Survey Form: How to Create a Questionnaire and Collect Quantitative Data

What is a survey form in quantitative research

A survey form is the tool you use to turn the concepts in your research model into questions and items that can be answered with numerical data. A complete form usually includes an introduction, screening questions, groups of items for each construct, demographic information, and a thank-you message.

In a quantitative thesis, a survey form is more than a link sent to respondents. It determines whether you reach the right participants, whether the items can be entered into SPSS, and whether the results from Cronbach's Alpha, EFA, or regression can be interpreted. If the questionnaire contains vague or repetitive items, or items that are not connected to your hypotheses, collecting more responses does not solve the underlying problem.

You should distinguish between three concepts that are often used interchangeably. A questionnaire is the complete instrument, including the questions and response instructions. A survey form is the version of that questionnaire deployed on paper or through an online platform. Survey data are the responses exported into a data table, usually as .xlsx, .csv, or a file imported into .sav. If you need the basic concepts first, see what is a questionnaire and what is a survey.

How to determine the size of your survey form

You need to determine two separate numbers: the number of items in the questionnaire and the number of valid responses required. A form with 25 items does not mean that you need only 25 respondents. The sample size must fit the analysis method, number of variables, research model, and your ability to reach the target population.

If the population is finite and you know its size, the Yamane formula is commonly presented as follows:

n = N / (1 + N × e²)

Here, n is the required sample size, N is the population size, and e is the acceptable margin of error. This formula is widely used in Vietnamese theses and can be cited as (Yamane, 1967). State the values of N and e that you used instead of copying only the final sample-size result into Chapter 3.

When your study uses a scale with multiple items, a commonly referenced rule is 5 to 10 observations for each item (Hair et al., 2010). This is a planning guideline, not a substitute for considering your model and analysis method. If you have 30 items, the reference range under this rule is 150 to 300 responses.

The levels in the table above are illustrative output calculated using the rule of 5 to 10 observations per item, with the source (Hair et al., 2010). When planning, collect more than the minimum number of valid responses because some responses may be removed for blank fields, extremely fast completion, or selecting the same scale point throughout the questionnaire.

For example, if you need 200 valid responses and expect around 15% of responses to be unusable, your collection target should be higher than 200. You can divide the target by participant group, location, or recruitment channel so you can see which group is still underrepresented instead of sending the link everywhere without tracking coverage.

Designing the questionnaire for this research objective

Before opening Google Forms or another form builder, place your research model beside you. Each hypothesis must be measured through one or more constructs, and each construct needs suitable items. Each question should measure one main idea and use wording that respondents in the relevant Vietnamese context will understand in the same way.

Define the structure of the sections

A survey form for a thesis commonly has the following structure:

1. Introduction: the research purpose, who should respond, expected completion time, and a statement that the data will be used for research purposes.
2. Screening questions: determine whether the respondent belongs to the target population.
3. Behavioural or contextual questions: usage frequency, length of experience, service type, or job position when these details are relevant to the topic.
4. Groups of items: organize the items by construct in the model and code them as PU1, PU2, PU3 or SAT1, SAT2, SAT3.
5. Personal information: ask only for variables needed for analysis and sample description.

Give a short introduction before each item group, but do not explain the group in a way that suggests one answer is correct. For example, “Please rate your level of agreement with the following statements” is more neutral than “Please confirm that this service is very convenient.”

Choose question types and scales

Screening questions usually use a single-answer format. Questions about gender, age group, or experience can also use a single answer, while a question about multiple usage channels may allow respondents to select several answers. Check this setting carefully because choosing the wrong option can make the exported data difficult to code.

Items in the model commonly use 5-point or 7-point Likert scales. With a 5-point Likert scale, responses may range from 1, “Strongly disagree,” to 5, “Strongly agree.” Keep the scale consistent across the entire item group unless the study has a clear methodological reason to do otherwise. You can read more about what is a Likert scale before choosing the number of points.

If you have reverse-coded items, mark them clearly in the coding file. For example, a negative statement may need to be recoded from 1 to 5 and from 2 to 4 before you run the analysis. If you are working against a deadline, limiting reverse-coded items when they are not essential is safer because respondents can miss the word “not,” and you can forget to recode the values later.

Write questions that can be entered into SPSS

Each item should have its own code. Do not use the full sentence as the column name. In the coding table, you might record PU1 as “The system helps me complete my work more quickly” and PU2 as “The system increases my work efficiency.” When you export the data, short column names make the Reliability Analysis, Correlations, and Coefficients tables easier to read.

Do not combine two ideas in one statement such as “The application is easy to use and saves time.” If a respondent finds the application easy to use but does not think it saves time, there is no clear response option. Demographic questions should also include “Other” or “Prefer not to answer” when appropriate, but do not add these options mechanically to every question.

Deploying the data collection

Before sending the survey form widely, complete it yourself on both a phone and a computer. Check the path from the screening question to the final section, required fields, the display of long questions, and the exported data file. You should also ask a few people from the actual target group to read through the form and identify unclear wording. This checks the form itself, but it does not replace the formal analysis process.

Choose a form-building platform

Google Forms works well when you need to create a form quickly, share it through a link, and export the data to Google Sheets. It is familiar, accessible, and free to use in many situations. Its limitations are that the design, permissions, and data controls may not meet every requirement of your study. You can see the related guide in Google survey.

DoThesis has fillform.info, our own form tool for creating Vietnamese questionnaires and collecting responses. It is suitable when you want to manage the form and survey data within one workflow. Google Forms remains a reasonable alternative if your study needs only a simple online form.

Organize the respondent recruitment channels

Your recruitment channel must follow your sampling criteria. If your target group consists of students who have used an application, place a screening question about that experience before they reach the item groups. If your target group consists of employees in a particular type of business, define the conditions for job position, length of employment, or location clearly.

When sharing the form in social media groups, state who may respond, how long completion should take, and how the data will be used. Do not ask one person to send the form to many unrelated groups simply to increase the response count. A large sample with the wrong respondents can be harder to defend than a smaller sample that follows the eligibility criteria.

Track progress and collection time

The collection period depends on the target sample, how accessible the participants are, and the distribution channel. Instead of promising a fixed number of days, set checkpoints based on the number of valid responses received each day. After each round, review the distribution by age group, gender, location, or experience to identify groups that are underrepresented.

Do not enter additional responses yourself to reach the required sample size. Do not alter respondents' answers to make the data table look balanced. If you need to remove responses, keep an original data file and record the reason for removal in a separate column or tracking file.

Cleaning data before importing it into SPSS

After closing the form, download the raw data and name the file by date, such as khaosat_raw_2026-09-07.xlsx. Create a copy for cleaning and keep the original so you can check it when your supervisor asks about a record. Before running the analysis, standardize the variable names, value codes, and data types.

Check incomplete responses

Review rows with blank values in the screening questions, item groups, and dependent variables. If a respondent stopped halfway through the form, apply the removal rule you defined in advance. Required questions usually limit blank values on the platform, but invalid values can still appear because of how the response options were designed.

Check extremely fast responses and patterned responses

An extremely short completion time is a signal to investigate, not by itself proof that a response should be removed. Combine time with other indicators, such as selecting the same scale point for almost every item, giving contradictory answers to a check question, or failing the eligibility criteria.

Selecting one scale point continuously is called straight-lining. Do not remove every row with many identical responses because a genuine respondent may agree with several statements. Set a reasonable criterion and record the number of removed rows and the reason for each removal.

Standardize variable names and values

Change item columns to short names without diacritics or spaces. Check whether the Likert responses fall within the range you designed. If the scale uses 1 to 5 but the exported file contains 0 or 6, return to the raw data to determine whether the error occurred in the form or during data entry.

Columns such as timestamps, email addresses, respondent names, or tracking codes may not belong in the model. Delete them only after saving the original file and confirming that they are not needed for quality checks or anonymity protection. Keep demographic variables if you will use them to describe the sample or test group differences.

How to present the procedure in Chapter 3

In the methods chapter, describe the survey form in an order that lets the reader understand what you measured, who answered, and how you collected the data. You can present the survey objective, scale sources, language adjustments, questionnaire structure, Likert scale, sampling method, and data-cleaning criteria.

How to write this in your thesis

You can adapt the following paragraph: “The questionnaire consisted of [number of sections] sections. The [section name] section measured the construct of [construct name] using [number of items] items adapted from [scale source]. Respondents rated the statements on a [5/7]-point Likert scale, ranging from [lowest point] to [highest point]. Data were collected from [target group] during [period], and [removal criteria] responses were removed before the data were analysed using [SPSS/SmartPLS].”

Place the variable coding table in the appendix. It should include the variable code, statement wording, construct, and reference source. Chapter 3 only needs to summarize the structure, while the complete form can be attached in the appendix. If you translated or adapted a scale, describe the actual changes. Do not state that the scale has been validated if you have only consulted an existing source.

Common mistakes

The first mistake is creating the form before finalizing the model. You may then add questions based on intuition, leave out an item for a hypothesis, or include many items that you will never use. Create a table of the model, constructs, items, and variable codes before designing the form interface.

The second mistake is using leading questions. The wording “Do you agree that the product is very convenient?” already suggests the answer. Rewrite it as a neutral statement and let respondents choose their level of agreement.

The third mistake is failing to control respondent eligibility. Sending the form to anyone with the link may increase the sample count without representing the target population. The screening question should appear early, be easy to understand, and lead ineligible respondents to an appropriate end page.

The fourth mistake is downloading the file and immediately running SPSS. Keep the raw data, check the coding, review missing data, remove invalid responses according to the recorded criteria, and save each processed version. EFA, Cronbach's Alpha, and regression cannot fix a questionnaire that was designed incorrectly from the beginning.

Frequently asked questions

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

A questionnaire is the content and structure of the measurement instrument. A survey form is the version of that questionnaire deployed on a specific platform, such as Google Forms or fillform.info. Once the responses are exported, you have an additional layer of numerical data that can be analysed in SPSS or SmartPLS.

Is creating a survey form with Google Forms enough for a thesis?

Google Forms may be sufficient if you need to create questions, collect responses, and export the data. The quality of the thesis depends more on the model, participant selection, item wording, and data-cleaning process. You still need to test the form before sending it and keep the original data file.

How many questions should a survey form contain?

There is no fixed number of questions for every study. Start with the constructs and items required to test the model, then add the necessary screening and descriptive variables. A questionnaire that is too long can increase careless responses, while one that is too short may fail to measure the constructs adequately.

What sample size is sufficient for a survey?

If you use a rule based on the number of items, you can refer to 5 to 10 observations per item according to (Hair et al., 2010). If you know the size of a finite population, you can present the formula from (Yamane, 1967). The final sample size must fit the model, analysis method, and number of valid responses remaining after cleaning.

What should I do if survey data are missing or many respondents choose the same answer?

First, check the original file and determine which rows are genuinely invalid. Missing data, extremely short completion times, and straight-lining should be assessed together using criteria defined in advance. Do not change answers to make the results look better, and record the number and reason for every removed case.

Open your data file now and list the constructs, variable codes, and sampling criteria before creating the form. If you need help running the analysis on your own .sav or .csv file, you can use M4 Data Analysis by DoThesis.