Overview

You can create monitors manually either from the Monitor List Page by clicking the New button in the top-right of your screen, or from an Asset Page in the Sifflet Data Catalog.

How to Set Up

Monitor Templates

First, select the asset you want to monitor. Then, choose the monitoring category to apply. Your options depend on your data, your business case, and the criticality of that data. A wide range of monitors is available in the Template Library.

Modes: Static, Dynamic, and Relative

You can base a monitor on static, dynamic, or relative thresholds. Dynamic thresholds are computed with machine learning (ML) models. Use:

• Static: for known business cases. For instance, when you want to monitor a price that must remain within a given price range.
• Dynamic: when you don't have specific thresholds but still want to detect any anomaly.
• Relative: to compare with the value from the previous run, against either a single threshold or a compound threshold (e.g., new value > old value + 10 OR new value < old value - 10).

If you use an ML-based monitor, an important parameter is the training history: the number of minutes, hours, or days of recorded data on which ML models train. Select this parameter according to the business periodicity you want the model to detect.

WHERE Clause

For all monitors, you can add a WHERE clause using SQL syntax to filter records with conditions.

GROUP BY Field

Finally, GROUP BY lets you increase the granularity of your monitoring. Each run of your monitor checks each unique value of the selected field separately.

How to Access Results

You receive a granular success or failure status for each dimensional value tested.