From an accuracy standpoint, the strongest evidence will be generated if everyone completes a survey. However, most of the time, it is not possible or appropriate to survey all employees. A representative group of employees should receive surveys if the cost in time and other resources makes it impractical to survey everyone.

When it is necessary to survey some, but not all, employees, getting a representative sample of employees to complete surveys is important for gathering accurate information. The following issues should be taken into consideration when determining who should get surveys:

• There are psychological and political reasons for selecting a sample size. You will be calling on the work group to carry out needed changes. Asking for input now will build a spirit of cooperation and an increased awareness about your efforts. This is an argument for surveying as many people as possible. Leaders may have already established standards for valid surveys. You should take their expectations into consideration when determining who will get surveys.
• If you will be breaking down information by subpopulations (such as work areas, job classifications, race or gender), be sure to get an adequate number of respondents so each group will be represented.
• Up the sample size to assure the anonymity of respondents. In groups of less than 25 people, everyone should be surveyed. This will help assure anonymity.
• Up the number of people who get surveys if you expect a low response rate. One way to estimate response rate is to look at the response to past surveys.

Confidence intervals can be used to help determine the statistical implications of your sample. Information about the number of completed surveys (or your estimate for completed surveys) and the number of total possible respondents determines a statistical measure of scale accuracy. A confidence interval or standard margin of error provides a range of scores that most likely includes the "true value" that would have been the result if all employees had completed surveys. The confidence interval is normally expressed in percentage terms (for example, plus or minus 2%). A smaller interval means a more accurate and sensitive measure of climate.

In order to compute the confidence interval, you will need to find out the total number of employees in your group or organization. Take the total number of employees who you estimate will complete surveys and divide it by the total number of employees to determine the percentage of the possible respondents who will complete surveys. Use that percentage to select a cell in the table below. For example, if you estimate 446 surveys will be completed and you had a total workforce of 892, the confidence interval would be associated with the table value for 50% (plus or minus 3.3%). You will need to make changes of that magnitude in order to pick up statistically significant improvements in future survey scores.

Confidence Interval Estimates

Do not get discouraged by difficulties in conducting your survey. If the environment has a poor social climate, it may be hard to tackle the political hurdles and apathy. If getting the survey done is a challenge, this very fact is evidence for the need to address the work climate. Furthermore, you are asking for peoples’ perceptions about the work environment. Individual survey responses will tell a story of the bigger picture. Every survey answered helps complete the picture.