What Is a Survey in Quantitative Research?

What Is a Survey in Quantitative Research?

A survey in quantitative research is a way to collect data with a structured questionnaire, then code the responses as numerical data for analysis in SPSS, SmartPLS, or another statistical program. You can conduct the survey on paper, send an online form, or distribute the questionnaire to a group that fits your research topic.

In a thesis, surveys are commonly used to measure respondents' perceptions, attitudes, intentions, behaviors, or evaluations of the research variables. For example, a study of e-wallet usage intention might measure perceived usefulness, perceived ease of use, trust, and intention to continue using the service. Each construct is measured with several items, such as PU1, PU2, PU3 or INT1, INT2, INT3.

You should distinguish between a survey and a questionnaire. A survey covers the entire process, from identifying the target respondents and designing the instrument to collecting responses and processing the data. A questionnaire is the instrument used in that process. If you need a separate explanation of the instrument itself, read what is a questionnaire.

A good quantitative survey must connect three parts: the research questions, the variables to be measured, and data that can be analyzed. If a questionnaire item does not serve any hypothesis, you may collect a large amount of data without getting closer to answering your research question. If the questionnaire is too short and leaves out important indicators, you may have difficulty testing the model later.

How to Determine Survey Sample Size

Sample size is the number of valid responses you need for the analyses you plan to run. You should not choose it simply because other students in your class used the same number. Start with the target population, the number of items, the planned analysis method, and how easily you can reach suitable respondents.

If your study has a finite population and you know its size, the formula is often presented as Yamane's formula: n = N / (1 + N e²), where N is the population size, e is the allowable error, and n is the required sample size. When you apply the formula, state the values of N and e in the methodology rather than reporting only the final number. The finite-population approach is attributed to (Yamane, 1967).

For a questionnaire with many items, a commonly used rule is 5 to 10 observations per item (Hair et al., 2010). Treat this as a reference rule, not as permission to ignore sample quality. If your questionnaire has 30 items, this approach gives a reference range of 150 to 300 responses.

For regression, another check is a sample size of at least 50 + 8m, where m is the number of independent variables (Tabachnick and Fidell, 2013). For example, a model with 6 independent variables needs at least 98 observations under this rule. If you plan to run EFA, regression, and a mediation test, use the stricter requirement among the approaches above and allow for invalid responses.

You should collect somewhat more responses than the expected minimum. Some responses may need to be removed because too many items were left blank, the same option was selected throughout the questionnaire, or the respondent did not pass a screening question. Do not create extra response records just to reach the target sample size. Data without a real source makes the entire analysis impossible to defend.

Designing a Questionnaire for This Purpose

Before writing the items, open your research model and list the variables you need to measure. Each independent variable, dependent variable, mediator, or moderator needs a clear operational definition. Then identify which items will represent each construct and record the source of the scale if you are using an adapted or previously established scale.

A quantitative questionnaire usually has four parts. The opening section explains the survey purpose, the eligible respondents, the expected completion time, and the confidentiality principle. The next section contains screening questions to establish whether the respondent fits the research context. The main section measures the variables in the model. The final section collects demographic or classification information needed to describe the sample.

Choose Appropriate Question Types

Closed-ended questions make responses easier to code. You can use single-choice questions, multiple-choice questions, age-group categories, or questions about usage frequency. Use open-ended questions only when the available options genuinely cannot cover the information you need, because open responses are difficult to standardize for quantitative analysis.

Each item should ask about one idea. A statement such as “This application is easy to use and secure” combines two constructs, so respondents do not know which part they are rating. You should also avoid leading wording, absolute terms, and double negatives. Before distributing the questionnaire, read every item as if you were someone who had never seen the research model.

Choose a Likert Scale

A Likert scale allows respondents to indicate how strongly they agree with a statement. A 5-point scale often runs from “Strongly disagree” to “Strongly agree,” but the important point is to label the points consistently throughout the questionnaire. You can read what is a Likert scale for more guidance on coding and presentation.

Each item should have its own code, such as SAT1, SAT2, and SAT3. Do not name an SPSS variable with the entire sentence in the questionnaire, because long names make the output harder to read and easier to confuse. Store the full item wording in the variable label or the data description table.

Check the Screening Questions

Screening questions must appear before the measurement section. For example, if your study concerns people who have purchased from an e-commerce platform, ask whether the respondent has made such a purchase. A person who does not fit the criteria should be screened out or sent to an end-of-survey page rather than continuing to produce data outside the target population.

You should also state the expected completion time. A questionnaire that is too long can cause respondents to quit halfway through or select answers quickly just to finish. Run a pilot with several people similar to your main target respondents to identify unclear wording, missing options, and page-logic errors.

Deploying the Data Collection

Before sending the questionnaire, specify who will be invited, where they will be reached, and how long data collection will remain open. Convenience sampling can be appropriate for a student thesis when access is limited, but you must describe the sample source accurately and avoid presenting it as representative of the entire population.

You can distribute the link through class groups, professional communities, social media, or a customer list when you have lawful access. The invitation should state the participation criteria, estimated time, academic purpose, and how the data will be used. Avoid sending repeated invitations to the same person, because this increases the risk of duplicate responses.

Survey form should be easy to follow on a phone. DoThesis has fillform.info, our own form tool for creating and collecting responses to Vietnamese questionnaires. Google Forms is another common option, especially when you need to share the form quickly and export the data to a spreadsheet. You can also read Google surveys if you are comparing ways to deploy the questionnaire.

During data collection, monitor the number of responses and the completion rate. If many respondents stop at the same item, the cause may be unclear wording or a form-configuration error. If the number of responses rises unusually quickly from one source, record that source so you can explain the sample's characteristics.

Do not close the survey simply because you have reached a minimum number. First check the data columns, the proportion of missing responses, and duplicate records before deciding to stop. Collecting more data cannot repair a badly designed questionnaire, but it can reduce the risk of having too few valid observations.

Cleaning Data Before Importing It into SPSS

After closing the form, export the data as .csv or .xlsx, keep an untouched original copy, and create a duplicate for cleaning. If you use SPSS, you can save the processed file as .sav. Column names should be short, consistent, and reflect the item codes.

Start by removing technical columns that are not used in the analysis, such as a timestamp if you are not using it to check response speed, a session code, or an internal notes column. Do not delete a research variable simply because its name is difficult to read. Create a list of variable codes and their labels so you can trace every processing decision.

Check for incomplete responses. A record missing one demographic item may still be usable depending on your objective, but a record missing most of the measurement items usually should not enter the analysis. State the exclusion rule in advance or apply it consistently. Do not change the rule according to whether a particular row makes the results look better.

Check for straight-lining, meaning that a respondent selected the same point for nearly every statement. This is a reason to inspect the record, but you should not remove it automatically just because the values are identical. A respondent may genuinely agree with many statements. Consider completion time, screening questions, reverse-coded items if any, and the response pattern as a whole.

Next, check for out-of-range values. A Likert scale from 1 to 5 cannot contain 0 or 6. Categorical variables such as gender, age group, and occupation need clearly defined codes. If a question includes an “Other” option, decide how to code it and record the decision before running the analysis.

Finally, open the file in SPSS and check Variable View: variable name, variable label, data type, missing values, and measurement level. Do not enter “Do not know” directly into a Likert variable if you plan to run Cronbach's Alpha or EFA. Define a separate missing-value code, then check the number of valid observations again.

How to Report This in Chapter 3

In Chapter 3, describe the survey in an order that lets the reader understand how the data were produced. A typical description includes the survey objective, target respondents, sampling method, collection period, questionnaire structure, measurement scales, and data-processing procedure.

How to write this in your thesis

You can use the following template and replace the bracketed parts: “Primary data were collected through a structured questionnaire from [start date] to [end date]. The survey respondents were [description of respondents] and met the criterion of [screening criterion]. After excluding [number] invalid responses, the study used [number] observations for analysis in [SPSS or SmartPLS].”

If you use a Likert scale, state the number of points and the meaning of each point. If you adopted or adapted a scale, report its source, the original number of items, and the coding scheme. Sources such as (Likert, 1932) can be used when explaining the basis of the Likert scale.

You should also report how the sample size was determined, the number of questionnaires distributed, the number of responses received, and the number of valid responses. These figures must match across Chapters 3 and 4 and the appendix. Do not write “the survey covered 300 people” in Chapter 3 when the actual data file contains only 247 valid rows without explaining the difference.

Common Mistakes

The first mistake is collecting data before finalizing the model. You may then discover that an item is missing after the survey has closed, and adding a question halfway through creates an inconsistent dataset.

The second mistake is using a long questionnaire without running a pilot. An unclear item can cause many respondents to select the same answer or skip an entire page. Test the questionnaire from beginning to end on a phone before sending the official link.

The third mistake is treating every online response as valid. The data must be checked for screening eligibility, completeness, coding range, and signs of careless responding. Keep an exclusion log with the row code, reason, and processing date.

The fourth mistake is changing the questionnaire after many responses have been collected without marking the version. If a revision is necessary, save the earlier version, record when the change occurred, and check whether the two data groups still use the same measurement approach.

The final mistake is reporting the sample size without explaining the sampling method. A large sample does not automatically turn convenience sampling into probability sampling. In a thesis, transparency about sample limitations is more defensible than an unsupported claim of representativeness.

Frequently asked questions

What is a survey, and do I have to use a questionnaire?

In quantitative research, a survey commonly uses a structured questionnaire to convert responses into numerical data. Other data-collection formats are theoretically possible, but for a quantitative thesis about attitudes or behaviors, a questionnaire is a suitable and easily standardized instrument.

How many respondents do I need for a survey?

There is no single number that applies to every study. Compare the population size, the number of items, and the planned analysis method, such as the rule of 5 to 10 observations per item (Hair et al., 2010) or the finite-population formula (Yamane, 1967).

Calculate the number of valid responses you need, then allow for records that may be removed. This allowance does not replace data-quality checks.

Can I use an online survey for a thesis?

You can use one when the people you can reach fit the target population and you describe the distribution process transparently. State the collection channel, period, screening criteria, and limitations of the online sample.

How many items should a survey questionnaire have?

The number depends on the constructs and the items you need to measure. The questionnaire must represent the model adequately, but you should not add items simply to make it longer. A pilot helps you identify the actual completion time and the sections that make respondents tired.

What should I do if survey data are missing?

First check whether missingness affects only a few variables or nearly the entire questionnaire. Set the processing rule before analysis, preserve the original file, and record the number of excluded rows. Do not fill in answers based on guesses, because that creates data that did not come from a participant.

Can I use the same questionnaire for every study?

No. The questionnaire must match the study's model, context, respondents, and variable definitions. You can use the structure from survey questionnaire as a reference, but adjust the wording, screening criteria, and variable codes before collecting data.

Open your data file now, create a variable-code table, check the number of valid rows, and compare it with the sample size stated in your proposal. If you need help running the analysis on your own .sav or .csv file, DoThesis M4 analysis performs this step with SPSS and SmartPLS.