What is the best way of choosing a sample to statistically represent a population?

You want to select a sample that is completely random. You don't want to select all types of data that are the same. For example, often in a survey, the person conducting the survey will select people in a random order. For example, maybe every 10th person in a given order.