User Community Service Desk Downloads

Working with Matching

This article covers the practical aspects of using matching in ONE MDM, including the different approaches available and how to manage matching decisions over time.

How the matching process works

The matching process operates consistently regardless of how data enters the system. It works for both batch operations (full system loads, single system loads, or incremental updates) and real-time scenarios (record creation via MDM online APIs, streaming data ingestion, or identify operations).

When new or updated records enter the system, they are compared against existing data using your configured matching rules. When potential matches are identified, the system evaluates the quality and confidence of these matches.

After duplicates are detected in data, each record in the group is assigned specific metadata (for example, the name of the matching rule that identified it). Records that meet your matching criteria are grouped together and assigned a unique identifier (master ID). This identifier is then used in the merging process to combine all relevant information from the matched records into a single, accurate golden record, also referred to as master record.

As part of the matching process, related or child entities are deduplicated. Deduplication ensures that related silver records (such as multiple addresses, phone numbers, or email addresses for the same customer) remain unique within the context of their parent entities. This process assumes a one-to-many relationship where each parent entity can have multiple related child entities.

Matching primarily operates at the instance layer, working with individual records. However, the same logic can be applied to master data in centralized or mixed MDM styles. See Matching on master entities.

Matching approaches

In ONE MDM, matching can be performed in three ways:

Matching proposals provide an intermediate option between the two other methods. They are generated automatically but resolved manually.

These approaches complement each other and are used in combination, allowing you to handle different data scenarios effectively.

Automated matching

Matching and merging are automatically performed in ONE MDM based on predefined rules set in the matching configuration (that is, the configuration of the Matching step within the matching plan).

In cases where automated matching runs into ambiguities or constraints, and in the case where some special proposal rules are defined in the matching configuration, matching proposals are generated instead and need to be resolved manually. Even when proposals are generated, the instance record is still automatically processed (a master_id is assigned and the golden record is created), ensuring data remains accessible while awaiting manual resolution.

This behavior can be customized through solution-specific configuration of the model and filtering between instance and master layers.

You’ll rely more heavily on automated matching when:

  • Your data is relatively clean and consistent.

  • You have well-defined matching rules.

  • The cost of occasional false matches is low.

For example, automated matching is commonly used if you’re dealing with similar data:

  • Customer records from well-structured CRM systems.

  • Product catalogs with standardized naming conventions.

  • Financial transactions with consistent formatting.

Manual matching

How effective automated matching is depends on the accuracy of the rule configuration and the quality of the processed data. In situations where records are incomplete or automated matching does not produce the desired results, it is possible to browse, compare, and merge records manually.

You can also create manual matching tasks for your team members, and perform split and rematch operations to refine matching outcomes.

During the manual matching process, the MDM Engine offers a merge preview, where you can see how the golden record would appear after the matching is performed. You can modify the result by selecting or entering new values (similar to the editing process).

As mentioned in the previous section, matching proposals, which are generated from automated matching, are also reviewed and resolved manually.

Manual matching is necessary for high-stakes decisions or complex cases but is limited to smaller volumes due to the individual review requirement. You’ll need more manual oversight when:

  • Data quality varies significantly across sources.

  • Business rules are complex and require human judgment.

  • The cost of false matches is very high (for example, financial or healthcare data).

Typically, it is used for situations such as these:

  • Legacy system migrations with inconsistent data formats.

  • Customer records requiring business context (for example, same name, different companies).

  • High-value entities where accuracy is critical.

For step-by-step instructions, see Merging and Splitting Records.

Matching proposals

Matching proposals represent a hybrid approach between automated and manual matching processes. They are generated as a result of automated matching and suggest potential duplicates but require manual resolution to confirm, reject, or adjust the suggested matches.

This resolution process is identical to the manual matching workflow. Additionally, you can search for related matching proposals and resolve them in bulk.

Matching proposals are generated when:

  • There are specific proposal rules defined in the matching configuration.

  • A record on MDM input matches several existing groups but cannot be automatically merged to more than one record.

  • There are some constraints applied that prevent the application from correctly assigning records to individual master records within a matching group.

  • AI Matching was used.

    This is an experimental feature. To learn more, see AI Matching.

For details, see Matching Proposals.

Currently, it is not possible to generate matching proposals for already processed data when records are updated.

Instead, you can reprocess the data with rematching enabled. Keep in mind your manual match decisions will not be preserved in this case.

How to manage identities over time

Once matching decisions are made, you might need to adjust them as business requirements change or data quality improves. ONE MDM provides several ways to modify matching decisions.

Identity change triggers

Changes to record identities can be initiated using the following methods:

  • Manual match and matching proposals (preferred): The recommended approach for identity changes involves manual review and decision-making through the matching proposals workflow. This is particularly important for validating complex matching scenarios.

    This is done through the web application, see Merging and Splitting Records and Matching Proposals.
  • Rematch functionality: Allows for re-evaluation of existing matching decisions. Typically used after updating matching rules or when data quality improvements require a fresh matching analysis.

As a best practice, focus on identifying suspicious or uncertain matching cases, such as significant data change events or potentially problematic matches, rather than attempting to automate their resolution. Automatic resolution of such scenarios typically produces unpredictable results and ultimately requires manual intervention, making the initial automation attempt counterproductive.

Rematching

Rematching allows you to reassess (and potentially alter) existing matching decisions when rules change or data quality improves. It can be initiated through various mechanisms within the MDM system:

Currently, all types of rematch operations except Rematch If Changed configuration override all previous manual interventions, such as manual matches, splits, and resolved matching proposals.

Consider avoiding those operations in the scenarios where manual changes are of great importance.

Full or partial reprocess batch operation

This is a specialized type of load that reprocesses all or selected records in MDM storage by sending them to the MD consolidation process.

When rematch is enabled, all rematch flag columns are set to true, causing the records to be both reprocessed and rematched. Currently, the rematch operation overrides manual matching decisions.

For details, see Batch Interface.

Rematch If Changed configuration

This is an advanced matching configuration in the model that automatically triggers a rematch if specific columns are altered. These columns are listed in the Rematch If Changed section in the entity Advanced Matching Configuration.

Only the records satisfying the condition are rematched while the rest of the matching group remains untouched (including identity). This is used by default for foreign keys changes.

If the ID of a parent master record changes (for example, two master parties were matched and merged manually), child records are rematched (automatically) and all related record details (such as contacts) are then added to the resulting parent entity (party) master ID.

By default, this configuration does not override manual matching decisions. However, you can turn on this option by setting the nme.match.rematchIf.removeManualMatch runtime parameter, see Runtime Parameters > Process control.

Rematch using MDM core service (processMatch)

The ProcessMatchService allows external applications to modify record matching.

For example, the service rematch function reruns specified records through matching plans while skipping records with eng_isolate_flag set to true. See processMatchService.

Currently, the rematch operation overrides manual matching decisions.

Rematch action in MDM Web Application

In some cases, it might be necessary to perform a match operation on the instance record that was previously matched. You can do this from the MDM Web App.

To rematch a record, navigate to the relevant instance record, and select Actions > Rematch. For details, see editing-master-data:merging-and-splitting-records.adoc#rematch-operation.

Advanced matching scenarios

Certain matching scenarios require specialized handling, particularly when dealing with inactive records or implementing centralized and mixed MDM styles.

Matching inactive records

By default, inactive records are included in the matching process and treated the same as active records. However, in the mastering phase, inactive records are often filtered out and not used in the computation of golden records.

If the source deletion strategy is set to delete, inactive records are removed from the MDM storage entirely and not considered in any consolidation operations (that is, cleanse, match, merge). This ensures that only relevant and active data contributes to the golden records.

The advantage of including inactive records is maintaining continuity. For example, if a client returns after a contract termination, the inactive record retains its master ID during the inactive period. In other words, when the client returns, this ID can be reused, preserving the customer identity.

Matching on master entities

In centralized and mixed MDM implementation styles, instance records can be matched to authored master records that exist only at the master layer. This ensures that incoming instance data can be properly associated with manually created master records.

However, authored master records themselves cannot be matched to other existing records through the standard matching process. Instead, they are identified through the Identify action in the web application or using the IdentifyMaster service.

When creating new authored master records, MDM populates duplicate_id values to flag potential duplicates, but this identification process is separate from the matching workflow. A specific configuration is required on the master entity to map attributes from the master layer to the Matching step for this functionality to work.

Next steps

Now that you understand matching principles and processes:

  • How the Matching Step Works - Learn how the matching algorithm evaluates rules and makes decisions before configuring your own.

  • Matching Configuration - Determine your matching configuration path depending on your business requirements and MDM style.

  • Matching Architecture - Dive into the technical foundations to better understand your matching outcomes.

Was this page useful?