Self-administered paper and pencil surveys
Self-administered computer surveys (typically online)
Define the research question: This is critically important to the success of a survey research project. Without a clearly defined question, it is difficult to determine the best approach for conducting the survey. For example, based on the research question, are the needed data exploratory, descriptive, or causal? The answer to this basic question has huge implications for the entire research process, yet it is often not directly addressed.
Specifying the population of interest: This simply refers to determining who you want your data to represent. If you want generalizable information about your customers then your population of interest is your existing customer base. If you want generalizable information about the U.S. then the population of interest is the entire U.S. population.
Identify a sample frame: Identifying a sampling frame is the process of determining how you will reach the population of interest. If you are surveying your existing customer base then you could use frames such as mailing addresses, telephone numbers, email addresses, or other existing points of contact that you have with them.
Choose a data collection mode: The choice of data collection mode is largely driven by the sampling frame that was selected. If the sampling frame is a database of customer email addresses then the mode of data collection will typically be an online self-administered survey.
Design and pre-test questionnaires: Designing the questionnaire carefully and then pre-testing it before fielding it to your entire sample is crucial to getting data that are valid and reliable. For example, careful questionnaire design and pre-testing can help reduce the chance that respondents may interpret the meaning of questions differently. Future posts in this series will tackle these important steps in much greater detail.
Select a representative sample: Selecting a representative sample from your sampling frame is also important for collecting valid and reliable data about the population of interest. For example, if you are sampling from a large database of customer email addresses and only wanted one response per household, you might want to cross-check each email address against mailing addresses and remove duplicates to avoid some households having a greater probability of selection. Then you would likely draw a random sample from the remaining list of email addresses.
Recruit and measure the sample respondents: This simply refers to the process of sending the survey out for data collection. Here it is key to put substantial effort into getting responses from everyone in the sample, this will determine the response rate of the survey.
Code and edit the unadjusted data and the conduct post hoc data adjustments: Coding and adjusting the data is often necessary, particularly if open-ended questions were asked or if certain branched variables need to be combined. This is also when the data are often weighted to match known population parameters.