Surveys on the other hand may provide information on both stocks and flows. Unlike Censuses, surveys involve taking a sample of the population under investigation, rather than visiting or contacting every household. The advantage of sampling is that it reduces costs (think how much cheaper it is to visit 3,000 people than it is to visit 300 millions). However, its main drawback is that it introduces sampling error. We cannot be sure of the values we obtain through sampling: if we were to draw multiple samples of the same population, each sample would yield slightly different results. There is some margin of error, but through statistical techniques, we can quantify this uncertainty (confidence intervals, cf. your biostats class).
Surveys come in various forms. Some surveys are cross-sectional, i.e., they only attempt to interview a sample of respondents once. Such cross-sectional surveys generally provide only data on stocks. Other surveys however are longitudinal, i.e., that they repeatedly re-interview the same respondents and aim to assess change. Such surveys may provide valuable information on flows. For example, a longitudinal survey may enroll a sample of newborns and aim to assess their health repeatedly over the first 5 years of life. Such a survey would provide valuable information on an important flow: the mortality of children.