Determine who you are going to assess, whether it is a sample or the whole population. Also, consider whether or not to follow a particular population over time (cohort).
There are two types of samples: "simple" and "heterogeneous." A simple sample assumes that each member of the population has an equal chance of being selected, therefore, all are treated equally, such as taking a random sample of all students. This is useful for studies with hypotheses that do not require the data to be more granular. In contrast, a heterogeneous sample splits the population into different groups but asks the same question, such as sampling undergraduates and graduate students so their responses can be compared, or testing students enrolled in a gateway course and capstone course within a major to test for differences in skills, attitudes toward libraries, etc.
To do a survey of a large population (>/=500 persons), where you know its actual size (e.g., number of freshman students enrolled in fall 2012) you only need a sample to produce generalizable data that you can apply to that entire population (hypothesis testing). Use a sample size calculator or chart such as that provided by SurveyMonkey to determine how many emails you need gathered for the sample to produce generalizable data. It is also a good idea to over sample to allow for bounced emails. There are many such freebies on the web. Google it to find others.
Assessment or evaluative research studies can also look at a whole population, e.g., a group of students in one class, chemistry faculty, etc. Therefore, you would need to survey the whole population (unless it happens to be >/=500 persons, in which case, sampling is recommended). For the goals of program assessment, this is enough to generate usable data if you only care to report on a single group. For some journals, this may be enough as well. If publication is intended, it is best to read through the assessment related articles in targeted journal(s) to see what types of populations are assessed.
A cohort is a single group that is followed from a fixed point in time until another fixed point in time. For example, students enrolled in a program who share a common course of study. This can involve sampling, such as asking a random sample of freshman students to take a survey or test at the beginning of their studies, then ask the same group again as sophomores, juniors, and graduating seniors, e.g. It can also involve a whole population, such as Bridge or Stretch cohorts that already exist. Other uses include using embedded assessment techniques to track student work after provision of tiered library instruction to test for increasing sophistication in resource choice, etc. In any case, cohort populations require participation from faculty, and attrition must always be accounted for due to lack of student persistence or programmatic changes.