Changed behavior

The match process now retains only one null match pair for each source record for a match model, irrespective of the number of Directed AI match rules that you configured. This behavior prevents inaccurate or large clusters from forming and applies to the Search Match REST API too. For example, you configure 10 Directed AI match rules with null matching enabled across two match models. A source record matches up to other two other records, creating one null match pair for each match model.

To optimize match results and avoid potential undermatching based on this change in null-to-null matching, choose to retain all match pairs. You don't need to reindex records after you choose this option.

For example, consider a list of customer records that don't contain data in some of the fields. Configure a match model with a Directed AI match rule that has null matching enabled. Select Full Name as a fuzzy match field, and Phone Number, Address, and SSN as exact match fields. Enable null matching for the SSN field.

The following table lists the sample records that don't contain values in the SSN field:

If you choose to retain the match pair with the highest score, it might cause undermatching because the match process retains only one match pair for each source record and deletes the rest. The merge process merges records 1 and 2 into one master record and records 3 and 4 into another master record.

To reduce this risk of undermatching, choose to retain all match pairs so that the match process retains all the match pairs. The merge process then merges records 1, 2, 3, and 4 into a single master record.

Previously, the match process generated one null match pair per source record for each Directed AI match rule in a match model. For example, if you configured 10 Directed AI match rules with null matching enabled across two match models, a source record could match up to 10 other records, creating 10 null match pairs.

Consider the preceding example with a match model that has three Directed AI match rules. The match process could generate up to three null match pairs for a single source record:

For more information about null matching, see Null matching.

When you map fields in a modernization project, you can now download the latest reference data and data model mapping files directly from their corresponding sections.

Previously, you could download these files from the Download Files section on the Map fields page.

For more information about mapping files, see Modernization mapping files.

MDM SaaS now limits the number of resolve long-running operations jobs that can run concurrently on a POD. When the number of concurrent jobs reaches the limit on a POD, API requests for operations might return an error with the response code 429. You can perform the operations again after some time.

Previously, MDM SaaS didn't limit resolve long-running operations jobs. The jobs could overuse resources and cause operations to fail repeatedly.

For more information about the resolve long-running operations job, see Resolve long-running operations job.

When you use the Merge API to merge a record pair that is already merged by using credentials that require approval, the API now returns a 400 error response code.

Previously, the API returned a success response for the same request, creating redundant approval tasks in MDM SaaS.