Clusters

A Cluster is a system entity that virtually groups one or more records based on the Matching Rules defined for a given entity. Unlike standard relationships, the connection between records within a cluster exists on a virtual level – no physical link is created between them at this stage.

Clusters are used as part of the duplicate detection workflow. When duplicate search is enabled for an entity – within the master data, within the staging data, or between master and staging – the system automatically groups the detected duplicates into clusters to facilitate review and processing.

A correctly formed cluster contains one master record and one or more staging records. Once the user confirms a cluster item, a real relationship between the master and staging records is established.

Creating and Deleting Clusters

Clusters are created automatically by a scheduled job based on the Matching Rules configured for the entity. Users cannot create clusters manually, nor can they add or remove cluster items manually.

Regardless of the cluster's current state, items may be added to or removed from it on each subsequent job run.

A cluster can only be deleted if it has no cluster items (i.e., its state is Empty).

Cluster Fields

The State field of the Cluster can have the following values:

• Empty – the cluster has no cluster items.
• Review – the cluster has items, but is not yet fully merged.
• Merged – all cluster items have been confirmed.
• Invalid – the cluster contains more than one master record, or contains no staging records.

Cluster Items

Cluster Items panel lists all master and staging records that the system has linked to the current cluster based on the entity's Matching Rules. Each item displays the Entity Name, the linked Record, and its Matched Score.

Cluster items cannot be added or removed manually. If a Matched Record is deleted, the corresponding Cluster Item is not removed – instead, it is reassigned to another cluster (existing or newly created).

Every staging record must belong to a cluster. If the system detects no duplicates for a given record, a cluster with a single item is created and that item is confirmed automatically.

Cluster items are created, populated, and automatically confirmed by the Create Clusters scheduled job. If matching rules change, or a record is created or updated causing a Matched Records pair to be added, removed, or have its Matched Score changed, the cluster is not updated immediately – changes take effect on the next job run.

Cluster View Modes

The Cluster entity supports two view modes:

• Standard View – displays all cluster fields along with two relation panels: Cluster Items and Rejected Cluster Items.

• Comparison View – displays a comparison table with the fields and attributes defined in the Selection Layout of the corresponding entity.

Field values that differ between records are highlighted with a vertical red line. The comparison table shows a maximum of 5 items at a time.

Each column in the comparison table includes a Matched Score row for the corresponding item. Field values in the table can be edited directly.

The Golden Record is indicated by a horizontal gold line in the comparison table. A confirmed staging record is indicated by a horizontal blue line.

To control which items are shown in the comparison table, open the Item tab in the left sidebar. It lists all items belonging to the cluster, grouped by entity. Use the eye icon next to each item to show or hide it in the table.

Cluster Item Actions

The following actions are available for cluster items in both Standard and Comparison views:

• Confirm – creates a link between the staging and master records. Two scenarios apply:

  • If a master record already exists in the cluster, the staging record is linked to it.
  • If no master record exists yet, a new master record is created based on the selected staging record and linked to it.

  An already confirmed item cannot be confirmed again.

• Reject – removes the cluster item from the current cluster and reassigns it to another cluster (existing or newly created). The rejected item remains visible in the Rejected Cluster Items panel. Confirmed items can be rejected as well – in this case, the link between the master and staging records is removed.

• Unreject – Available for items in the Rejected Cluster Items panel. Returns the item to the cluster. If the item was previously confirmed, then rejected, and then returned to the cluster, it will have an unconfirmed status and must be confirmed again.

• Delete Deletes the corresponding record and its cluster item.

Confirming Cluster Items

When a cluster item (master or staging record) is confirmed, the system creates a Golden Record for the cluster if one does not yet exist. If a Golden Record is already set, confirming additional items links the corresponding staging records to that master record. The linked staging records become visible in the relevant relation panel on the master record's detail page.

Merging Script

During confirmation, all cluster items are processed through the Merging Script defined in the Master Data Entities configuration for the corresponding staging entity.

This script describes the logic by which the master record is created or updated based on the staging record – specifying which fields should be copied, under what conditions, and whether any field transformations should be applied.

Automatic Confirmation

Cluster items can be confirmed manually by the user or automatically by the Create Clusters scheduled job. To enable automatic confirmation, check the Confirm Automatically checkbox in the configuration of the corresponding staging entity under Master Data Entities. When enabled, the Minimum Matching Score field becomes required and defines the threshold for automatic confirmation.

A cluster item is confirmed automatically if its Matched Score is greater than or equal to the Minimum Matching Score. If the score falls below this threshold, the item will not be confirmed automatically, even if the setting is enabled for the entity.

Items confirmed automatically are marked with a dedicated icon in the cluster.

Note: The Matched Score is copied from the corresponding Matched Record to the cluster item. If the matching chain is A → B with a score of 80 and B → C with a score of 60, the scores assigned to the cluster items will be: A – 80, B – 80, C – 60.

Automatic Master Update

When Update Master Automatically is enabled in the Master Data Entities configuration, any update to a staging record will automatically trigger an update of the corresponding master record according to the Merging Script – without requiring a manual re-confirmation.

If automatic field locking on manual modification is enabled, updates from staging records will not overwrite fields in the master record that have been modified manually.