From QualtricsWiki

How do you create a survey? First, you need a better understanding of the characteristics of people who engage in the behavior you are interested in. Building a focused and effective survey questionnaire will help you to pinpoint the nuggets of information required to make more informed decisions.

Building a survey questionnaire is both an art and a science. Just as an artist has a palette with variety of different colors to choose from, a researcher has a variety of different question formats with which to build a question that gives an accurate picture of important customer segments, client perceptions or other issues.

The following question types are basic to building a great survey questionnaire.

Contents 1 The Multiple Choice Question 1.1 The Dichotomous Question 2 The Matrix/Multiple Choice Battery Question 2.1 The Likert Type Rating Scale 2.2 The Semantic Differential Scale 2.3 The Staple Scale 3 The Open-Ended Question 4 The Constant Sum Question 5 The Rank Order Question

[edit] The Multiple Choice Question

Multiple-choice questions have three or more exhaustive, mutually exclusive categories.

When building a survey, be aware that multiple choice questions can be used for both single and multiple answers. Radio buttons are present for single answers and check boxes for multiple answers.

We could ask (and even have the survey engine force) the respondent to select:

• only one answer from the 7 possible
• exactly 3 of the 7
• as many as 3 of the 7 (1,2,or 3 answers can be selected)
• as many as apply (any number of answers can be selected)

Example of select only 1 of 7 options:

For this type of question it is very important to consider including an "other" category because there may be other answers.

[edit] The Dichotomous Question

The dichotomous question provides two options and is usually a yes/no question. Questions can be displayed in many different formats. Examples of the dichotomous question include:

Dichotomous questions are often seen in screening questions that ask whether or not the respondent has done something such as purchased or used a product or service. Researchers use "screening" questions to make sure that only qualified individuals participate in the survey.

Yes/no questions can also be used when building a survey, to separate people or even to branch people to questions for those who "have purchased" and those who "have not yet purchased" your products or services. Once separated, different questions can be asked of each of these groups.

You may want to ask the purchasers how satisfied they are with your products and services, and ask the "non-purchaser" what the primary reasons are for not purchasing. In essence, your questionnaire branches to become two different sets of questions.

[edit] The Matrix/Multiple Choice Battery Question

The Matrix/Multiple Choice Battery Question is actually a series of questions, all of which have the same answer scale. For example, multiple attributes of a dealership could be evaluated on a 5 point satisfaction scale that includes the scale items: Delightful, Very Satisfied, Satisfied, Somewhat Satisfied, Failure

Grouping identically scaled multiple choice questions into batteries keeps similar questions together and thereby reduces response time and respondent fatigue.

The multiple choice matrix questions may use either the radio button (one answer), check box (multiple answers), or spreadsheet (text input) formats. These formats give you a lot of versatility and breadth. Typical questions might include measures of agreement with statements about the degree of preference or degree of satisfaction with a list of product, service, or attributes.

Randomization bias may occur if the same question appears at the top of the list for each respondent. Randomization corrects this bias by randomly rotating the order of the multiple choice matrix questions for each respondent.

The sections immediately following this section will explain scales that are frequently used in Matrix questions.

[edit] The Likert Type Rating Scale

Rating scale questions are used when building a survey that requires a person to rate a product or brand along a well-defined, evenly spaced continuum. Rating scales are often used to measure the direction and intensity of attitudes. Rating scales can be used in individual question, matrix question and side by side question formats. The following is an example of a comparative rating scale question:

[edit] The Semantic Differential Scale

The semantic differential scale asks a person to rate a product, brand, or company based upon a seven-point rating scale that has two bi-polar adjectives at each end. The following is an example of a semantic differential scale question.

Notice that unlike the rating scale, the semantic differential scale does not have a neutral or middle selection. A person must choose, to a certain extent, one or the other adjective.

[edit] The Staple Scale

The staple scale is another variant of the multiple choice question that asks a person to rate a brand, product, or service according to a certain characteristic on a scale from +5 to -5, indicating how well the characteristic describes the product or service. The following is an example of a staple scale question:

[edit] The Open-Ended Question

The open-ended question seeks to explore the qualitative, in-depth aspects of a particular topic or issue. It allows the respondent to respond in detail, but places few constraints on the nature of their response. Although open-ended questions are important, it is time consuming to code their responses and should not be over-used. Examples of open-ended questions might be:

[edit] The Constant Sum Question

The constant sum scale question produces what is assumed to be a ratio measurement data. Ratio data is the most powerful of all measurement scales because it is characterized by an absolute zero point and an equal interval scale. Ratio scales are often used to measure the magnitude of a characteristic and scale the differences between alternatives.

Constant sum data is obtained by asking the respondent to "Assign 100 points (or percent) across the answer options so as to reflect your degree of preference, importance, or other evaluation" Typical questions might include identifying not only liking or preference, but attribute strength, or intention to buy, look for a new job and so forth.

Randomize: We know that in elections, being the first on the list increases chances of election. Similar bias occurs in all questionnaires when the same answer appears at the top of the list for each respondent. Randomization corrects this bias by presenting a random choice order for each respondent.

Example: The following question asks you to divide 100 points between a set of options to show the value or importance you place on each option. Distribute the 100 points giving the more important reasons a greater number of points. The computer will prompt you if your total does not equal exactly 100 points.

This type of question is used when you are relatively sure of the reasons for purchase, or you want input on a limited number of reasons you feel are important. Questions must sum to 100 points and point totals are checked by a java script.

[edit] The Rank Order Question

The rank order question provides more powerful data than a simple multiple choice question for selecting nominal or categorical items. Ordinal scales apply order to the data. Unlike rank order data, most multiple choice scales do not permit us to say that one item is greater than another.

The rank order question provides direction and relative position, but not absolute difference. That is, rank order data indicates order, but does not tell us how much one item is preferred over another. The rank order question is a powerful tool because respondents often misuse traditional rating scales, indicating ties and evaluating all choices within a narrow 1 or 2 point range.

Rank order data uses an answer format that requires the respondent to assign a rank position for the first, second... up to the nth item to be ordered. This format of assigning position numbers is very versatile. Respondents may be asked to rank a specified subset from the list (such as their first, second, and third choices from a list), or to rank all items in the list. Typical questions might include identifying preference rankings, attribute association strength, first to last, oldest to youngest, or relative position (most , next most, and so forth, until either a set number of items is ordered or all items may be ordered).

Randomize Rank Order Items: We know that in elections, being the first on the list increases chances of election. Similar bias occurs in all questions where the same answer appears at the top of the list for each respondent. Randomization corrects this bias by presenting a random choice order for each respondent.

Are Rank Order Ties Allowed? If ties are permitted, several items may be evaluated as having the same rank. In general this is not a good idea because it weakens the data. However, if ties truly exist, then the ranking should reflect this. If branching is selected and a tie occurs, the first item with a tie is selected. This is an arbitrary rule, but one that makes sense if answers are randomized.