After a survey is completed, some stations will interpret the data themselves; others choose to have their data analyzed by an experienced consultant and research professional. If your station wants to analyze the survey data on its own, here are a few suggestions about how to begin.

Depending on the type of questions asked, answers to each closed-ended question should provide one of the following results:

• Frequency distributions
  (Example: 59% agree that our station sound has greatly improved in the past year, 20% feel it has gotten worse, 10% perceive no change, and 11% are unsure or have no opinion)

• Averages (medians and/or means)
  (Example: When asked "How much do you like Marketplace?" on a 1- 5 scale, the average response from listeners was 1.97)

Exactly how the results to each question can be interpreted depends entirely on what was asked in the survey. Some areas of interest to your station may be covered in only one question; others may be the subject of related multiple questions.

Once the answers to each question are calculated, make sure that the standard error value for each question is known. That is the most basic way of determining whether or not a result is statistically significant. If it is not, the results of a question are inconclusive.

For this example, let's assume that two hundred valid surveys were returned to your station, and that the question to be interpreted was a "yes/no" question: "Should WXYZ broadcast traffic reports at least once an hour on weekdays?" If the response was 56% "yes" and 44% "no," the standard error would be plus or minus 7%, and the answer would be inconclusive, because 7% subtracted from 56% equals 49%, and 7% added to 44% equals 51%. With the standard error, the true majority could either be a "yes" or a "no" response. However, if 80% responded "yes," and only 20% responded "no," the answer would be significant, since even if the standard error of 7% was subtracted from the "yes" total and added to the "no" total (equaling 73% "yes" and 27% "no"), the "yes" responses would still outnumber the "nos" by a wide margin.

While answers reflecting the overall sample of a survey are useful, it is often even more revealing to use cross-tabulation as a method to gain a wider depth of understanding of the data. Cross-tabulation (or cross-tabbing, as it is often called) is used to segment the responses of different significant groups of listeners. When the survey is tabulated, you'll have totals that represent the answers from all respondents. But the results often mean more if they are presented so that they show the opinions of different listener groups, such as members vs. nonmembers, or members whose annual gifts are above and below a specified amount. As long as there is a question or some other indication on the survey that will allow identification of these different groups, cross-tabulation is possible. There are several ways to go about this, depending on whether a mail or telephone survey is used, and on the nature of the questions asked.

For mailed surveys, there are several options that can easily be used to separate various groups of respondents. Surveys can be printed on different colors of paper, each of which is mailed to specific groups of listeners, to separate the responses into groups without having to ask a specific question on the survey. Each survey sent can contain a specific code number, so it can be separated when it is returned, or, cross-tabulation can be accomplished through responses to questions on the actual survey.

For telephone surveys, cross-tabulation is more dependent upon the questions asked. But calling can take place from two or more distinct lists, each of which reflects a certain trait of the listeners being contacted. When a survey is successfully completed, the caller can indicate which list the person was on. This information can be coded into the database software, allowing for cross-tabulation.

For example, in asking a question about the amount listeners are using the station compared to a year ago, the answer among all respondents might be: 28% using the station more, 55% using the station about the same as a year ago, and 17% using it less. But if these figures are cross-tabulated to reflect only members of your station, the responses may be quite different. The only way to determine if there are significant differences in the responses is by cross-tabulation.

Any good database software program that is used to process the results should be capable of cross- tabulating data.

Regardless of who analyzes the survey data, and whatever the outcome, remember that research results only benefit a station if they are implemented. Be sure to determine the conclusive, significant results from your survey, and adapt them into a plan of action to take advantage of the knowledge gained from this research.

[ Chapter 7 Table of Contents | Previous Section | Next Section | Toolkit Home ]