When talking about automated functional tests, the three most-used types are unit tests, integration tests, and system tests. These types of test can be compared along several axes: the time it takes to create a test, the time it takes to execute a test, and the scope that they test:

• Unit tests: Unit tests are the quickest to write, and they execute very quickly, often in less than a millisecond. They test the smallest possible scope in an application, often a single class or method. This means that, once written, it should virtually never be necessary to change a unit test. For many systems, it is more likely that a test will be deleted rather than changed.
• Integration tests: Integration tests take more time to write since they concern themselves with multiple units that have to be set up to work together. The execution of these tests should still be fast, averaging from below a second up to tens of seconds. Integration tests have a larger test scope, which means that, in return for this, they will cover a larger part of the code and are more likely to detect defects that are introduced with a change.
• System tests: System tests test a fully assembled and running application. Depending on the type of application, these are often API tests or automated UI tests. These tests take a lot of time to create since they rely on a deployed system to run and often require the setting up of an initial state in a database or another persistent store. The tests take a long time to execute, sometimes minutes per test. They are also less reliable and much more fragile than unit and integration tests. Even a minor change in an interface can cause a whole series of tests to fail. On the other hand, system tests can detect errors that both unit and integration tests cannot, since they actually test the running system.