How Much Testing Is Enough?

This post originally appeared on the Software Carpentry website.

When we teach testing, we are sometimes asked, "How much is enough?" The short answer is, "It depends." To help construct a longer answer, we'd like to find out how much testing scientists do in other contexts. For example, the Cosmic Origins Spectrograph (COS) and the Space Telescope Imaging Spectrograph (STIS) are two of the instruments on the Hubble Space Telescope. During Cycle 21 (which starts this month, and runs for about a year), they will spend roughly 25% and 67% of their time respectively on calibration:

What are your figures? How much of the time in your lab/experiment/life is spent on calibration, testing, validation, and other tasks, compared to how much you spend on actual experiments?