How Do You Know Whether Your Trading Edge Is Large Enough to Matter?

A strategy may show a small historical advantage that is difficult to distinguish from randomness or execution costs. What evidence would you want before concluding that the advantage is meaningful enough to justify continued use?