Overview
This article details the new features and improvements that are included in the dentsu.Composable Match 1.2.4 release.
Improvements and updates
Per-job billing events emitted to Snowflake telemetry
The Match App now emits per-job billing events to the Snowflake event table using snowflake.telemetry.add_event("billingEvent", …). Each job generates a ROWS_IN event at acceptance, capturing the input row count, and a ROWS_OUT event at finalization, capturing the total output along with individual, household, and address match breakdowns.
Each event includes the following fields: requestId, dentsuClientKey, rowsToProcess, rowsReturned, rowsMatchedTotal, rowsMatchedIndividual, rowsMatchedHousehold, rowsMatchedAddress, and the consumer account organization name and locator.
Self-service recovery for failed or stuck jobs
You can now diagnose and recover jobs that hang or fail, most commonly due to stale streams after a failed file reconstruction or issues with the build cache, without requiring engineering intervention. Two recovery procedures are available in APP_PUBLIC and are granted to APP_USER_ROLE, enabling you to identify root causes and resume processing directly within the application.
Request status surfaces failures
REQUEST_STATUS now transitions to a terminal failure state when errors occur, instead of remaining in ACCEPTED. This ensures you can clearly identify when a job has failed. Failure scenarios include DQ-prep errors and sub-procedure exceptions, with an associated error message captured for each failed request.
Change tracking enabled on request status view
You can now create a Snowflake Stream on the APP_PUBLIC.REQUEST_STATUS view to react to job status changes in real time, such as triggering downstream processing when a job completes. Streams created by you persist across Match application upgrades.
Runtime parameters captured in request status for history
REQUEST_STATUS now persists job-level runtime parameters, including output ID type, NCOA flag, and audience configuration. This allows you to review the settings used for a job without parsing logs. Historical records are backfilled when upgrading to version 1.2.2.
Container image security update alerts
You are now notified when the container image falls behind security baselines through a new notify_update procedure, allowing you to identify and address security update gaps.
DQ handling for # character in PAF customer names
DQ processing no longer fails when a # character appears in PAF customer name fields. Escaping is corrected to ensure jobs complete successfully.
Stream offset reset and transient table collision fixes
An issue where app upgrades left streams pointing to stale offsets is resolved, ensuring streams resume correctly after upgrade. An additional fix addresses a transient table name collision that could occur during concurrent runs.
We value your feedback
To provide feedback on this release, submit a request.