Key information
- ✓Match each question with a research problem or a necessary sample characteristic.
- ✓Avoid suggestive questions, doubles and imprecise times.
- ✓Use consistent response scales and describe all their points.
- ✓Conduct a pilot before starting the actual data collection.
Start with a table of research questions
Enter your research questions in the first column, the information you need in the second, and specific survey questions in the third. This makes it easy to spot redundant elements and problems for which data is missing.
Before writing the questionnaire, plan basic tables and charts. If you don't know how you will use the answer, the question probably needs to be changed or deleted.
Closed, open and semi-open questions
Closed questions make it easier to compare answers, but the list of options must be complete and disjoint. The "other, what?" option helps when it is impossible to predict all possibilities.
Open-ended questions provide free expression, but require coding and are more likely to be omitted. Use them sparingly where you really need content that cannot be captured in a ready-made scale.
How to build clear scales?
The scale should have clear endpoints and stay in the same direction throughout the survey. A sudden change from "1 - definitely not" to "1 - definitely yes" increases the risk of accidental responses.
Don't combine two things in one statement, such as quality and speed of service. The respondent may have different opinions about each of them, and a single answer will become impossible to interpret.
Imprint, anonymity and order
Specific questions should cover only the features needed in the analysis. Too detailed a combination of age, position and location may enable the person to be recognized even without asking for the name.
Start with simple and topic-related questions, place the more difficult ones in the middle, and usually put the information at the end. In the introduction, state the purpose of the study, the expected time and how the answers will be used.
Pilot and test the form
Ask a few people who are similar to your future respondents to complete the survey and describe any areas that are unclear. Time yourself, check skipped questions, and whether answer options cover common situations.
Test transition rules, mandatory questions, data export and mobile view. Once you have launched the actual survey, do not change the content of the questions unnecessarily, as different versions make it difficult to compare the results.
Build the questionnaire from the analysis plan
Before drafting items, decide which summaries will answer the research problems. For each research question, identify the variables needed and how to measure them. If you plan group comparisons, collect information that distinguishes the groups, but do not add demographic questions without a purpose. Every item increases participant burden. Retain questionnaire elements whose use you can justify in relation to the project.
Educational example: assessing lecturer feedback may require separating timeliness, clarity and usefulness. One general satisfaction question will not show which aspect causes difficulty. Equally, many nearly identical items do not automatically form a good scale. If using an existing instrument, check its intended application and conditions of use. Describe your own question set according to its actual status.
One question should address one issue
A question about a quick and helpful response combines two characteristics. A participant may have received a quick but unhelpful answer and find no appropriate option. Split this into two items. Remove assumptions participants need not share and suggestions pointing towards a preferred response. Neutral language helps collect material reflecting participants' experience. Avoid jargon, double negatives and unclear time references.
Instead of asking about frequent use, specify a period and a response format appropriate to the problem. For numerical categories, check that ranges do not overlap and cover possible results. Explain how many answers may be selected in a multiple-response question. Do not know, not applicable and prefer not to answer mean different things. Do not automatically treat them as the scale midpoint or a negative response during analysis.
Order, conditional navigation and form length
Begin with brief information about the study's purpose and participation conditions, consistent with university rules. Follow with easy questions and introduce more demanding ones after establishing the context. Group items addressing the same issue. Position demographic questions so they do not disrupt the flow. Do not collect identifying information merely because the form allows it. The information requested should arise from the project's needs.
Test every form pathway. Someone who has not used a service should not be forced to evaluate it in detail. On a phone, check long labels, scale readability and button operation. Estimate the stated completion time from a pilot rather than the author's preference. If the form is too long, remove questions unrelated to the analysis. Shortening instructions at the expense of understanding usually does not solve the problem.
A pilot that checks how questions are understood
Ask people resembling the intended participants to complete a trial version and identify difficulties. What matters is not only whether the form can be submitted but also how individual concepts are understood. Ask about missing response options, similar questions and situations where choices felt forced. Record revisions and their rationale. Piloting helps uncover errors but does not by itself guarantee the entire instrument's validity.
Before the main study, perform a trial export. Check column names, multiple-response recording, empty answers and character encoding. Create a variable dictionary containing names, question wording, permitted values and missing-data rules. Survey analysis in Excel or statistical software will then rest on clear conventions. Do not change category meanings after seeing results simply because another coding scheme produces a more attractive comparison.
Describe the survey in the methodology and appendices
Explain the questionnaire's structure, the relationship between its sections and the research problems, distribution method and response-collection period. Describe recruitment, eligibility and handling of incomplete forms and duplicates. Link views are not respondent counts. An open link in a social-media group does not automatically produce a random sample. Conclusions must account for recruitment and possible self-selection.
Include the version actually used in the appendix, with instructions and response options. If it contained conditional navigation, explain the pathways so readers can understand the process. Separate test responses from real study data. Do not invent respondents or answers to reach a desired sample size. It is better to describe the material's limitations and agree next steps with your supervisor than to create a falsely complete dataset.
Summary
The value of a survey depends not on the number of questions, but on their relationship to the purpose of the study and the quality of the answers obtained. First plan the analysis, then select the indicators and only then write the questionnaire based on this. A short pilot allows you to detect ambiguities that the author usually does not notice. Thanks to this, a proper study provides data that can be reliably described and interpreted.
Frequently asked questions
How many questions should a work survey have?
The questionnaire should be as short as possible and as long as the purpose of the study requires. Most often, what is more important than the number is the expected pace of completion and the lack of unnecessary questions.
How many answers are needed for a diploma thesis?
It depends on the population, sample selection and planned analysis. The number should be agreed with the promoter and justified methodologically, instead of adopting a random threshold.
Can the survey be prepared in Google Forms?
Yes, if the tool meets data protection requirements and allows you to configure questions correctly. Before publishing, you need to check the form on your phone and computer.
Should every survey question be compulsory?
No. Mandatory responses need justification and must account for cases where a question does not apply. Forced answers can distort data or cause participants to leave. Align voluntary-participation rules and handling of sensitive information with university procedures.
Need support with your work?
Describe the topic, scope and deadline. We will review the materials, recommend the right support and prepare a clear quote.
