A Step-by-Step Guide to Conducting a Survey in Research

Surveys are one of the most powerful tools in research, letting you collect data from large numbers of people in a structured and comparable way. But a survey is only as good as the planning behind it. A rushed questionnaire sent to the wrong people produces data that is unreliable at best and misleading at worst. Whether you are a student working on a dissertation or a researcher studying user behaviour in a library, the difference between a useful survey and a wasted one comes down to following a disciplined, step-by-step process. This guide walks you through that process, from defining your research problem to ethically approaching respondents, so the data you collect actually answers the question you set out to study.

Defining the research problem
Setting inclusion and exclusion criteria
Choosing the target population
From population to sample
Designing and pretesting the questionnaire
Why pretesting is non-negotiable
Pretest versus pilot test
Approaching respondents
Informed consent and transparency
Collecting accurate, unbiased data
Bringing the steps together

Defining the research problem

Every survey begins with a clearly stated research problem. This is the foundation on which everything else rests. If you cannot articulate exactly what you want to find out, you cannot design questions to find it out. A vague problem like “I want to study reading habits” leads to a scattered, unfocused questionnaire. A sharp problem like “I want to measure how often postgraduate students use e-resources versus print resources in university libraries” tells you precisely what data to collect and from whom.

Start by writing your problem as a specific, answerable question. From there you derive your research objectives, the concrete actions that will help you answer it. The objective shapes the entire study: it defines who you survey, what you ask, and how you interpret the answers. As one widely used framework puts it, identifying the goal first guides every step that follows, including finding the right audience and designing appropriate questions.

Setting inclusion and exclusion criteria

Once your problem is defined, you need to decide who counts as a valid participant. This is where inclusion and exclusion criteria come in. These are pre-set rules that determine which members of your target population can or cannot take part in your study. Inclusion criteria are the characteristics a person must have to be eligible, while exclusion criteria are the features that disqualify them.

For example, if you are studying e-resource usage among postgraduate students, your inclusion criteria might be enrolment in a postgraduate programme and active library membership. An exclusion criterion might be students on leave who have not accessed the library in the past six months. The key practice is to write these criteria before you begin recruiting. Defining them in advance prevents you from changing the rules midway, which would introduce bias and weaken your findings. Typical criteria include demographic characteristics such as age, gender, education, occupation, and geographic location, alongside variables specific to your topic.

Clear criteria do two things. They ensure your sample genuinely matches your research aim, and they make your selection process consistent and defensible. A guide from a public-health research team notes that well-defined criteria increase the likelihood of producing reliable and reproducible results while reducing the risk of selecting people whose characteristics are unrelated to the research question.

Choosing the target population

The target population is the complete group of people your research question is about. If your study concerns library usage by postgraduate students in a particular university, then all postgraduate students at that university form your target population. Identifying this group precisely matters, because every later decision about sampling and data collection flows from it.

Two qualities make a target population workable. First, it must be relevant, meaning the people in it can actually provide the information your research problem requires. Surveying undergraduate students about postgraduate research behaviour would produce irrelevant data. Second, it must be measurable and reachable. You need a clear, practical method for identifying and contacting the individuals in this population. A population you cannot access is useless, no matter how relevant it is on paper.

From population to sample

In most studies you cannot survey every single person in your target population, especially when it numbers in the thousands. Instead you select a smaller subset, called a sample, that represents the larger group. The quality of your sampling determines whether your conclusions can be generalised back to the full population. A representative sample reflects the key characteristics of the population in roughly the same proportions, so the patterns you observe in your sample are likely to hold true for everyone.

This is also where you need to watch for selection bias. If your sampling method systematically over-represents or under-represents certain groups, your results will be skewed. One large study found that even with a high consent rate, the people who agreed to participate differed significantly from those who refused in age, education, and other characteristics, meaning a high response rate alone does not eliminate the possibility of selection bias. Being aware of who is likely to respond, and who is likely to be left out, helps you interpret your data honestly.

Designing and pretesting the questionnaire

The questionnaire is the instrument that turns your research objectives into actual data. Designing it well is part science and part craft. Each question should map directly back to an objective; if a question does not help answer your research problem, it does not belong in the survey. Keep the language simple, avoid jargon, and make sure every question asks about only one thing at a time. Response options for closed questions should be relevant, comprehensive, and mutually exclusive, so respondents always find an option that fits and never feel forced between two that both apply.

One of the most common and damaging mistakes is the leading question, which is phrased in a way that nudges respondents toward a particular answer. A question like “How much do you appreciate the library’s excellent new digital catalogue?” assumes a positive view before the respondent has expressed one. Neutral, balanced wording produces honest data; loaded wording produces flattering but useless data.

Why pretesting is non-negotiable

No matter how carefully you write your questions, you cannot be sure they work until you test them on real people. This is the purpose of pretesting. Pretesting is the process of evaluating your questionnaire and survey procedures in advance to find out whether they will cause problems for respondents before you launch the full study. As an extension-research guide explains, no one writes perfect questions on the first attempt, and a survey designed for a specific population needs input from that population to confirm the questions capture what you intended.

Pretesting reveals problems you simply cannot see yourself: ambiguous wording, confusing instructions, unfamiliar terminology, questions that take too long, or a survey flow that loses people halfway through. A practical rule of thumb is to test your survey with at least five people from your target group, or as close to that group as you can find. Even this small number surfaces a surprising number of improvements. Some researchers test on around five to ten percent of the total target audience before going wider.

Pretest versus pilot test

It helps to distinguish two related stages. A pretest is a small, often informal check of the questions themselves, focusing on wording, clarity, format, and logic. A pilot test is a larger dress rehearsal of the entire survey procedure, run under conditions as close to the real thing as possible, with the same mode of delivery and instructions. Pilots tend to surface practical problems with implementation, such as logistics of distribution, data-entry errors, or response rates, testing the feasibility and logistics rather than just the wording. After each round, analyse the feedback, prioritise the issues, revise, and if the changes are significant, test again. This iterative loop is what transforms a rough draft into a reliable instrument.

Approaching respondents

With a tested questionnaire ready, the final step is reaching your respondents and collecting data tactfully and ethically. How you approach people directly affects both your response rate and the honesty of their answers. The goal is to make participation feel voluntary, respectful, and worthwhile.

Before anyone answers a single question, they should understand what they are agreeing to. This principle is called informed consent, and it sits at the core of ethical research. Respondents should be told the purpose of the survey, how their data will be used, who is conducting it, and any potential risks or benefits. In many survey studies, completing and submitting the questionnaire after reading a clear information statement can itself constitute informed consent, though sensitive topics or vulnerable groups may require explicit written permission. Note that you generally cannot collect data from minors without permission from a parent or guardian.

Equally important are the assurances you give. If you promise confidentiality or anonymity, you must honour it. That means storing data securely and keeping personally identifiable information separate from survey responses. Any promise of confidentiality, once made, must be kept by the researcher, and participants should never feel pressured, cajoled, or coerced into taking part.

Collecting accurate, unbiased data

Ethical approach and data quality are closely linked. People give more honest answers when they trust the process and feel their privacy is protected. To encourage truthful responses, give clear instructions, use neutral language throughout, and reassure respondents that there are no right or wrong answers. Watch out for social desirability bias, the tendency for people to answer in ways that make them look good rather than telling the truth. Self-administered and anonymous surveys, where responses are returned privately rather than handed directly to the researcher, help reduce social desirability bias and responder bias.

Finally, guard against bias in the survey flow itself. The order in which questions appear can influence answers, so think carefully about sequencing. Professional bodies that set standards for survey research are explicit that researchers must guard against bias in survey flow and wording to avoid skewing data. When your wording is neutral, your sample is representative, and your respondents are treated with respect, the data you collect will genuinely reflect reality rather than the artefacts of a flawed process.

Bringing the steps together

Conducting a survey is a sequence of connected decisions, and each step depends on the one before it. A well-defined research problem with clear inclusion and exclusion criteria tells you who your target population is. A properly identified target population lets you draw a representative sample. A representative sample is only useful if you collect data through a questionnaire that has been carefully designed and pretested. And even the best questionnaire produces poor data if respondents are approached carelessly or unethically. Treat the process as a chain, where the weakest link determines the strength of the whole, and your survey will yield data that is reliable, valid, and genuinely useful for answering your research question.

What do you think? If you were designing a survey on library usage at your own institution, what inclusion and exclusion criteria would you set, and why? And which single step in this process do you think researchers most often rush or skip, with the greatest cost to their results?

How useful was this post?

Click on a star to rate it!

Average rating 4.8 / 5. Vote count: 4

No votes so far! Be the first to rate this post.

References

PDF 📄

Table of Contents

Defining the research problem 🔗

Setting inclusion and exclusion criteria 🔗

Choosing the target population 🔗

From population to sample 🔗

Designing and pretesting the questionnaire 🔗

Why pretesting is non-negotiable 🔗

Pretest versus pilot test 🔗

Approaching respondents 🔗

Informed consent and transparency 🔗

Collecting accurate, unbiased data 🔗

Bringing the steps together 🔗