
Critical Threshold (σ) 4.0σ
2.5σ (Sensitive) 5.5σ (Quiet)
Standard deviations from baseline to trigger a critical alert
Cooldown Period 60 min
15 min 180 min
Minimum time between alerts for the same anomaly type
Minimum Deviation 50%
10% 100%
Ignore anomalies with less than this percentage deviation from baseline
Detection Features
Time-of-Day Awareness
Use separate baselines for each hour of the day to account for normal daily usage patterns
Seasonality Detection
Detect and account for weekly patterns (weekday vs weekend)
Correlate Related Events
Group related anomalies that occur together to help identify root causes
Maintenance Window Awareness
Suppress anomaly alerts during scheduled maintenance windows

Understanding Standard Deviation (σ)

The sensitivity thresholds are measured in standard deviations (σ) from the learned baseline:

  • 1.0σ: ~32% of normal values fall outside this range (very sensitive)
  • 2.0σ: ~5% of normal values fall outside this range (balanced)
  • 3.0σ: ~0.3% of normal values fall outside this range (quiet)

Best Practices

  1. Start with Balanced: Use the balanced preset initially, then adjust based on your experience
  2. Give it Time: Wait at least 2 weeks before adjusting settings to let the baseline stabilize
  3. Use Feedback: Marking anomalies as "Expected" helps the system learn your patterns
  4. Consider Criticality: Use lower thresholds for business-critical systems
  5. Review Regularly: Check your anomaly detection effectiveness monthly

Was this article helpful?

¿Aún necesitas ayuda? Contactar con soporte

Buscando...
Buscar en reuniones, grabaciones y participantes
Presione ESC para cerrar