Bookshelf Home | Contents | Index | Search | PDF |
Siebel Data Quality Administration Guide > Siebel Data Quality Matching Server >
Siebel Data Quality Matching Server Functionality
The Siebel Data Quality Matching Server provides the following functionality:
- Key generation. The Siebel Data Quality Matching Server uses multiple keys for each record to detect duplicate records. Keys are generated for each customer record based on a set of deduplication input fields on the customer record for prospects and contacts, or account name for accounts (typically a combination of personal name, company name, address, and identifier information). Later in the Search functionality, the matching server detects possible matches by comparing these stored keys against the key for the active record at run time. Keys are generated based on a person's name (first name, middle name, last name) for prospects and contacts or the account name for accounts. If no keys are generated for a certain record, that record is ignored as a potential candidate when search and match takes place.
NOTE: Usually you need to run batch-mode key generation before you run real-time data matching. The Siebel Data Quality Matching Server requires generated keys in the key tables first before you can run real-time data matching. The Siebel Data Quality Universal Connector also has a similar requirement, but the key generation is done within the deduplication task. For more information, see Batch Mode.
- Key refresh. Key refresh is a server component operation type specified by the administrator. Because key data can become out of sync with the base tables, you need to refresh periodically. A key refresh updates keys only for records that are new or have been modified since your last key generation. Key refresh is a useful administrative tool because it is much faster than key generation.
- Search. The Search functionality specifies what ranges of possible keys should be considered for a given record when performing a match.
- Match. A match score is computed for each candidate record. The match score is a number that depicts the similarity of a record with a matched one. It is calculated taking into account a large number of rules along with a number of other factors and weightings.
The following is the matching process used by the Siebel Data Quality Matching Server:
- Keys are generated for the existing customer records in the database.
Typically keys are generated and refreshed on a periodic basis by the data administrator. In addition, if real-time deduplication is enabled for end users, keys are also automatically generated for a customer record whenever a user inserts or modifies an existing record. For more information about key generation, see Batch Mode.
- When a user enters or modifies a record or the administrator submits a batch deduplication request, the Siebel Data Quality Matching Server identifies candidate matches for each record by locating existing records whose corresponding keys fall within a range around the master record. Like the keys, these ranges are based on a person's name (first name, middle name, last name) for prospects and contacts, and account name for accounts.
- A match score is computed for each candidate record.
The match score is a combination of a large number of rules that compensate for how frequently a given name or word appears in the real world. The rules then weigh the similarity of each field on the record according to the real-world frequency of the name. For example, Smith is a common last name, so a match on a last name of Smith would carry less weight than a match on a rare last name. For more information about matching rules, see Matching Rule Modification.
- The match scores are returned by the matching routine.
Any existing records in which match scores exceed the threshold specified in the Data Quality Settings are considered as matches. For more information about match threshold and Data Quality Settings, see Applying Siebel Data Quality Settings.
For more information about the search functionality and matching process, see Match Process for Siebel Data Quality Matching Server.
Bookshelf Home | Contents | Index | Search | PDF |
Siebel Data Quality Administration Guide Published: 15 May 2003 |