In a science experiment, your initial experimental design may yield uninterpretable results. So, prior to investing large amounts of grad student hours and resources into a particular experimental design (common rookie inclination), you do a pilot study. Just collect 5-7 data points.<p>Just run it and see what happens. Since you only spent a relatively small amount of time/resources on the pilot study, it's not a terrible waste if you get null/negative/inconclusive/uninterpretable results, and you could get a quick turnaround on the next iteration. If you're lucky, you might get to estimate the number of data points you need to collect based on the variability in the sample. It might tell you that this setup doesn't have a good signal to noise ratio. You may need to go back and change your design.<p>Or you might find something completely unexpected, and find yourself asking a different question altogether, that you need to change your overall direction. Your project proposal might have an elaborate tree of potential experiments that depend on the outcomes of each one before. And you find something so different you need to redesign everything, armed with this new information. That's a good thing. Trust me.<p>Entrepreneurship is like science in that experiment is the sole judge of truth.