Refer to this topic for information on fixed issues and known limitations associated with Data Quality.
| Issue | Summary | Workaround | Noted on |
|---|---|---|---|
| Run parameters with AI Assist in the Custom coding step | AI Assist does not support run parameters in the Custom coding step. A script generated by AI Assist does not declare run parameter tokens, so the Select properties to map panel shows no properties for the step. | Declare each token manually in the script in the @{name:default} format, then map it to a run parameter with the same name as the token. For more information, see Configure run parameters for quality pipelines. |
November 2026 |
| Cross-environment run configurations with Databricks target datasets | When a Databricks target dataset is selected in a cross-environment run configuration that uses an agent-based pipeline engine, cluster-based cross-connection configurations can perform very slowly. | Recommended: Use SQL Warehouse when a Databricks target dataset is selected. | September 2026 |
| Limitations for the Candidate Finder operator |
As a result, ordering or grouping may vary depending on partitioning strategy and data distribution. |
NA | September 2026 |
| SQL rule queries with duplicate output column names | When failed-record capture or output is turned on for an SQL rule, a query that returns more than one output column with the same name, including names that differ only by case, can treat the columns as ambiguous and can fail. If failed-record capture and output are turned off, the rule run can complete. This can occur when a query selects fields with the same name from joined datasets. |
Assign aliases so each selected output column has a unique name. Instead of: Use: |
September 2026 |
| Snowflake NUMBER(38,2) Columns May Lose Decimal Precision |
When a Snowflake NUMBER(38,2) column is cataloged in Data Integrity Suite, it may be interpreted as an Integer. As a result, the pipeline may store the column as BIGINT or NUMBER(38,0), causing the decimal portion of the value to be lost. This is an existing product limitation and is not related to the recent data type mapping update. |
|
August 2026 |
| Verify and Geocode operator across different DBR versions | The Verify and Geocode operator is not supported across different DBR versions when the pool type is arm type. | NA | August 2026 |
| Multi-level matching issue on Snowflake runtime | The Create unified match groups across rules option does not work with the Multi-Level Matching with Transitive Grouping (Inter and Intra Flow) step on the Snowflake runtime environment. For more information, see Snowflake support case. | NA | July 2026 |
| Custom data quality rule evaluation with CSV files | When evaluating custom data quality rules using assets stored in CSV files within the Amazon S3 datasource, the process tends to take longer compared to working with other file formats such as Parquet or JSON. This delay occurs because, during the rule evaluation, all fields in the dataset are read from the CSV file, rather than selectively accessing only the required fields. | NA | February 2026 |
| Geolocation API source preview fails with "NullPointerException" | When you attempt to preview a transformation step that uses the Geolocation Address API source in a data quality pipeline, the preview operation fails. This limitation affects only the preview functionality in the transformation step editor. | NA | December 2025 |
| Step preview row limit in Data Quality pipelines | The Step Preview feature supports a maximum of 500 rows in total, distributed across all input sources for an operator. For example, if a step has two input datasets, only 250 rows from each will be included in the preview. This may result in incomplete previews when working with larger datasets. | For Upload File sources used in sample generation, ensure that the most relevant records are positioned at the top of the file. | June 2025 |
| Quality score generation with profiling | When configuring a datasource with Profile Datasets toggle enabled and the Run Quality Rules and Calculate Scores toggle disabled it is noticed that a quality score is generated for certain datasets, providing additional insights beyond the expected behavior. | NA | March 2025 |
| Data-type limitation in quality rules | The 'Time' datatype is not supported when setting up Catalog rules for Dataproc or BigQuery connections. It defaults to 'String', which may affect rule accuracy. | Currently, there is no direct workaround. To ensure accurate rule evaluation when using Dataproc or BigQuery connections, consider using alternative data-types. | December 2024 |
| Agent selection impact | If no agent is selected, the Insights feature will be disabled due to the absence of agent selection. This means that in order to enable the Insights feature, an agent must be actively chosen. | NA | November 2024 |