Random sampling evolved as a time- and cost-saving process. Simple random sampling (SRS) is used in a variety of legal matters. One example is the collection of data for a random sample of class members to determine average overtime hours worked in an FLSA or state wage and hour case. Another example is a forensic accountant drawing a random sample of invoices to determine the number of fraudulent claims. Still another is a random sample of work orders used to estimate the amount of R&D expenditures to be claimed for tax purposes.
In each of these examples, a large database must be scientifically and appropriately analyzed, but the cost of examining every observation in the database is prohibitive. Discussion of sampling typically begins with SRS. In litigation, there is a preference for SRS over other sampling techniques or data analytics because it is accessible and inexpensive. The central idea is that one can analyze a small number of observations (sample) in order to make general statements about an entire population. Results of measurements generated by a sample are projected to the population.
In the litigation context, the relevant population may be a class of employees, universe of invoices, or total number of work orders. The trick is to draw a sample of observations that closely approximates the information contained in the population. Typically, the first question is determining how large a sample is needed to give a precise description of patterns in the population. For example, how many employees must be in a sample to accurately estimate total number of hours of overtime worked by the entire class?