The growth of online survey research over about the last 15 years has been truly remarkable. While estimates vary, most agree that online now accounts for somewhere between one third and one half of the total volume of survey data collection being done by US market research firms (See, for example, ESOMAR, 2009). Key to this growth has been the positioning of online as a faster and less expensive method capable of delivering the same or better survey results as traditional methods (telephone, mail, and in-person). Indeed, online often is the least expensive and fastest survey method for studying

a wide range of business problems, although this speed advantage is clearest when samples sizes are large and/or expected incidence is low. It is less certain that the survey results achieved are comparable to other methods, especially in those cases where one of the objectives is to accurately measure some characteristic, attitude, or behavior in a target population, whether it be a customer base, individuals with certain characteristics, or the general population.

This issue of accuracy is the main focus in this paper. In it we offer a framework for evaluating an online design versus a telephone design when a central goal of the research is to produce an accurate estimate of some personal characteristic,

attitude or behavior in a target population. In this context, accuracy means a survey result that is as close as possible to the “true value” in that population. We recognize that not all research has as its goal the development of accurate estimates of population values.

For example, in some instances directional measures may be sufficient. In others we might only want to test hypotheses about the interrelationships among characteristics, attitudes, and buying behavior. Under these and similar circumstances we might well use different criteria than described below to select a data collection method.