What Is a Questionnaire? How to Create a Quantitative Survey Questionnaire

What is a questionnaire in quantitative research

A questionnaire is a set of questions and items arranged deliberately to collect data from respondents. In quantitative research, a questionnaire turns research concepts such as satisfaction, purchase intention, service quality, or perceived usefulness into variables that can be coded numerically.

You can think of the questionnaire as the bridge between your research model and your data file. Each hypothesis requires one or more constructs. Each construct then needs suitable items. When respondents select a level on the questionnaire, their answers become values in a .sav or .csv file that you can analyse with SPSS or SmartPLS.

Questionnaire and survey questionnaire are different terms, but in practice they are often used almost interchangeably. If you need to distinguish more carefully between the data collection instrument and the activity of collecting data, you can read what is a survey. The article on survey questionnaires is more useful when you need to see the structure of a complete form.

A quantitative questionnaire usually has five content groups: an introduction, screening questions, respondent profile questions, the main scales, and a thank-you message. Your study may not need all five groups, but you should define the purpose of each section before creating the form.

How to determine how many items your questionnaire needs

The number of items should not be decided by instinct or by making the questionnaire long enough to fill a certain number of pages. Start with your research model, the number of constructs you need to measure, and suitable measurement scales. A construct such as satisfaction may be measured with several items, while demographic questions are often single descriptive variables.

For quantitative research, Hair et al. (2010) give a reference rule of 5 to 10 observations for each item. This helps you estimate sample size, but it is not a reason to add items that do not support your hypotheses. If your questionnaire has 25 items, this rule suggests approximately 125 to 250 observations as a reference minimum.

You also need to check the requirements of the analysis method. For regression, Tabachnick and Fidell (2013) propose a sample size of 50 + 8m, where m is the number of independent variables. This formula does not replace consideration of your model, sampling method, or the proportion of invalid questionnaires.

If you use an adapted scale, record the author, year, number of items, and original context. An English questionnaire may need wording adjustments so that respondents understand each statement, but changing the wording too far can mean that the item no longer measures the same construct. The article What is a Likert scale explains how to choose and interpret response levels in more detail.

Designing a questionnaire for your research objectives

How to create a questionnaire should begin with a cross-checking table that connects each hypothesis to the data you need to collect. For each latent variable, record the construct name, operational definition, variable code, statement, and scale. This helps you find questions that are not connected to your analysis before the form is distributed.

Write a short, clear introduction

The opening section should tell respondents what the questionnaire is for, who is conducting it, the expected completion time, and how the data will be used. You do not need to present the entire theory in this section. A short, neutral paragraph usually gives respondents a clearer understanding of what they need to do.

Avoid an introduction that lets respondents guess the answer you want. For example, the question, “Do you agree that this service is excellent?” already contains a positive evaluation. Replace it with a neutral statement about their actual experience.

Add screening questions

Screening questions help you exclude people who do not belong to the target population before they reach the main section. For example, if your study concerns people who purchased a product within the last three months, a screening question can ask when they made the purchase and end the form for respondents who do not meet the criterion.

A screening question needs a specific criterion. Decide in advance which values will be retained in the data file and which values will be excluded. Without this step, you may collect many responses without obtaining the correct research sample.

Choose the right question format

Closed-ended questions are suitable when the response options are already known, such as gender, age group, usage frequency, or level of agreement. Open-ended questions should be used only when a free-text answer is genuinely necessary, because they are harder to code and usually do not fit a basic quantitative analysis.

For categorical variables, the response options should cover the reasonable possibilities without overlapping too much. For questions about income or age, you can divide responses into ranges. Set the coding scheme while designing the questionnaire, such as 1 for the first group and 2 for the second group.

Use the Likert scale consistently

A Likert scale is commonly used to measure the level of agreement with a statement. You can choose 5 or 7 points, but you need to keep the same logic across the scale groups. For example, 1 can mean strongly disagree and 5 can mean strongly agree. The instructions should appear clearly before the group of items.

Each item should express one idea. The statement “The application is easy to use and processes transactions quickly” measures two aspects at once. If a respondent agrees with one part but not the other, the resulting data become difficult to interpret.

Arrange the questions in a clear order

You can use the following order: introduction, screening, main questions, classification information, and thank-you message. The main questions should move from familiar experiences to overall evaluations. Avoid placing sensitive questions at the beginning before you have established the necessary context.

Before distributing the questionnaire, complete it yourself on both a phone and a computer. Check required questions, page-branching conditions, the “other” option, the one-response limit, and the data export function. This is how you find errors that are easy to miss when you only read the draft.

Collecting responses

After completing the questionnaire, determine where your sample will come from, what makes a response valid, and how long data collection will last. The sample source must fit the target population in your thesis. If you are studying customers who have used a service, sending the questionnaire to people who have never used it will put your sample outside the intended scope.

You can create an online form with Google Forms. It is familiar, easy to share, and suitable for a basic questionnaire. DoThesis also has fillform.info, our own form tool for creating questionnaires and collecting responses in Vietnamese. Choose the tool based on your data controls, file export process, and the functions you actually need.

When sharing the link, state the eligible respondent group, estimated completion time, and the rule that each person should respond only once if your research design requires it. Posting a link in one group and calling the resulting responses representative is not enough. Record the collection channel, the period when the form was open, and the number of responses received in each round.

A questionnaire that is too long can cause respondents to stop halfway or select answers quickly just to finish. Test it with several people from the target group to check completion time and identify statements that are understood differently. This tests the instrument. It is not official data to transfer directly into your analysis.

If you want to see more layout examples, you can refer to survey form templates. When you need to create a form quickly on a common platform, the article on Google surveys focuses on the implementation steps.

Cleaning data before importing it into SPSS

Data exported from a form are not necessarily ready for running Cronbach's Alpha, EFA, or regression. First, save an original copy of the data and create a duplicate for cleaning. Do not edit the only copy because you may need to check the original responses later.

Check for rows with too many unanswered items. If a respondent leaves most of the main scale blank, you need a removal rule before processing the file. For a few blank cells, the appropriate treatment depends on the missing proportion and the method you plan to use. Record the number of removed rows and the reason for removal.

Next, look for responses that show straight-lining, meaning that the respondent selected the same level for almost every item. This sign alone is not enough to conclude that the data are invalid, because some respondents may genuinely have consistent evaluations. Also check completion time, screening answers, and reverse-coded items if your questionnaire uses them.

Check variable codes and data types. Column names should be short, contain no accents or spaces, and match the codes in your measurement-scale table, such as HL1, HL2, HL3. Do not leave one column containing both the text “Agree” and the number 4 if you intend to analyse it in SPSS.

Columns for submission time, email, or identification codes should be retained only if they support data checking or the research design. If you will not analyse these columns, you can separate them from the analysis file to reduce the risk of exposing personal information. Then open the file in SPSS and check Variable View and Data View before running any test.

How to present the questionnaire in Chapter 3

In the methodology chapter, describe the questionnaire in enough detail for the reader to understand how the data were created. Present the survey objective, respondents, sampling method, collection period, number of questionnaires distributed, number returned, and the criteria for excluding invalid questionnaires if your study contains this information.

You can create a table describing the item groups as follows. The table below is illustrative output.

The table is illustrative output, not a validated scale for your thesis. In the final version, replace it with the construct names, variable codes, and sources that match your own model. If you use a scale from the literature, explain how you translated it, adjusted the wording, and tested the questionnaire.

How to write this in your thesis

You can write the following: “The questionnaire consisted of [number of sections] sections. The [section name] section used [number of items] items to measure [construct]. The items were measured with a [5/7]-point Likert scale, ranging from [lowest level] to [highest level]. Data were collected from [target respondents] during [period].” Replace every bracketed placeholder with the actual information in your file.

Common questionnaire design errors

Adding questions that are not connected to the model

A question may sound interesting but still fail to support any variable. It makes the questionnaire longer and makes the results harder to explain. Before keeping a question, identify which construct it measures and which analysis will use it.

Using leading or ambiguous wording

Words such as “frequently,” “reasonable,” and “quick” can be understood differently by different respondents. When necessary, connect the statement to a specific time, behaviour, or experience. Leading wording can also push respondents toward one answer.

Reverse-coding an item and forgetting to recode it

If an item has the opposite meaning to the other items, record it and handle the coding before analysis. Forgetting to reverse-code it can reduce correlations between items and cause you to remove the wrong item. Use reverse-coded statements only when you have a clear methodological reason.

Failing to test the form link

A required question set incorrectly, a broken page branch, or a response option with no suitable answer can all lead to lost data. Complete the entire form at least once as a respondent, then open the exported file and check the column names.

Treating the number of responses as evidence of sample quality

A large number of responses does not automatically mean that the sample is appropriate. You still need to check the target population, duplicate responses, unusually short completion times, and missing data. Set these criteria before removing data so that you do not select cases based on the result you want.

Frequently asked questions

What is a questionnaire, and is it different from a survey questionnaire?

A questionnaire is an instrument containing questions or items used to collect data. Survey questionnaire is often used with a similar meaning in English. The important point is to design the questions around your research objective and make sure that the responses can be converted into analysable data.

How many questions should a questionnaire contain?

There is no single number that fits every thesis. The number depends on the constructs, items, screening questions, and respondent profile information you need. Prioritise scales with a clear basis and avoid adding questions simply to make the questionnaire longer.

Should you use a 5-point or 7-point Likert scale?

Either can be used if it fits your research design and respondents' ability to distinguish between response levels. A 5-point scale is easy to explain and often convenient for broad surveys. The important requirement is to keep the meaning of the levels and the coding direction consistent throughout the questionnaire.

Should you use Google Forms to create a questionnaire?

Google Forms is suitable for a structured online questionnaire that needs to be shared quickly and exported as data. You can also use fillform.info, DoThesis's form tool, when you need an option focused on Vietnamese questionnaires. Whichever platform you use, check the form logic and the exported data file afterward.

What should you do when questionnaire data are missing or respondents select the same level?

Check each case against the rules you set in advance. A row with many missing items, an unusually short completion time, or the same response level for every item may require review, but you should not remove a response because of one sign alone. Keep the original file, record the number of removed rows, and state the criteria in the methodology section.

Open your questionnaire file now, create a cross-checking table between each hypothesis and its items, then test the form link and exported data file before distribution. If you need to build a research model, measurement scales, and questionnaire for your own topic, use DoThesis M3 for research models, scales, and questionnaires.