A survey is a written questionnaire used to gather data about people and their opinions. It is a valuable tool for gathering data from a large number of people about a specific subject.

Use a Survey to

• Gain detailed knowledge of a specific subject you are investigating
• Reach a highly reliable sample of individuals representing a cross-section of the organization
• Gather a significant amount of objective data that may be analyzed and cross-analyzed
• Ensure anonymity for the people whose response you need

The following chart lists approximate percentages that should be sampled in order to ensure an accurate sampling of the population.

Number of People	Percent in Sample
1-20 21-40 41-75 76-125 126-250 251-2500 2500+	100% 70-100% 55-70% 40-60% 30-50% 15-20% 10%

 The object is to get as many surveys back as possible. The careful design of a survey can significantly influence the response rate.

 When preparing survey questions, try to make them easy to answer. Specific responses or checkmarks are quicker and easier to analyze than open-ended questions.

 When designing a survey, it is important to be very clear about the information you're seeking. This will inform who you survey, how you survey and precise wording of the questions.

1. Define the population or the representative sample of the population that you want to focus on.
2. Plan how you are going to administer the survey (in groups by email, mail etc.)
3. Prepare the questions
4. Pretest the questionnaire with a few individuals. Ask them:

• how long it took to complete the survey
• if the questions were clear and understandable
• if the instructions were clear
• ask for their interpretation of the most difficult questions
• ask what they thought of the survey in general

5. If necessary edit the questionnaire
6. Administer the survey

How to Use It

Step 1: Defining the Population Sample

When surveying your group, if feasible, it is best to survey the entire group, or population. This ensures that you have not inadvertently built any biases into your study and that your data reflects the perspective of the entire group.

However, it is often the case that the entire population is either too large or too dispersed for a survey the entire group to be feasible or cost effective. This is the time when some sample or representative percentage of the population is selected.

In developing a representative sample, it is important to identify those factors which create relevant subgroups to ensure that the sample reflects the population at large. This is known as a stratified random sampling.

For example, in sampling the perceptions of the workgroup, such factors as length of employment, position by title, geography or department may be important. Other factors may not matter. For example, it is unlikely that eye color or number of children will matter in the sample. Sex may or may not matter.

There are no predetermined criteria. The core questions are whether any of these factors are likely to influence how people respond to the survey, and whether you would like to analyze the data for any of the subgroups to look for differences among them.

Remember, you only have one opportunity to ask these questions. If you do not think through the groupings prior to administering the survey and build these categories into the identification elements of the survey, you will not be able to break up the categories later.

Use a representative sample that reflects a mix of your population in categories that you want to focus on, such as department location, salary, grade, years with the company, age, etc.

Use random sampling within the categories that you choose in order to minimize bias.

Random sampling means that once you identify a subgroup, such as the members of a given department, it is important to ensure that the individuals selected do not have any pre-biases. For example, if you asked for volunteers, will be some self-selection bias already built-in. If you go to the people who work in one area, you may have built in another unconscious bias.

The best way to select randomly is to use some form of standard technique. For a small group, names in a hat or a similar process works just fine. For larger groups, you may get a computer list of the entire population and take every third name, or whatever the percentage is that you want and work down the list. For very large groups, there is a statistical tool called the table of random numbers that can drive of this process.

Use statistical techniques to determine sample size. The following is a good general guideline for sample size selection.

Step 2: Planning How to Administer the Survey

Surveys can be administered by email, on-line, mail, telephone, fax, group sessions or one-on-one meetings.

Each survey administration procedure has some advantages and disadvantages. If you want to be sure that you get everyone's response, bring them into a room in groups.  Ask them to sit down and complete the survey and put it in a sealed box as they leave the room. Those who did not want to respond may resent being boxed in and their anger will be reflected in their responses. The technique does however ensure full response.

How people respond to this technique is very much influenced by how it is presented to them when they're in the room.

Another technique that generates a high response rate is to send (email, mail or fax) the survey to people and asked them to return it to its administrator within some period of time. When they return the questionnaire, they place it in a sealed box, but the administrator must observe them drop in and checked their name off the list. This gives a complete and accurate record of who responded, protects response anonymity, and allows follow-up to ensure that everyone participates.

This process, of course, requires time and monitoring.

The most common methods are mail, on-line or email-based surveys. This is best for very large groups. Mail surveys are the least threatening and absolutely anonymous. However, they also elicit the smallest percentage response, typically from 20 - 60 percent based upon the group being surveyed and the nature of the survey.

The dilemma, of course, is that you do not know who responded so a follow-up reminder with another batch of questionnaires must be sent to the full population again.

If mail surveys are used with several reminders, note that each succeeding wave of returns based on successive requests will be increasingly negative. Those who are more negative will tend not to respond. The more you ask them, the more they are likely to let you have it. Therefore, if you only do a single mailing release with no follow-up requests, then responses will be skewed to the more positive pro-active segment of the population.

Step 3: Preparing Survey Questions

The shorter the questionnaire, the more attentive respondents will be.

Closed-ended questions are best for service-oriented questionnaires. Use the following four basic close-ended question types to make the responses easier to analyze.

• Selection from several responses in defined categories:

1. How long have you been with the company?

a. Less than 1 year
b. 1-3 years
c. 3-5 years
d. 5-10 years
e. over 10 years

• Bipolar:

1. Do you currently contribute to the company's 401(k) plan?

Yes _________
No  _________

• Rank order:

1. Rank order (by importance of need) the type of training you need most for improving your job performance. (Rank one as most necessary, to five as least necessary)

___ Software related training
___ Writing skills
___ Total Quality problem-solving training
___ Presentation skills training
___ Other (specify)

• Multiple-choice answers:

1. In selecting a car, what is the single most important factor to you?

____Price
____Appearance
____Performance
____Features

 The greatest problem of surveys developed by those not professionally trained in survey design has to do with the creation of the individual questions.

The most common problem tends to be with incorporating several different ideas or thoughts into the same survey question. Be sure that each question covers only one thought and requests a response to that single thought. Test-run the questions before you finalize the survey to make sure that each question asked reflects a single focused thought.

 The second biggest problem in designing questions is the use of the words that are subject to wide interpretation. “Our company has a good benefits program.”  This question will elicit all kinds of responses.

Some will interpret the question to be asking about the entire size or value of the benefits. Others will respond to a particular on a program that they like or do not like. Still others respond to the way the program is administered, or explained to them individually. If you find this is the case when you test your questionnaire, then it would be best to further refine the question into several different questions, each of which is designed to gather data about a specific area.

Open-ended questions

Open ended questions can be used if a sample is relatively small and the number of questions is not too large.  Open-ended questions should be designed so that they do not structure or limit the respondents' answers. Distribute the questions randomly throughout the questionnaire to minimize bias and to enhance objectivity and reliability.

Examples:

• Are there any other issues you would like considered?
• Are there any other comments you would like to make?

 It is important to assure participants of anonymity, even if the information you're requesting is not controversial. Anonymity encourages honest answers.

Step 4: Gathering the Responses

It is important to have a plan for the distribution and collection of the survey. How the survey is circulated and how much time is allowed to respond depends in large part on the method being used to gather the information.

Ensure the survey responders have adequate time to respond, but not too much time.

Step 5: Interpreting the Data

Once the surveys are completed and the data has been collated, a number of tools (Histogram, Pareto Chart, Scattergram, etc.) can be used to convert the raw data into meaningful information.

 People react well to a visual display of the data, the more graphics you can make yourself the easier it will be to translate the raw data into meaningful information.