It may be the case, and it is very common, that the universe is very large or difficult to delimit. For example, it is very difficult and costly, in time and money, to interview or survey each and every voter in a given constituency, because it is about millions of people in a wide geographical area, although through the population census we can know exactly how many and who they are. It can also be difficult to delimit. In these cases, sampling is used. We propose the following table to understand when a sample is needed and when it is recommended, to help decide if we choose to analyze the universe or the sample: 

Recommended sample

It's faster.

Lower costs.

It makes control easier.

When we have a limited time

On the other hand, the samples present two difficulties that must be considered: it is not easy to elaborate a good sample, especially that it is representative, and it does not always collect well the complex variability of the universe to study.