The Reference Matcher process compares one entity with another entity, while the Relationship Linker process is used to compare records in the same entity. The Reference Matcher is mainly used to update the reference (master) entity file with new records.
For example, suppose after running the initial matching, you received a new set of records. In this case, take the new records as your input (transaction) entity and the initial matched records as your reference (master) entity. You can compare the input entity with the reference entity and verify the existence of new records in the reference entity.
Before configuring and running the Reference Matcher process, you need the following preparation:
- Create window keys using the Window Key Generator. The window keys must be produced and included in the input and reference entities.
- Sort rows by the window key using the Sort Utility. After creating the window keys, but before running the Reference Matcher, the input record must be sorted by the window key.
Reference Matcher Process
The Reference Matcher completes its process in six major steps:
Step 1: Create match windows
The Reference Matcher uses the window keys to generate the match window for both the input and reference entities.
Step 2: Compare records
The Reference Matcher processes each record in the reference entity match window one at a time and compares them to each record in the input entity match window. This process works in the same way as the Relationship Linker.
The matching criteria is built based on matching attributes, comparison routine, score, grades, and pattern. See Determining Matches for details. Based on the matching pattern returned by the routines, the result of the comparisons is considered a Pass (matched) , a Suspect (suspect match), or a Fail (unmatched). Matching is evaluated based on the type of matching. See Specifying Reference Matcher Settings.
Step 3: Store match information
If a match is determined, a reference match key is copied from the reference record to the matching input record. If no reference match key is found on the matched reference record, a new key is generated.
The match information is stored based on the match type, Pass or Suspect, and match level, Level 1 or Level 2. The matches are maintained as an ordered list based on the value of the pattern ID with the match with the lowest pattern ID being at the head of the list. In the event of multiple matches with the same lowest ID, the first match encountered will be the first in the list.
Step 4: Copy attributes
If configured, data for selected attributes can be copied between the matched input and referenced records. See Specifying Attributes to Copy.
Step 5: Run window matching
After all records are passed through the reference matching process, if configured, normal relationship linking begins on the input records. This process attempts to match all records within a window to all other records within the same window. You can turn on or off this process (the default is on). See Specifying Reference Options.
Step 6: Generate results
When the Reference Matcher process completes, it produces an output entity with reference match information added to each output record, and also produces a statistics file summarizing the results of the matching process.
The standard output from Reference Matching contains records from the input entity with reference match keys and other reference matching information. The number of output records is the same as in the input records.
You can specify an optional second output entity which includes records from the reference entity that had a matching record in the input entity. See Specifying Reference Output.
In the end, the user can use the reference match keys to update the reference entity. See Viewing Reference Matcher Results.
Click any of the following topics to learn about how to set up and run the Reference Matcher process.
- Sample Reference Matching
- Determining Matches
- Setting Up a Reference Matcher Process
- Specifying Reference Matcher Settings
- Running the Reference Matcher Process
- Viewing Reference Matcher Results
- Tuning the Linking Rules
By default, Name and Address projects include a Relationship Linker process. If your business goals require reference matching, you can convert Relationship Linkers to Reference Matchers. You can also convert Reference Matchers to Relationship Linkers.