In SPC terms, a population is every output of the process being monitored (e.g. every single bottle cap produced by a process throughout its working life).
A sample is, therefore, a small selection of the process's outputs, drawn at random from that population, from which the probability of future outcomes can be predicted.
When we wish to make predictions about a population based on a sample, the actual number of items in the sample should be a rational one, i.e. it should be of a size sufficient to give a result which can be relied upon.
Where it is possible to obtain measured data from the process, we can make esitmates of probable future outcomes with reasonable confidence once we have sampled around 125 individual items. These items are usually measured in subgroups of 4 or 5 items at regular intervals in the process cycle.
This measured data (which could be length, weight, diameter, temperature, duration, etc.) is known as variable data.