Skip to main content
Version: 3.2.0

Baseline Configuration in Thresholds

Baseline in thresholds allows dynamic comparison of current metric values against historical data patterns. It supports both simple time-shift comparison (e.g. yesterday vs today) and recurrent baselines based on daily or weekly patterns.

Custom threshold metric

To enable baseline comparison, activate Custom threshold metric.
This allows defining a specific data source, metric, and time range for the baseline calculation.

ParameterDescription
SourceData stream used as a baseline reference (e.g. netflowByIpAggr).
DirectionDefines traffic direction: Client, Server or Both.
MetricThe measured value (e.g. Sum Bytes).
Time rangeDefines the window of current data (e.g. Last 15 minutes).

Time shift

Specifies how far back the baseline should look.

  • Typical setting: 1 day
  • Shifting by 1 day compares the current 15-minute window with the same 15-minute window from the previous day.
  • Time shift = 0 includes today’s data in baseline calculations, which may lead to inaccurate results.

Recommendation: Always use Time shift = 1 day for daily and weekly recurrence to exclude ongoing data.

Recurrent interval

When enabled, baseline is computed from multiple historical intervals (rolling baseline).

ParameterDescription
Recurrent typeDaily - compares the same hour/minute ranges from previous days.
Weekly - compares the same day of week (e.g. Monday to Monday).
Start recurrence fromDefines the historical window used for averaging, typically Last 7 days.
Filter daysRestricts comparison to: Auto days, All days, Only working days, Only weekend days.

Auto days

Automatically adjusts the baseline according to the current day type:

  • If today is a working day, only working days from history are used.
  • If today is a weekend, only weekend data is used.

This ensures comparable context between weekdays and weekends.

Aggregation per category

Optional parameter for grouping baseline calculations by selected field (e.g. per Country).
This allows independent baselines for each group, improving anomaly detection accuracy.

Example:
Baseline of Sum Bytes can be calculated separately for each country, providing distinct thresholds per region.

Example configuration

Critical:
Sum Bytes > avg(Source: netflowByIpAggr, Metric: Sum Bytes, Timerange: Last 15 minutes, Time shift: 1d)

Baseline:
Last 7 days, recurrent interval: Daily, Filter days: Auto, Aggregate per category: Country

This setup compares the current 15-minute traffic volume to the average traffic observed in the same 15-minute window during the last seven corresponding weekdays or weekends.

Best practices

ScenarioRecommended settingsNotes
Daily comparison (standard)Recurrent type: Daily
Filter days: Auto
Time shift: 1 day
Start recurrence: Last 7 days
Most common and stable configuration. Suitable for most production environments.
Weekend/weekday differentiationRecurrent type: Daily
Filter days: Auto
Ensures weekend traffic is compared only to weekend data.
Weekly trend detectionRecurrent type: Weekly
Time shift: 1 day
Useful when traffic strongly varies by weekday (e.g., Monday spikes).
Short-term anomaly checkNo recurrent interval
Time shift: 1 day
Lightweight comparison with minimal historical load.
Granular baseline per groupAggregate per category: Country, Interface, or ServiceRecommended for multi-tenant or geographically distributed networks.
info

For high-volume datasets, limit recurrence range (e.g. 7-14 days) to optimize database performance.
Increasing historical depth (up to 28 days) may improve baseline accuracy when traffic patterns are stable.