DQ Rules
To set up automated DQ evaluation, you need to create DQ evaluation rules (DQ rules) to define criteria by which the data quality is measured.
There are multiple ways to create DQ rules in ONE. All of them involve defining the rule implementation logic (that is, the conditions used to evaluate records), testing the rule, and then publishing it.
All rules are saved to the rule library under Data Quality > Rules, where you can access and edit them. The rules can be re-used and applied to multiple attributes.
Ways to create DQ rules
- Rule creation in the rule library
-
This method gives you access to all possible rule-creation options, including variables, parameters, scoring, and option to generate rules with AI. It enables you to search for similar rules before creating a new one, helping to avoid duplicates. This approach is recommended when you want to create rules intended for re-use in multiple catalog items, or by multiple teams. See Create DQ Rule in the Rule Library.
- Quick rule creation from the attribute
-
Workflow optimized for quick and easy rule creation from your data. The rule template is prefilled with input and test data from the catalog item, and contains only the most commonly used options from the rule creation workflow. For more advanced options, such as variables, you can modify the rule in the rule library once it’s created. The rule is automatically applied to the attribute it was created from. See Quick DQ Rule Creation From Attribute.
This method is recommended to define business rules (rules specific to your use case).
- Rule creation using AI Agent
-
AI Agent can create rules in bulk, and apply them to attributes. See Create DQ Rule Using AI Agent.
This method is recommended to define technical rules (that is, rules that focus on integrity and structure of the data) and to create rules based on descriptions.
Best practices for creating rules
How to name the rule
Rule names should be phrased positively, that is, in a way that describes desired outcomes or acceptable conditions. This approach makes it easier to interpret results and reduces ambiguity.
For example, a 100% data quality result for a rule named "Id is unique" clearly indicates good quality. But with rules like "Id is not unique" or "Name is not null," a 100% result is ambiguous, making it unclear whether the result is positive or negative.
Next steps
Use the newly-created DQ rule to evaluate your data quality.
Was this page useful?