Match entities

The match phase specifies how to combine your datasets into a unified customer profile dataset. After completing the map step in the data unification process, you're ready to match your entities. The match phase requires at least two mapped entities.

The match page consists of three sections:

  • Key metrics that summarize the number of matched records
  • Match order and rules for cross-entity matching
  • Rules for deduplication of match entities

Specify the match order

Go to Data > Unify > Match and select Set order to start the match phase.

Each match unifies two or more entities into a single, consolidated entity. At the same time, it keeps the unique customer records. For example, we selected two entities: eCommerce:eCommerceContacts as the primary entity and LoyaltyScheme:loyCustomers as second entity. The order of the entities specifies in which order the system will try to match the records.

Screenshot of the Match page in the Unify area of the data unification process.

The primary entity eCommerce:eCommerceContacts is matched with the next entity LoyaltyScheme:loyCustomers. The dataset that results from the first match step is matched with the following entity if you have more than two entities.

Important

The entity that you choose as your primary entity will serve as the basis for your unified profiles dataset. Additional entities that are selected during the match phase will be added to this entity. This doesn't mean that the unified entity will include all of the data included in this entity.

There are two considerations that can help you choose the hierarchy of your entities:

  • Choose the entity with the most complete and reliable profile data about your customers as primary entity.
  • Choose the entity that hast several attributes in common with other entities (for example, name, phone number, or email address) as primary entity.

After specifying the match order, you'll see the defined match pairs in the Matched records details section on Data > Unify > Match. The key metrics will be empty until the match process completes.

Define rules for match pairs

Match rules specify the logic by which a specific pair of entities will be matched.

The Needs rules warning next to an entity name suggests that no match rule is defined for a match pair.

Screenshot of the Matched record details section with control to add rules highlighted.

  1. Select Add rules under an entity in the Matched records details section to define match rules.

  2. In the Create rule pane, configure the conditions for the rule.

    Screenshot of an opened match rule with conditions added.

    • Entity/Field (first row): Choose a related entity and an attribute to specify a record property that is likely unique to a customer. For example, a phone number or email address. Avoid matching by activity-type attributes. For example, a purchase ID will likely find no match in other record types.

    • Entity/Field (second row): Choose an attribute that relates to the attribute of the entity specified in the first row.

    • Normalize: Select from following normalization options for the selected attributes.

      • Whitespace: Removes all spaces. Hello World becomes HelloWorld.
      • Symbols: Removes all symbols and special characters. Head&Shoulder becomes HeadShoulder.
      • Text to lower case: Converts all character to lower case. ALL CAPS and Title Case becomes all caps and title case.
      • Unicode to ASCII: Converts unicode notation to ASCII characters. /u00B2 becomes 2.
      • Numerals: Converts other numeral systems, such as Roman numerals, to Arabic numerals. VIII becomes 8.
      • Semantic types: Standardizes names, titles, phone numbers, addresses, etc.
    • Precision: Set the level of precision to apply for this condition.

      • Basic: Choose from Low, Medium, High, and Exact. Select Exact to only match records that that match 100 percent. Select one of the other levels to match records that aren't 100 percent identical.
      • Custom: Set a percentage that records need to match. The system will only match records passing this threshold.
  3. Provide a Name for the rule.

  4. Add more conditions or select Done to finalize the rule.

  5. Optionally, add more rules.

  6. Select Save to apply your changes.

Add conditions to a rule

To match entities only if attributes meet multiple conditions, add more conditions to a match rule. Conditions are connected with a logical AND operator and thus only executed if all conditions are met.

  1. Go to Data > Unify > Match and select Edit on the rule you want to add conditions to.

  2. In the Edit rule pane, select Add condition.

  3. Select Done so save the rule.

Add rules to a match pair

Match rules represent sets of conditions. To match entities by conditions based on multiple attributes, add more rules

  1. Go to Data > Unify > Match and select Add rule on the entity you want to add rules to.

  2. Follow the steps in Define rules for match pairs.

Note

The order of rules matters. The matching algorithm tries to match on the basis of your first rule and continues to the second rule only if no matches were identified with the first rule.

Change the entity order in match rules

You can reorder entities for match rules to change the order in which they are processed. Rules that are conflicting because of a changed order will be removed. You have to recreate removed rules with an updated configuration.

  1. Go to Data > Unify > Match and select Edit.

  2. In the Edit rule pane, select the Move up/down control or drag and drop entities to change the order.

    Options to change in which order entities are processed in the match phase.

  3. Select Done so save the rule.

Define deduplication on a match entity

In addition to cross-entity match rules, you can also specify deduplications rules. Deduplication is another process when matching records. It identifies duplicate records and merges them into one record. Source records get linked to the merged record with alternate IDs.

Deduplicated records will be used in the cross-entity matching process. Deduplication happens on individual entities and can be configured every entity used in match pairs.

Specifying deduplication rules isn't mandatory. If no such rules are configured, the system-defined rules are applied. They combine all records into a single record before passing the entity data to cross-entity matching for enhanced performance.

Add deduplication rules

  1. Go to Data > Unify > Match.

  2. In the Merged duplicates section, select Set entities. In case deduplication rules are already created, select Edit.

  3. In the Merge preferences pane, choose the entities you want to run deduplication on.

  4. Specify how to combine the duplicate records and choose one of three options:

    • Most filled: Identifies the record with most populated attribute fields as the winner record. It's the default merge option.
    • Most recent: Identifies the winner record based on the most recency. Requires a date or a numeric field to define the recency.
    • Least recent: Identifies the winner record based on the least recency. Requires a date or a numeric field to define the recency.

    Deduplication rules step 1.

  5. Once the entities are selected and their merge preference is set, select Add rule to define the deduplication rules at an entity level.

    • Select field lists all the available fields from that entity. Choose the field you want to check for duplicates. Choose fields that are likely unique for every single customer. For example, an email address, or the combination of name, city, and phone number.
    • Specify the normalization and precision settings.
    • Define more conditions by selecting Add condition.

    Deduplication rules step 2.

You can create multiple deduplication rules for an entity.

  1. Running the match process now groups the records based on the conditions defined in the deduplication rules. After grouping the records, the merge policy is applied to identify the winner record.

  2. This winner record is then passed on to the cross-entity matching, along with the non-winner records (for example, alternate IDs) to improve the matching quality.

  3. Any custom match rules defined overwrite deduplication rules. If a deduplication rule identifies matching records, and a custom match rule is set to never match those records, then these two records won't be matched.

  4. After running the match process, you will see the deduplication stats in the key metrics tiles.

Deduplication output as an entity

The deduplication process creates a new entity for every entity from the match pairs to identify the deduplicated records. These entities can be found along with the ConflationMatchPairs:CustomerInsights in the System section in the Entities page, with the name Deduplication_DataSource_Entity.

A deduplication output entity contains the following information:

  • IDs / Keys
    • Primary key field and its alternate IDs field. Alternate IDs field consists of all the alternate IDs identified for a record.
    • Deduplication_GroupId field shows the group or cluster identified within an entity that groups all the similar records based on the specified deduplication fields. It's used for system processing purposes. If there are no manual deduplication rules specified and system defined deduplication rules apply, you may not find this field in the deduplication output entity.
    • Deduplication_WinnerId: This field contains the winner ID from the identified groups or clusters. If the Deduplication_WinnerId is same as the Primary key value for a record, it means that the record is the winner record.
  • Fields used to define the deduplication rules.
  • Rule and Score fields to denote which of the deduplication rules got applied and the score returned by the matching algorithm.

Run the match process

With configured match rules, including cross-entity matching and deduplication rules, you can run the match process.

Go to Data > Unify > Match and select Run to start the process. The matching algorithm takes some time to complete and you can't change the configuration until it completes. To make changes, you can cancel the run. Select the status of the job and select Cancel job on the Progress details pane.

You'll find the result of a successful run, the unified customer profile entity, on the Entities page. Your unified customer entity is called Customers in the Profiles section. The first successful match run creates the unified Customer entity. All subsequent match runs expand that entity.

Tip

There are statuses for tasks and processes. Most processes depend on other upstream processes, such as data sources and data profiling refreshes. Select the status to open the Progress details pane and view the progress of the task or process. Then, select the See details link for more progress information, such as processing time, the last processing date, and any applicable errors and warnings associated with the task or process.

Review and validate your matches

Go to Data > Unify > Match to evaluate the quality of your match pairs and refine them if necessary.

The tiles on top of the page show key metrics, summarizing the number of matched records and duplicates.

Cropped screenshot of the key metrics on the Match page with numbers and details.

  • Unique source records shows the number of individual source records that were processed in last match run.
  • Matched and non-matched records highlights how many unique records remain after processing the match rules.
  • Matched records only shows the number of matches across all of your match pairs.

You can assess the results of each match pair and its rules in the Matched records details table. Compare the number of records that came from a match pair against the percentage of successfully matched records.

Review the rules of a match pair to see the percentage of successfully matched records at the rule level. Select the ellipsis (...) and then select Match preview to view all these records on the rule level. We recommend that you take a look at some records to validate that they were matched correctly.

Try different precision thresholds on conditions to find the optimal value.

  1. Select the ellipsis (...) for the rule that you want to experiment with and select Edit.

  2. Change the precisions values in the conditions you want to revise.

  3. Select Preview so see the number of matched and unmatched records for the selected condition.

Manage match rules

You can reconfigure and fine-tune most of the match parameters.

Screenshot of the dropdown menu with match rule options.

  • Change the order of your rules if you defined multiple rules. You can reorder the match rules by selecting the Move Up and Move Down options or by drag and drop.

  • Change rule conditions by selecting Edit and choose different fields.

  • Deactivate a rule to retain a match rule while excluding it from the matching process.

  • Duplicate your rules if you've defined a match rule and would like to create a similar rule with modifications, select Duplicate.

  • Delete a rule by selecting the Delete symbol.

Specify custom match conditions

You can specify conditions that override the default match logic. There are four options available:

Option Description Example
Always match Defines values that are always matched. Always match Mike and MikeR.
Never match Defines values that never match. Never match John and Jonathan.
Custom bypass Defines values that the system should always ignore in the match phase. Ignore the values 11111 and Unknown during match.
Alias mapping Defining values that the system should consider as the same value. Consider Joe to be equal to Joseph.
  1. Go to Data > Unify > Match and select Custom match in the Matched records details section.

    Screenshot of the match rules section with Custom match control highlighted.

  2. In the Custom pane, go to the Records tab.

  3. Choose the custom match option from the Custom type dropdown and select Download template. You need a separate template for each match option.

  4. A template file downloads. Open it and fill in the details. The template contains fields to specify the entity and the entity primary key values to be used in the custom match. For example, if you want primary key 12345 from Sales entity to always match with primary key 34567 from Contact entity, fill in the template:

    • Entity1: Sales
    • Entity1Key: 12345
    • Entity2: Contact
    • Entity2Key: 34567

    The same template file can specify custom match records from multiple entities.

    If you want to specify custom matching for deduplication on an entity, provide the same entity as both Entity1 and Entity2 and set the different primary key values.

  5. After adding all the overrides, save the template file.

  6. Go to Data > Data sources and ingest the template files as new entities.

  7. After uploading the files and entities are available, select the Custom match option again. You'll see options to specify the entities you want to include. Select the required entities from the dropdown menu and select Done.

    Screenshot of the dialog to choose overrides for a custom match scenario.

  8. Applying the custom match depends on the match option you want to use.

    • For Always match or Never match, proceed to the next step.
    • For Custom bypass or Alias mapping, select Edit on an existing match rule or create a new rule. In the Normalizations dropdown, choose the Custom bypass or Alias mapping option and select Done.
  9. Select Save on the Match page to apply the custom match configuration.

  10. Select Run on the Match page to start the matching process. Other specified match rules are overridden by the custom match configuration.

Known issues

  • Self-conflation doesn't show the normalized data in deduplication entities. However, it applies the normalization internally during deduplication. It's by design for all normalizations.
  • If the semantic type setting is removed in the Map phase when a match rule uses Alias mapping or Custom bypass, the normalization won't be applied. It only happens if you clear the semantic type after configuring the normalization in the match rule because the semantic type will be unknown.

Next step

After completing the match process for at least one match pair, continue to the Merge step.