• For tests, use data that makes the tests easy to read and follow.
  • Don’t scatter data values around.
  • Differences in data should be meaningful. If there’s no conceptual difference between 1 and 2, use 1.
  • If system handles multiple inputs, tests should reflect multiple inputs. If there’s no difference between handling list of three items and list of ten items, use the one that is shorter.
  • Never use the same constant to mean more than one thing. For example, don’t call plus(1,1), use plus(2,3) instead.

Realistic data is the alternative to test data.