Analyse Your Survey Results

printer Printer Friendly Version

Author John Towler, Ph.D.

How to manage your employee survey and what to do with the results

You'll need to figure out what trends and issues the data you collect reflects. Survey software offers data analysis capability, and other good statistical packages can help you analyze your survey results. Beware of the pitfalls of statistical analysis. You don't want to draw erroneous conclusions. Data can be and have been manipulated to support numerous conclusions.

Investigate Your Data Analysis Options
There are many ways to analyze the data you get, and the more knowledgeable you are, the more analytical you can be. All the survey software programs have some kind of statistical tools in them. If they aren't suitable, you can always use stand-alone statistical packages, but doing so requires some familiarity with the procedures they offer. offers good statistical packages. Harvard University has an excellent review of statistical packages; the packages reviewed range from the simple to the ultracomplex.

Beware of Data Analysis Pitfalls
Make sure you aviod erroneous conclusions. You'll want to pay particular attention to percentages if your sampling is small, because the margin for error increases as the size of the sample decreases. If you need to be statistically accurate with your results, you will need to know something about statistics and how to apply them to your survey.

Don't fall into the trap known as "paralysis through analysis." Keep your analysis simple if you can. Measures of central tendency, such as mean (average), median (the midpoint), and standard deviation (the distance from the mean), make the most sense to most people.

Beware of bending the results to make them fit what you want them to be. It has been said that "liars figure and figures lie." Be sure that you are rigorously honest and accurate in your conclusions. Suppose you found that 51 percent of your customers like brand A and 49 percent like brand B. Saying that the majority of people like brand A would be true but misleading, since there is really only a small difference.

Another statistical error can come from inaccurate or improper sampling procedures. If your survey was designed to find out whether teen-agers are smoking but you had no way of knowing the ages of the people who responded to the survey, you cannot say anything about teen-age smokers. Even if you did know the ages of your respondents and got 12 teen-agers among the total of 150 responses, you still cannot legitimately claim that your results show anything about teen-age smoking. The sampling is just too limited to allow you to draw a general conclusion.

John Towler is a Psychologist and the founder of Creative Organizational Design. Please send comments about this article to For more information, please contact us.

Re-printable with permission.