Online Survey Questionnaire: Design, Data Collection, and Cleaning

What an online survey questionnaire means in quantitative research

An online survey questionnaire is a questionnaire delivered through an online platform so respondents can complete it on a phone or computer. In a quantitative thesis, it turns research concepts such as satisfaction, purchase intention, or service quality into numerical data in a .csv or .sav file.

A thesis questionnaire usually has four parts: screening questions, items measuring the research variables, demographic questions, and a thank-you message. You need to identify which construct each item belongs to before adding the questions to the form. A questionnaire with many items that are not connected to the model and hypotheses will produce data that is difficult to analyse later.

If you are drafting the questions, you can also review customer survey questionnaire guidance to compare how the introduction, screening questions, and groups of items are arranged.

How many responses your online survey questionnaire needs

Sample size depends on the model, the number of items, the analysis method, and your access to respondents. Do not choose a fixed number simply because you saw another thesis use it. State clearly how you determined the sample size in your proposal and keep evidence of the number of valid responses.

If the population is finite and you know its size, N, the Yamane formula is often written as follows:

n = N / (1 + N × e²)

Here, n is the required sample size, N is the population size, and e is the allowable error. This calculation is cited from (Yamane, 1967). If you are estimating a proportion in a population, you may need the approach in (Cochran, 1977), depending on the research design and your supervisor's requirements.

For regression analysis, a commonly used rule is a minimum sample size of 50 + 8m, where m is the number of independent variables, according to (Tabachnick and Fidell, 2013). For EFA, some theses use the number of observations per item. The level of 5 to 10 observations for each item is cited from (Hair et al., 2010). This is a reference rule, not a substitute for considering the model and the quality of the data.

This table is illustrative output.

For example, if your model has 5 independent variables, the regression rule gives a minimum of 50 + 8 × 5 = 90 observations. You should still allow for responses that may be removed because of missing data, unusually fast completion, or selecting the same option for nearly every question. If you need 200 valid responses, the actual number of questionnaire submissions must be higher than 200.

Designing the questionnaire for this purpose

Before creating the form, prepare a mapping table containing the variable name, variable code, number of items, scale source, and response format. For example, the construct “satisfaction” may be measured with the items HL1, HL2, and HL3, with every item using the same Likert scale. Consistent coding makes it easier to import the data into SPSS without renaming dozens of columns later.

Put screening questions at the beginning

Screening questions determine whether a respondent belongs to the target population. You might ask when they used a product, whether they have purchased a service, or whether they belong to the age group covered by the study. If they select “No”, the form should end or move to an appropriate information page.

Screening questions should be short and limited to criteria directly related to the sample scope. Do not add many personal questions simply because the platform allows it. Every additional piece of information takes more time for respondents and creates another column to clean.

Use the Likert scale consistently

A 5-point Likert scale is commonly presented from “Strongly disagree” to “Strongly agree”. The Likert scale originates from the attitude-measurement technique in (Likert, 1932). You can use a 7-point scale if the model or original scale requires it, but you should not mix 5-point and 7-point scales within the same group of items without a clear methodological reason.

Each item should express one idea. A statement such as “The application is easy to use and processes transactions quickly” measures two characteristics at once, so respondents do not know which criterion to evaluate. Split it into two items if both characteristics need to appear in the model.

Check the content before sending the form

Read the entire questionnaire on a phone, complete one trial as a respondent, and complete another trial as the data manager. Check required questions, page-branching logic, the “Other” option, the one-response limit, and how the form appears on a small screen.

You can also review how to create a survey form to check the form structure. If you choose Google Forms, the Google survey form guide can help when you need to review how questions are organised and data are exported. Whichever tool you use, the scale content must still follow your research model.

Collecting the data

Before distributing the questionnaire, define where and when you will reach respondents and which criteria make a response valid. Distribution channels may include suitable customer groups, professional communities, email, or social media. Record when the survey opened and closed, the number of submissions, the number of completed forms, and the number of removed responses.

With Google Forms, data are usually linked to Google Sheets and then downloaded as an Excel file or CSV. It is easy to start with and familiar to many people. You can also review the Google survey form guide to compare the implementation steps.

fillform.info is DoThesis's Vietnamese form tool for collecting questionnaire responses. You can use it when you need to create a questionnaire, share a link, and monitor the data in one place. Google Forms is a suitable alternative if you already use the Google ecosystem or need to collaborate directly in Google Sheets.

When sharing the questionnaire, the introduction should state the academic purpose, estimated completion time, participation conditions, and how the data will be used. Do not write that the survey “takes only one minute” if respondents actually need 7 to 10 minutes. An accurate description can reduce dropouts and limit careless responses.

You should also test the questionnaire with a small group before distributing it widely. At this stage, check whether respondents understand terms such as “usage intention”, “perceived value”, or “service quality” in the same way. If you change the content after collecting many responses, record when the change was made and consider whether different groups received two different versions.

Cleaning the data before importing it into SPSS

After closing the questionnaire, download the original data file and create a copy for cleaning. Do not edit the original file directly. Column names should be short, contain no accents, and avoid special characters, such as GT1, GT2, HL1, and HL2. Keep time, email address, or tracking-code columns only if they support checking and you are permitted to use them.

Remove responses with too much missing data

Determine the proportion of unanswered items in each row. A response that stops after only a few opening questions usually does not qualify for analysis. Required questions can reduce the number of blank cells, but you still need to check cases where data were corrupted during file export.

Check for straight-lining

Straight-lining occurs when a respondent selects the same level for nearly every item, such as choosing level 3 for all items. This is a sign that requires review, but it is not enough to remove the response automatically. Combine completion time, reverse-coded items if any, and the plausibility of the answers before making a decision.

Check completion time and duplicate patterns

An unusually fast completion, several rows with the same demographic information, or identical response patterns may indicate duplicate data. There is no single time threshold that applies to every questionnaire. Examine the completion-time distribution in your own data and record the removal criterion in the methods section.

Standardise codes and variables before running the analysis

Categorical choices such as gender, age, or usage frequency need consistent coding. Check whether an option appears as “Male”, “male”, or “1” in different rows. For reverse-coded variables, define the coding formula before calculating Cronbach's Alpha. Do not delete identifier columns or screening questions until you have saved the original data file.

A short checklist before importing the file into SPSS includes the number of response rows, the number of valid responses, blank cells, removed rows, variable coding, reverse-coded variables, and the final file version. Keep a cleaning log so that when your supervisor asks why the sample size decreased, you can explain each step.

How to present this in Chapter 3

In the methods chapter, describe the online survey questionnaire according to a process that readers can check. Report the target respondents, sampling method, collection period, distribution channels, number of questionnaires received, and number of valid questionnaires. If the questionnaire includes screening questions, state the criteria used to retain or remove responses.

You can include a scale summary table with the construct name, variable codes, number of items, scale type, and reference source. Put the full questionnaire in the appendix, while Chapter 3 only needs to describe its structure and design principles. Explain the scale used in the context of your study instead of copying the wording from another questionnaire without adjustment.

You can adapt the following wording directly: “The study collected data using an online survey questionnaire during [period]. The respondents were [target respondents], selected according to [criteria]. After removing [number] invalid responses because of [reason], [sample size] observations remained for the subsequent analysis steps.”

If you use a sample-size rule, report both the number of variables and the calculation. For example: “The model contains [m] independent variables, so the minimum sample size under the 50 + 8m rule is [result], according to (Tabachnick and Fidell, 2013).” Replace the placeholders only with actual numbers from your file.

How to write this in your thesis

“The study collected data using an online survey questionnaire during [period]. The respondents were [target respondents], selected according to [criteria]. After removing [number] invalid responses because of [reason], [sample size] observations remained for the subsequent analysis steps.”

If you use a sample-size rule, report both the number of variables and the calculation. For example: “The model contains [m] independent variables, so the minimum sample size under the 50 + 8m rule is [result], according to (Tabachnick and Fidell, 2013).” Replace the placeholders only with actual numbers from your file.

Common mistakes when creating an online survey questionnaire

The first mistake is creating the form before finalising the model. You may then add questions based on intuition without knowing which variable will test which hypothesis. Return to the conceptual model, variable list, and scales before creating each group of questions.

The second mistake is using self-written questions without checking the scale source or the context in which it was applied. A close translation can still be difficult for respondents to understand. Adjust the wording, test the questionnaire, and keep the final version so you can explain what you did.

The third mistake is collecting many responses without controlling the target population. A data file with 500 rows does not help if most respondents do not belong to the target population. Sample quality, screening criteria, and the way responses are recorded matter more than simply increasing the count.

The fourth mistake is deleting data without keeping a log. When you run SPSS, you may need to review a removed row or explain why the sample size in Chapter 3 differs from the number of completed forms. Save the original file, the cleaned file, and the list of removal reasons separately.

Frequently asked questions

What is an online survey questionnaire, and can it be used for a thesis?

An online survey questionnaire is a questionnaire completed on an online platform and then exported as data for quantitative analysis. It is suitable for a thesis when you clearly define the target respondents, sampling criteria, scale content, and method for checking valid responses.

Is Google Forms enough for creating an online survey questionnaire?

Google Forms can handle basic needs such as creating questions, branching, limiting responses, and exporting data. The tool is only a data-collection medium, so it does not determine whether a questionnaire is valid. You still need to build the questions around the model, check the logic, and clean the file before running SPSS.

How many questions should an online survey questionnaire contain?

There is no fixed number of questions for every study. The number depends on the constructs, the items required to measure them, screening questions, and sample-description information. The questionnaire should measure the model adequately without becoming so long that respondents select answers quickly just to finish. Complete it on a phone to see the actual time required.

How many responses are enough to run SPSS?

The required sample size should be based on your analysis method and model. For regression, the 50 + 8m rule is cited from (Tabachnick and Fidell, 2013). For EFA, the rule of 5 to 10 observations per item is cited from (Hair et al., 2010). These are reference levels, so discuss the sample with your supervisor if the model is complex.

What should I do if the online survey data are missing or respondents selected the same level?

Keep the original file, mark the problematic rows, and define the removal criteria before deleting anything. Check the missing-data proportion, completion time, repeated response patterns, and screening questions. Straight-lining should be assessed together with other signs because some respondents may genuinely select the same level for several similar statements.

Can I edit questions after data collection has started?

You can correct technical or wording errors, but record when the change was made and what was changed. If the change alters the meaning of an item, data collected before and after the change may no longer be fully equivalent. Report this clearly in the methods section and ask your supervisor before combining all responses.

Open your current data file, list the variables, and check the sample size, invalid responses, variable coding, and questionnaire version in that order before running the analysis. If you need to run the analysis on your own .sav or .csv file, DoThesis M4 data analysis supports this step.