The notions of a

null hypothesis and

alternative hypothesis are relevant in studying type I and type II errors. The null hypothesis is, basically, : "The proposed improvement is not actually an improvement.". The alternative hypothesis is : "The proposed improvement is actually an improvement.". Accepting the alternative hypothesis when it is not true is a type I error. Failing to accept the alternative hypothesis (and sticking to the null hypothesis) when it is in fact true is a type II error.

Consider the following outline of an example. Suppose it is known that 5% of all light bulbs currently produced by company X are defective. Someone has redesigned the light bulbs and claims that the new design will lead to less defectives. In order to determine whether or not this is true, the company decides to test 200 of the new light bulbs. Now 5% of 200 is 10, so the company would say that if there are less than 10 defects in the batch of 200, then the new light bulbs are better. However, these 200 are only a sample, and the test does not truly reflect the actual probability that one of the new light bulbs will be defective. So if less than 10 new light bulbs come out defective and yet there is actually greater than a 5% chance that a new light bulb will fail, the company has committed a type I error. Similarly it may turn out in the test that more than 10 light bulbs were defective and yet less than 5% of light bulbs with the new design are actually defective. Rejecting the new design in this case would be a type II error.