To ensure that all of the major groups of customers are covered in the sample, you could use a stratified random sample. In selecting a stratified random sample you first identify the different categories of your population that you want to be sure and have represented in the sample you survey. For example, you might want to make sure different geographic regions are represented, or alternatively you might want to ensure each of the different groups of your customers by size are represented. In setting up a stratified random sample — break your population list out by the groupings you want to be sure and include. Then, one group at a time, draw a random sample. This will ensure you have representation from each of the groups that you desire.

How to draw a random sample — the idea here is that you want to draw a sample from your total population so each individual has the same chance of being selected to be in the sample. One way you can do this is to put your total list into a spreadsheet with each individual being in a new row. Next, insert a column (we like to have it on the far left-hand side) labeled “Number.” Start with the number No. 1 for the first individual and number each of the individuals sequentially. Now suppose that you have 500 individuals in this spreadsheet and you want to randomly draw 25 of these names to send the survey to. Next, go to a random number generator (http://www.randomizer.org/form.htm) and submit a request for 25 random numbers between 1 and 500 with no repeats. Then take the numbers that are generated and the individuals with those numbers are the ones that you will survey.

Finally we note a commonly used sampling approach — and that is the convenience sample. Just as the name would suggest, this is when you sample according to what is convenient. If you wanted to sample from a broad list of potential and new customers, you might have a list from an industry trade organization that included many, but not all of your population group. Using this industry trade organization list would be an example of a convenience sample.

So why would you want to use a random sample vs. a convenience sample, vs. a survey of the entire population? A random sample gives you a truer reading in a statistical sense of what is happening in the population you are sampling. A convenience sample, while not as accurate in a statistical sense — is easier, quicker and cheaper to perform — and the same is true for a sample compared to surveying the entire population.

Another issue associated with how you select your sample is how you will communicate with the individuals you are surveying. You will need email or mail addresses if you plan to use electronic or mail surveys. If you plan to do a phone survey, then you will obviously need phone numbers.

Even with the very best lists of individuals, you need to plan on the fact that you will never get a complete set of responses and therefore you will have nonresponse bias in your survey results. We will discuss this in more detail below under “Analyzing the Data from Your Survey.” Keep in mind as you work with the issues related to selecting the desired population and the most appropriate sample, what your overall need for the information is and how you will most likely get the information that is most helpful to you.

Formulating the survey questions

Writing the survey questions seems like it should be easy — yet there are a number of pitfalls you can easily fall into. There are some questions your audience may find intrusive. Examples of these include items like income and age. Asking these questions so that the respondent selects a category (or range) reduces the risk that the respondent will find this intrusive and increases the chance that he/she will answer your question. Another common approach in surveys is to keep your questions related to demographics (age, income, gender, etc.) until the end of the survey. There is a greater chance individuals will respond to these questions at the end of the survey compared to at the beginning of the survey. If you are using an online survey instrument and the respondent does “click out” of the survey when he/she reaches a “sensitive,” question you at least have the responses to the first questions. Similarly if you are conducting a phone survey, you will have responses to the early questions.

The wording of a question can and does influence the responses that you get. It is really easy to word a question so that virtually everyone will answer “yes” and then you will fail to get the information that you need. Taking the time to pretest your survey can be critical here as the pretesting is where you will learn about poorly worded questions. Doing a pretest will also help you uncover questions that people misinterpret.