Version 26.9.0

Date: 20th September 2026

Data Reliability

• Consistent Reliability Scores Across ADOC: Reliability scores are now computed using the same formula and displayed consistently everywhere they appear — Discover Assets, Asset Overview, the Policies tab, Reliability Explorer, and Data Products (Marketplace and Manage Overview) — so the same asset shows the same score on every page. Scores reflect each rule's most recent execution result, aggregated according to the scoring methodology selected under Settings > Score aggregation methodology (Policy Based Simple Averaging, Rule Based Simple Averaging, or Rule Based Row Weighted Averaging). The score delta shown alongside each score reflects the change from the previous execution, and is no longer scoped to a fixed time window. For more information, see Score Aggregation Methodology.

• Absolute Change and Percentage Change Operators for Data Equality Reconciliation: Data Equality reconciliation rules now support two new operators — Absolute Change and Percentage Change — that match source and target values within a defined tolerance instead of requiring an exact match. Absolute Change accepts a fixed amount of deviation from the source value; Percentage Change accepts a percentage of the source value; both support symmetric, one-sided, or asymmetric tolerance using independent Lower Bound and Upper Bound settings. For more information, see Threshold-based matching (Absolute Change / Percentage Change).

Data Sources and Connectors

• Global Storage (S3) Support for Redshift Serverless (Pushdown Engine): ADOC now supports persisting data quality policy results to Global Storage (S3) when running Redshift Serverless policies in Pushdown engine mode. Previously, this persistence was only available on the Spark engine, forcing customers to choose between Pushdown engine performance and S3 result storage. Good and bad row records are now accessible both through the Data Violations view in the ADOC UI and directly from your S3 bucket. For more information, see Configure Redshift Spectrum for External Tables.

• Crawler No Longer Archives Policies When S3 Assets Are Temporarily Unavailable: The S3 crawler no longer deletes an asset when its underlying file or object is missing during a crawl — for example, due to a delayed upload or a brief storage interruption. The asset is now retained in its last-known state, so policies attached to it stay active and continue running on their configured schedule without requiring manual intervention. For more information, see Crawler Behavior on Temporarily Unavailable Assets.

• Configurable Crawler Batch Size for S3, Kafka, and Pub/Sub Data Sources: Crawlers for AWS S3, Kafka, and Pub/Sub data sources can now crawl multiple assets in a single batch instead of one asset at a time, improving crawl speed and preventing a single asset's failure from stopping the entire crawl. The batch size is controlled by the CRAWLER_ASSET_BATCH_SIZE setting (default: 5), configurable per data source under Crawler Resources, or globally for all data sources under a Data Plane's Application Config. For more information, see Amazon S3, Apache Kafka, and Google Cloud Pub/Sub.

Data Products

• Custom Fields for Data Products: Administrators can now define custom metadata fields — Text, Number, Date, DateTime, Email, Checkbox, Pick List, or Multi-Select — that appear automatically on every Data Product create and edit form. Fields can be marked required, given a default value, and optionally surfaced as a Marketplace detail or as a filter in the Data Products listing. For more information, see Data Products | Custom Fields.

• Data Product Tags — Associations and Filters: Data Products can now be tagged using ADOC's unified tag model, bringing them into the same tagging framework already used for assets, policies, and other catalog objects. You can associate one or more tags from the shared tag library on any Data Product's detail page, and filter the Manage Data Products listing by tag to quickly locate related products. For more information, see Data Products | Detail Page.

Pipelines

• AWS Step Functions Pipeline Integration: ADOC now supports AWS Step Functions as a pipeline source, giving you visibility into Step Function executions alongside your other orchestrators. States within a Step Function appear as job nodes on the pipeline canvas, with nested visualization for Map and Parallel states. When one Step Function triggers another, you can link the two executions into a single connected pipeline view in ADOC by passing parent run context between them — the same pattern used for Airflow-to-Spark and Airflow-to-Dbt lineage. This integration supports Standard Step Functions only; Express Step Functions are not supported, and results reflect a completed execution rather than a live, in-progress view. For more information, see AWS Step Functions.

• Cross-Layer Pipeline and Data Lineage Stitching: ADOC now automatically stitches lineage across orchestration, processing, metadata catalog, and storage layers into a single, navigable view — connecting an Airflow DAG task to the Databricks, Trino, AWS Glue, or dbt Cloud job it triggers, and resolving that job's input and output assets against the Glue, Unity, Trino, or Redshift External catalog. The Pipeline Details lineage graph now renders this as a multi-level hierarchy, with triggered pipelines shown as expandable nodes so you can drill into their upstream and downstream assets without leaving the top-level view. For more information, see Pipeline Run Details.

• Namespace-Based Filtering for Spark OpenLineage Pipelines: The Pipelines overview now supports namespace-based filtering when Spark is selected as the pipeline source, bringing it into parity with other OpenLineage integrations such as Airflow. Selecting one or more namespaces scopes the pipeline list, summary metrics, and all overview charts to just those namespaces, giving teams a focused view of a specific cluster, environment, or team's Spark workloads. For more information, see Filtering and Searching Pipelines.

• Optional Control Over Built-In dbt Artifact Extraction for Airflow-to-dbt Cloud Lineage: The acceldata-openlineage package (v1.1.0) adds a new dbt_cloud_lineage_extraction_enabled setting that lets you skip Airflow's built-in dbt Cloud artifact extraction when the ADOC dbt Cloud Data Source connector already collects model, test, and dataset lineage, avoiding duplicate or conflicting lineage. This setting defaults to true, preserving existing behavior for current deployments, and requires no configuration change unless you want to disable built-in extraction. For more information, see Link Airflow Tasks to dbt Cloud Pipeline Runs in ADOC Cloud.

• Configurable Delay for Pipeline Automations: Pipeline automations for Profiling, Data Quality, and Reconciliation can now be configured with a delay, in minutes, before they run after their trigger (Pipeline, Job, or Span) completes. This helps stagger automation execution across a shared Data Plane cluster, reducing resource contention when multiple automations are triggered at the same time. For more information, see Pipeline Run Details.

Security & Authorization

• Resource-Based Access Management (RBAM) for Pipelines: Pipeline access can now be scoped by domain, matching the access model already available for assets and reports. Admins can group pipelines into Pipeline Resource Groups — by data source or individual pipeline selection, or by matching pipeline names against one or more regex patterns — and grant view or modify access to specific teams via a Domain Role's new Pipeline Management permissions. New pipelines, whether pulled in or pushed via OpenLineage, automatically join any resource group whose pattern matches, with no manual tagging required. At rollout, every tenant retains a default resource group covering all pipelines, so existing access is unaffected until an admin deliberately narrows it. For more information, see Pipeline Groups.

• Resolved an issue where table references inside Common Table Expressions (CTEs) and subqueries in Custom SQL policies were not transformed during policy migration when using asset mapping, causing migrated policies to retain source environment table paths. (Manonandan)

• Resolved an issue where Collibra OAuth token requests lacked retry logic, causing intermittent policy quality score synchronization failures during transient network interruptions or token renewal delays. (Rahul Kumar)

• Resolved a UI styling issue where the "Add terms" modal dialog box in the Data Products Terms tab had an unfixed height, leading to layout shifts and clipped form controls.

• Resolved an issue where data quality rules associated with previously deleted policies continued to appear on the Rules listing page.

• Resolved an issue where persisted column selections were unexpectedly cleared and reset to defaults after revalidating SQL queries while editing Data Quality Custom SQL policies. (Ankit Ranjan)

• Resolved an issue where the ADOC UI inserted extraneous whitespace characters into calculated SUM aggregate values displayed in the Sample Data preview for Custom SQL rules.

• Resolved an issue where the dbt Cloud lineage graph failed to trace dependencies through ephemeral dbt models, resulting in disconnected upstream source assets from downstream models and snapshots.

• Resolved an issue where incremental Data Quality policies running on Databricks Pushdown engine failed with a NullPointerException when good/bad record persistence was enabled and no new incremental data was detected.

• Resolved an issue where timestamp-based incremental Kafka reconciliation jobs evaluated the latest partition offsets at execution time rather than the job-start boundary, leading to window misalignment across streaming sources and false outlier alerts.

• Resolved an issue where Oracle profiling jobs failed entirely upon encountering unsupported large object data types (CLOB/NCLOB) instead of skipping unsupported columns and completing profiling for remaining fields.

• Resolved an issue where dbt Cloud pipeline views rendered an empty canvas without nodes, events, or lineage traces for specific multi-task DAGs.

• Resolved an issue where Reconciliation policy imports swapped left (source) and right (target) asset column mappings during cross-environment migration validation.

• Resolved an issue where the "Treat Zero Rows as Success" configuration flag on Data Quality policies was dropped and failed to persist across environment migrations.

• Resolved an issue where the Job Schedule API response returned inconsistent and mismatched values for the schedule Name attribute across scheduled policy executions.

• Resolved a regression where the Kafka Reconciliation execution details page failed to display partition offset ranges and execution boundaries in the UI.

• Resolved an issue where Create Table As Select (CTAS) operations did not establish lineage connections between source and target tables in OpenLineage pipeline graphs.

• Resolved an issue where filtering policies by data source on the Policy Listing page resulted in gateway timeouts and loading failures on large tenants with 80,000+ configured policies.

• Resolved an issue where crawler executions remained indefinitely stuck in the RUNNING state when a Spark driver pod terminated unexpectedly without triggering pod failure detection.

• Resolved an issue where crawler job state was not transitioned to failed when the crawler container pod was terminated unexpectedly, leaving the job in RUNNING status until the subsequent scheduled crawl.

• Resolved an issue where crawling large Excel (.xlsx) files on Amazon S3 failed with OutOfMemoryError due to unbounded heap memory allocation during DOM-based Apache POI workbook loading.

• Resolved an issue where differences in precision and scale representation for NUMERIC and DECIMAL data types between Teradata and ADOC caused calculation mismatches in reconciliation and profile checks.

• Resolved an issue where creating SQL Views containing CTE WITH clauses against Hive tables failed with a TABLE_OR_VIEW_NOT_FOUND error during query planning.

• Resolved an issue where OpenLineage parent pipeline runs were incorrectly marked as SUCCESS even when child tasks or sub-jobs emitted explicit FAIL events.

• Resolved an issue where Data Product overall reliability scores did not match the calculated average of underlying member assets due to differing aggregation and scoring methodologies applied across overview and detail pages.

• Resolved an issue where long-running Databricks JDBC pushdown policy executions failed upon token expiration due to missing automatic OAuth token refresh handling.

• Resolved a security vulnerability where users with view-only permissions on Account Configuration could view sensitive SCIM bearer tokens in the UI.

• Resolved an issue where Reconciliation job execution status remained displayed as Running in the UI even though the underlying data plane job had completed successfully.

• Resolved an issue where lineage links between Airflow tasks executing DbtCloudRunJobOperator and dbt Cloud pipelines were missing when using dbt Cloud Data Source along with Airflow OpenLineage integration.

• Resolved an issue where Data Reliability scores displayed inconsistent values between the Asset Overview summary and the Policies tab for the same asset.

• Resolved an issue where Spark jobs deployed on a Data Plane namespace with Kubernetes ResourceQuota failed immediately with an error rather than remaining queued until cluster compute capacity became available.