Alert Policies

Telmai provides a set of out-of-the-box policies that are automatically created when a new dataset is added. Users also have the flexibility to create their own custom policies

Out-of-box Policies

Below is the list of pre-defined policies

Policy	What is Monitored?
Completeness Drifts	Changes in percentage of empty records
DQ Rules Violation	Detection for any rule violation
Data Difference	Addition/Removal of different records across tables. Note: Data diff must be enabled first
Data Drifts	Changes in various value formats (length, special characters, number of tokens, etc.)
Data Process Faillures	Detection of any scan failure
Freshness Drifts	Table and record level freshness drifts Note: For this policy to take effect, timestamp attribute must be identified
Numeric Value Drifts	Changes in averages of numeric values for an attribute
PII Exposure	Presence of PII data in your dataset (email, IP, etc). Note: This policy is disabled by default
Record Count Correlation Drifts	Changes in data volume across correlated tables with defined lineage. Note: For this policy to take effect, you need to define the tables lineage
Record Count Drifts	Changes in the number of scanned rows
Record Id Uniqueness Drifts	Changes in data uniqueness of record Id Note: For this policy to take effect, you need to enable table Id
Schema Drifts	Changes in schema; example: column added or removed
Top Values Drifts	This feature monitors the changes for the top 10 values. An alert is generated if new values are added or removed
Uniqueness Correlation Drifts	Changes in data uniqueness across tables with defined lineage Note: For this policy to take effect, you need to define the tables lineage
Uniqueness Drifts	Changes in data uniqueness across attributes
Value Distribution Drifts	Changes in the distribution of the top 20 most frequent categorical values
Correctness Drifts	Changes in the percentage of correct records based on defined expectations

Modifying/Reviewing Existing Policies

You can view, modify, or disable existing policies anytime by navigating to the “Alerting Policies” page and selecting the relevant dataset. The policies are displayed in a table with the following properties:

• Name: Policy name (User-defined policies are marked in yellow; Telmai out-of-the-box policies are marked in green)

• Metric: The monitored metric

• Status: Indicates whether the policy is enabled or disabled

• Actions:

  • Edit: Modify policy properties or scope.

  • Notifications: Toggle notifications on or off

  • Delete: Remove the policy (Available only for custom policies)

User Defined Policy

Users can create custom policies to monitor any of the calculated metrics. These metrics are tracked by specifying a threshold that triggers an alert when violated. The supported thresholds include:

Supported Threshold

• Telmai-ML Telmai’s built-in anomaly detection using machine learning

• Relative Drift Percentage change in data compared to previous scans or the average of scans

• Acceptable Range Check if a metric falls within a specified range

Steps to Create a Custom Policy

1. Navigate to “Alerting Policies” page & chose the “Policies” Tab

2. Click the "New Policy" button to start defining your policy.

3. A side menu will appear similar to this image

   
4. Select desired “Tracked Metric” from the list of

   • User-defined metrics, or

   • Health metric calculated by Telmai

5. Based on Metric selection, the user may need to select tracked attributes (ex: correctness, the user will need to specify associated attributes)

6. Select detection type Note: In case of Relative and Range detections, the user can specify how to handle the metric when not available (replace by zero or average)

Only follow the below steps if you want to be notified when an alert is generated Click Next to set notification destinations

1. Navigate to the Notifications tab within the New Policy page

2. Click +Add Channel

3. Select a value from the list of defined channels.

4. Repeat step 2&3 for all desired channels

5. After providing the required info, click Save