What types of testing should be performed? What types of test data should be used?

What will be an ideal response?

An effective test plan must include unit, integration, and system testing. The first step is unit testing, which involves a test of each data entry program or function using both valid and invalid data, as well as a test of each report generated by the system.
The next step is integration testing. Programs or functions that create or modify files used by other programs must be link tested with those other programs. In a database environment, this would involve testing all relationships and multi-table queries. Finally, all programs must be tested together in a system test, along with all system utility functions, including backup and recovery procedures.
Users want to be sure that the newly developed information systems can handle their typical workload, both in terms of volume of work and types of normal business transactions. This is why analysts typically use live data during system testing. It also is true that creating good system test data takes time and effort, and there usually is considerable pressure at this point to get the system finished. In fact, live data is not always better than simulated data, because a limited sample of live data might not represent all possible data situations.
The best answer often is a combination of live and test data. If there are situations that might not arise during system testing with live data, test data often can be used effectively to simulate these conditions.