A Reference Matcher process runs as part of a Trillium Quality project.
To create a project
Open the Control Center, click the Develop > Projects, right-click and select Create project.
Enter a name and, optionally, a description for the project.
Select the project type.
For the input entity, select the entity containing your sample reference data.
Finish creating the project and click Now in the scheduler notification bar.
Tune the processes as you normally would to suit your business needs. The project must include the following processes, all of which are included in the standard Name and Address project template:
Note: Postal Matchers for China (CN), Japan (JP), Korea (KR), and Taiwan (TW) do not require the Sort process.
Transformer. The Transformer assigns unique sequence numbers to each row in the entity, which is required for reference matching.
Customer Data Parser or Business Data Parser. The Customer Data Parser and Business Data Parser standardize name and/or address data according to country-specific rules and tables. The standardized data is used for matching. (Business Data Parser is used for non name and address data).
Sort for Postal Matcher. (Not required for Business Data Projects.) Input to the Postal Matcher must be sorted in geographic order.
Note: If you change the rules by which the Window Key is generated, ensure that any process required to create the key is included in your project.
Postal Matcher. (Not required for Business Data Projects.) The Postal Matcher verifies and enriches address data by matching the data to postal directories, and then populates appropriate attributes.
Transformer Data Reconstruction. The Transformer Data Reconstruction creates a common set of derived attributes, and formats data for the batch to real-time implementation and the Global Postal Matcher.
Window Key Generator. The Window Key Generator creates keys that are used to identify groups of records that conform to a set of rules.
Note: The Trillium Quality Name and Address Project template includes the Commonizer and Transformer Label Lines or Transformer Unified Output processes, which are not required for a Reference Match project. Unless they are required for your customized workflow, you can delete them.
Sort for Linking. For the Relationship Linker to compare records in the reference entity, the rows must be organized in ascending order by window key, which is the default sort order.
Relationship Linker. The Relationship Linker uses comparison routines to determine the level of similarity between rows in an entity. For efficiency, it only attempts to match records that have the same window key.
If you will be deploying your project for real-time use, you must add the MATCHER_USER_KEY attribute to the output DDL of the Relationship Linker process. The Matching service uses MATCHER_USER_KEY to uniquely identify records during the match process. Create the attribute with a data type of NOTRANS. The size must be big enough to hold a unique value for each record you will send to the matcher in one call. For example, if you will be sending 1000 records at one time, the MATCHER_USER_KEY attribute must be a minimum of four bytes long.
When you have finished tuning the project to suit your needs, run the project and ensure that it runs successfully.