Apply Policies and Monitor Reliability

After profiling a dataset, the next step is to define what “good data” looks like and monitor it continuously. In ADOC, a policy is a set of Rules or Rule Sets that defines your expectations for a dataset. Each policy can include one or more rules, and a policy is considered passing only if all its rules are met.

Policies help you enforce data quality, freshness, reconciliation, anomaly detection, data drift, and schema drift rules, ensuring datasets remain reliable over time.

All policy actions are performed per dataset in the Asset Details page, which serves as the hub for monitoring and management.

Example:

• Data Quality Policy: Rules such as “customer email cannot be null” and “order total must be greater than zero.”
• Freshness Policy: Rules like “sales data must refresh daily before 6 AM.”

Workflow Overview

1. Open the Asset Details Page

1. In Discover Assets, locate the dataset you want to monitor.
2. Click the dataset name to open the Asset Details page.

2. Access Policies

1. Navigate to the Policies tab in the Asset Details page.
2. Here you can:
   • View existing policies
   • Create new policies
   • Modify rules or thresholds

Apply Policies

Policies are applied per dataset, and you can configure multiple policies for the same dataset depending on business needs. Each policy type has its own criteria and configuration:

• Data Quality Policy
• Reconciliation Policy
• Data Freshness Policy
• Data Anomaly Policy
• Data Drift Policy
• Schema Drift Policy

Rule Identifiers

When you create a rule within a policy, ADOC automatically generates a unique name for that rule. You can edit this name only at the time of rule creation. Once the rule is saved, the name becomes read-only.

The rule name uniquely identifies the rule within a policy version. It remains unchanged across policy edits as long as the rule definition itself is not modified. If a rule is updated — for example, its configuration or logic is changed — ADOC generates a new rule name for it. Rules that are newly added also receive new names, while rules that remain unchanged retain their existing names.

Rule names appear in execution results, are used for column labelling in Good and Bad records, and are included in API responses. This ensures that specific rules can be reliably referenced in reports, monitoring workflows, and external integrations.

Choosing the Right Policy Type

Policy Combinations for Comprehensive Monitoring

Complete table monitoring:

• Data Quality policy: Validate business rules
• Data Freshness policy: Ensure timely updates
• Schema Drift policy: Protect against structure changes
• Profile Anomaly policy: Detect quality degradation

ETL pipeline monitoring:

• Data Freshness policy: Monitor source data arrival
• Reconciliation policy: Validate transformation accuracy
• Data Quality policy: Check output data quality
• Schema Drift policy: Ensure schema compatibility

Machine learning feature monitoring:

• Data Drift policy: Detect feature distribution changes
• Profile Anomaly policy: Catch statistical anomalies
• Data Quality policy: Validate data completeness
• Data Freshness policy: Ensure training data is current

Getting Started Checklist

Phase 1: Critical Assets (Week 1-2)

• Identify 3-5 most critical tables
• Create Data Quality policies with 3-5 essential rules
• Enable Data Freshness monitoring with basic SLAs
• Set up notification channels

Phase 2: Data Movement (Week 3-4)

• Add Reconciliation policies for ETL processes
• Enable Schema Drift monitoring on production tables
• Configure alerts to appropriate teams

Phase 3: Advanced Monitoring (Month 2)

• Enable profiling on critical assets
• Create Profile Anomaly policies
• Add Data Drift policies for ML features
• Fine-tune sensitivity and thresholds

Phase 4: Expand and Optimize (Month 3+)

• Extend monitoring to more assets
• Optimize alert thresholds based on feedback
• Document data quality standards
• Establish incident response procedures

Monitor Compliance

Once policies are applied, ADOC continuously checks your data.

• Alerts are raised if any policy fails, allowing early intervention before downstream impacts.
• Execution history shows which rules passed or failed.
• Quality scores help track overall data health.

Example alerts:

• Missing or null values in a critical column.
• Daily data not arriving on schedule.
• Totals mismatching across systems.
• Revenue spikes outside normal trends.

Take Action

• Investigate alerts directly in the Alerts page.
• Use system recommendations to fix data issues.
• Adjust policy thresholds or rules as needed.

Managing Policies

• Export & Import: Share policies across teams or environments.
• Policy Groups: Organize related policies for easier management.
• Combine with Profiling: Policies rely on profiling statistics, so ensure profiling is up-to-date before applying policies.

What’s Next

• Choose a dataset that has been profiled.
• Explore policy types.
• Apply a policy and monitor alerts to ensure data reliability.