One of the greatest myths regarding name search systems is that they are successful simply because they find what was expected or is known to be on file.

To truly measure the success of a name search, one also needs to have an understanding of what matches have been missed. In many organizations, missed matches are only discovered once they adversely affect the business, operation or system. While this is often too late from a business viewpoint, such discoveries are useful input for improving the name search process.

Missed name matches can also be discovered from within the organization’s data by finding existing duplicates based on attributes other than name (for example, address and date of birth), or by exhaustively running a background matching process that uses less of the name data in its keys. To be useful for tuning the name search, this requires expert users to review the missed matches now found and help establish rules to avoid missing these matches in future.

Whatever the method of discovering matches that otherwise would not have been found, the goal should be to create and maintain a set of model answers, based on both real data and expert user input, as a benchmark for the reliability of the name search process.

It is not enough for a user to test only with the difficult cases not found by the old system. Tests should be carried out on more common names to ensure the search finds them as well and does not return too many.

The Match Level should be set to Loose during testing to assist the discovery of matches which otherwise would be missed.

A batch test of an online customer name search which uses as search criteria a file of new business transactions, or even the customer file itself, provides a valuable report for users to evaluate the reliability of the search.

Because the system resource usage of the name search transaction is higher than most business transactions, it is vital that the expected volume and concurrency of searches be factored into any capacity planning.

When it is critical to a business or system to absolutely avoid missing data, then it is critical to implement procedures and processes to discover real world cases and examples of what can be missed. Only then can systems be improved.