Do not begin by choosing a survey, interview, statistical test, or software package. Begin with the research question. A defensible method is one that can generate appropriate evidence for answering that question.
Start With the Research Question
Your research question defines what the study is trying to investigate. The method should follow from that objective.
Compare these questions:
Exploring an Experience
How do remote employees describe the challenges they experience when collaborating with their teams?
This question asks about experiences and perspectives. Interviews or another suitable qualitative approach may provide relevant evidence.
Examining a Relationship
Is weekly remote-working time associated with employee job satisfaction score?
This question involves measurable variables and a relationship that could be investigated quantitatively.
If the research question changes substantially, the appropriate method may change with it.
Identify What the Study Is Trying to Do
Research questions can serve different purposes. Identifying that purpose helps narrow the methodological options.
| Purpose | What You May Be Trying to Understand |
|---|---|
| Explore | Experiences, perspectives, processes, or a topic that is not yet well understood |
| Describe | Characteristics, frequencies, patterns, or conditions |
| Compare | Differences between groups, conditions, or time points |
| Examine relationships | Whether variables are associated and how those relationships behave |
| Estimate effects | How an intervention, exposure, or condition relates to an outcome under a suitable design |
| Predict | Future or unknown outcomes from available information |
A study may have more than one objective, but each objective should connect logically to the evidence and analysis used.
Decide What Kind of Evidence You Need
Once the question is clear, ask what evidence would actually allow you to answer it.
Experiences & Meanings
Detailed accounts, conversations, observations, documents, or other context-rich evidence may be needed.
Measurements
Numerical variables may be needed when the question involves quantities, comparisons, relationships, or prediction.
Existing Evidence
Published research, existing datasets, records, or documents may already contain relevant information.
This step prevents a common problem: collecting convenient data first and only later trying to create a research question that fits it.
Choose a Broad Research Approach
At this stage, you can consider whether the question is best addressed through a qualitative, quantitative, or mixed methods approach.
| Approach | May Be Appropriate When... |
|---|---|
| Qualitative | You need to explore experiences, meanings, perspectives, context, or processes in depth |
| Quantitative | You need to measure variables, estimate quantities, compare groups, examine relationships, or model outcomes |
| Mixed Methods | Both qualitative and quantitative evidence are needed and their integration contributes to answering the research question |
Being comfortable with SPSS, interviews, Python, surveys, or another tool is useful, but familiarity alone does not make that method appropriate for the research question.
Will You Collect New Data or Use Existing Evidence?
Next, consider where the evidence will come from.
If suitable information already exists, secondary research may be able to address the question without collecting entirely new data.
If the necessary information does not exist, is inaccessible, or does not match the research question closely enough, primary data collection may be needed.
Does suitable data already exist?
Does it cover the population or setting I need?
Are the required variables or concepts available?
Is the existing evidence current enough for my question?
If not, can I realistically collect the evidence myself?
Choose a Data Collection Method
If you need new data, select a method capable of producing the evidence required by the research question.
| Method | May Be Useful For |
|---|---|
| Interviews | Detailed individual experiences, perspectives, and explanations |
| Focus groups | Group perspectives and interaction around a topic |
| Structured surveys | Collecting standardized responses across participants |
| Observation | Studying behavior, activities, settings, or processes |
| Experiment | Studying outcomes under controlled or manipulated conditions |
| Direct measurement | Generating physical, biological, behavioral, or technical measurements |
These methods can take many forms. The details of the instrument and procedure should be justified by the study design rather than selected from a template without adaptation.
Think About Who or What You Need to Study
A method cannot be evaluated properly without considering the sample.
Ask questions such as:
- What is the population or group of interest?
- Who or what should be included?
- Who or what should be excluded?
- How will participants or cases be identified?
- How many observations are realistically needed?
- Does the sampling strategy support the conclusions I want to draw?
Sample size should not be selected using one universal rule. The appropriate reasoning depends on the methodology, study design, analysis, variability, effect sizes or precision needs where relevant, and practical constraints.
Think About the Analysis Before Collecting Data
Analysis should not be an afterthought.
Before collecting data, consider how the evidence will be used to answer the research question.
Comparing Two Groups
If the objective is to compare a numerical outcome between two groups, you should know how that outcome and the group variable will be measured before collecting the data.
The eventual statistical method will depend on the design, data structure, variable types, and relevant assumptions.
Exploring Professional Experiences
If the objective is to understand how professionals experience a particular workplace change, the interview protocol and qualitative analysis approach should both align with that purpose.
Planning the analysis early can reveal that important variables, questions, measurements, or contextual information are missing from the proposed data collection.
Check Whether the Method Is Actually Feasible
A method can be theoretically appropriate but unrealistic for the resources available.
Consider:
- the project deadline;
- access to participants;
- access to datasets or records;
- equipment and software;
- researcher skills and training;
- financial resources;
- participant recruitment;
- data processing time; and
- institutional requirements.
An Unrealistic Proposal
A student has six weeks to complete a research project but proposes recruiting hundreds of participants from several countries, conducting interviews, administering a survey, and running a longitudinal follow-up.
Even if those methods could theoretically produce useful evidence, the design may not be realistic within the available time and resources.
Consider Ethics Before Collecting Data
Research methods must also be ethically and institutionally appropriate.
Depending on the study, issues may include:
- informed consent;
- participant privacy;
- confidentiality;
- sensitive questions or information;
- data storage and access;
- risks to participants;
- research involving vulnerable populations;
- use of existing personal or restricted data; and
- institutional ethics or review requirements.
If your institution requires ethics approval or another form of authorization, obtain the required approval before beginning activities that must be approved in advance.
Three Practical Method Selection Examples
Understanding AI Adoption
Question: How do small-business owners describe the barriers they experience when adopting AI tools?
Because the question focuses on experiences and perceived barriers, semi-structured interviews followed by an appropriate qualitative analysis could be considered.
Examining Customer Satisfaction
Question: Is service waiting time associated with customer satisfaction score?
This involves two measurable variables. A quantitative observational design could collect or use existing measurements and examine their association using a suitable statistical method.
Evaluating a Training Program
Question: How did employee performance change following a training program, and how did employees experience the program?
The first part requires measurable performance evidence, while the second asks about participant experiences. A mixed methods design may be worth considering if integrating both forms of evidence is necessary to answer the overall question.
These examples identify possible approaches, not automatic designs. The final method still depends on sampling, access, measurement, timing, ethics, and the conclusions the study intends to support.
A Practical Research Method Decision Workflow
Define the research problem.
Write a focused research question.
Identify what evidence is needed to answer it.
Decide whether the study needs qualitative, quantitative, or integrated evidence.
Check whether suitable existing evidence is available.
If collecting new data, choose an appropriate collection method.
Define the population and sampling approach.
Plan how the resulting data will be analyzed.
Check feasibility and available resources.
Address ethical and institutional requirements.
Confirm that every major methodological choice connects back to the research question.
Common Mistakes When Choosing a Research Method
1. Choosing software before choosing the method
SPSS, Python, Excel, R, NVivo, and other tools can support analysis, but software should not determine the research design.
2. Choosing a method because it seems easier
A convenient method that cannot answer the research question creates a weak study regardless of how quickly the data can be collected.
3. Using a questionnaire for every project
Surveys are useful for many questions, but they are not a universal research method.
4. Choosing the statistical test before defining the variables
The analysis should follow from the research question, design, variable types, data structure, and relevant assumptions.
5. Ignoring how the sample will be obtained
A proposed method is incomplete if there is no realistic plan for obtaining the participants, observations, or data it requires.
6. Collecting data without planning the analysis
This can leave you with data that does not contain the variables, detail, or structure required to answer the research question.
7. Making conclusions the design cannot support
For example, a cross-sectional observational association does not by itself establish that one variable caused another.
8. Ignoring feasibility until the project has started
Recruitment difficulties, limited time, inaccessible data, or unavailable equipment can make an otherwise reasonable design impossible to complete.
Final Research Method Checklist
A strong research method is not the most complicated one. It is the method that produces appropriate evidence for the research question while remaining methodologically sound, ethical, and feasible.
Need Help Choosing an Appropriate Method?
Share your topic, research question, project requirements, available data, proposed sample, and any method you are considering. The options can then be worked through in the context of your actual research project.