Your Research Question and Methods Do Not Match. Here Is How to Fix It
A study is aligned when its research question, objectives, design, population, sampling, measurement, analysis and conclusion all support the same inferential goal. A mismatch occurs when the chosen methods produce a different kind of evidence from the one the question requires. Fix the problem by tracing the study from question to conclusion and checking whether every methodological decision preserves what you are actually trying to learn.
You can have:
- a validated questionnaire;
- a respectable sample size;
- excellent statistical software;
- a beautifully formatted proposal.
And still design a study that cannot answer your research question.
The method is not automatically bad.
It may simply be answering a different question.
What does research alignment mean?
A useful alignment chain is:
Question → objective → design → population → sampling → measurement → analysis → conclusion
Every arrow should make sense.
For example:
If the question asks how people experience something, the method must generate data about experience.
If the question asks what proportion of a district population has an outcome, the sampling must support that population estimate.
If the question asks whether an intervention reduces an outcome, the design needs a defensible comparison and temporal structure.
Alignment sounds obvious.
Proposal reviews suggest it frequently is not.
Mismatch 1: asking “why” with a closed list of answers
Research question:
Why do adolescents avoid the clinic?
Method:
A closed questionnaire asking respondents to tick:
- distance;
- cost;
- confidentiality;
- provider attitude;
- lack of knowledge;
- parental disapproval.
What can this study answer?
It can estimate how frequently respondents endorse the six explanations you selected in advance.
It may test associations between those responses and clinic use.
What can it not necessarily answer?
Why adolescents avoid the clinic in the broader sense.
You have restricted the explanatory universe before asking participants.
Fix
Either narrow the question:
What proportion of adolescents report selected barriers to clinic use, and which barriers are associated with non-use?
Or change the method to include an inquiry capable of identifying unanticipated mechanisms, such as qualitative interviews.
The correct fix depends on the real research purpose.
Mismatch 2: causal wording with cross-sectional timing
Research question:
Does exposure to the programme reduce unintended pregnancy?
Method:
One cross-sectional survey measuring:
- current programme exposure;
- lifetime or previous pregnancy history.
The timing is problematic.
Some pregnancy outcomes may have occurred before the current exposure.
The direction of influence may be unclear.
Your temporal ordering is now doing gymnastics.
Fix
First clarify the estimand.
If you genuinely need a causal effect of programme exposure, consider what design, comparison and temporal ordering are needed to support that inference.
If only cross-sectional data are feasible, narrow the question:
Is current programme exposure associated with [clearly timed outcome]?
The conclusion must match the design.
Mismatch 3: prevalence question, convenience sample
Research question:
What proportion of health facilities in the district offer implant removal?
Method:
Interview managers from four conveniently selected facilities.
Those interviews may generate rich operational information.
They cannot support a district-wide prevalence estimate.
Fix
If prevalence is the goal, define the facility sampling frame and select facilities using a strategy capable of supporting the intended estimate.
If depth is the goal, change the question:
How do managers in selected facilities describe operational barriers to providing implant-removal services?
Both are valuable questions.
They are not the same question.
Mismatch 4: qualitative objective, quantitative conclusion
Objective:
To explore women's experiences of contraceptive side effects.
Data:
In-depth interviews.
Conclusion:
“Seventy percent of women in the district experience severe side effects.”
That is not what a purposive qualitative sample was designed to estimate.
Qualitative data can show:
- types of experiences;
- variation;
- meaning;
- processes;
- consequences;
- decision-making.
The sample usually does not provide a basis for district prevalence estimates.
Fix
Keep the qualitative conclusion aligned with the evidence:
Participants described side effects as influencing continuation through concerns about health, daily functioning and partner relationships, although experiences varied substantially.
If prevalence matters, design a quantitative component capable of estimating it.
Mismatch 5: objective says “compare,” analysis only describes
Objective:
To compare contraceptive knowledge between first-year and final-year students.
Results:
- Overall mean knowledge score: 7.4.
- Overall percentage with “good knowledge”: 62%.
Where is the comparison?
The study may have collected the right data.
The analysis simply did not answer the objective.
Fix
Build the analysis plan before data collection.
For every objective, state:
- outcome;
- exposure/grouping variable;
- descriptive estimate;
- comparison or model;
- uncertainty;
- any adjustment.
The results section should then mirror the objectives.
A method is appropriate only in relation to the question it is expected to answer.
There is no universally “strong” design floating independently of the research problem.
The Methods Bench alignment test
Take your study through seven checks.
1. Question
What exactly are you trying to know?
Not the topic.
The question.
Bad:
Family planning among university students.
Better:
What proportion of students used a modern contraceptive method during the previous six months?
Or:
How do students describe confidentiality concerns when deciding whether to seek sexual and reproductive health services?
Those questions demand different evidence.
2. Objective
Does the objective preserve the question?
Question:
How do students describe confidentiality concerns?
Objective:
To determine the prevalence of contraceptive use.
That objective has wandered into another study.
Your objectives should break the question into answerable components, not replace it.
3. Design
Can the design produce the type of evidence required?
If you need prevalence, you need a design and sample that support estimation.
If you need change over time, time has to enter the design.
If you need experience, the method must give participants space to describe it.
If you need a causal effect, the design must address more than statistical association.
4. Population and sampling
Who needs to be represented or heard?
If the question refers to:
all district facilities
but the sample covers only one private hospital and three convenient clinics, the population and sampling do not support the scope of the question.
If the question concerns:
experiences of women who discontinued a method because of side effects
then purposively selecting women with that specific experience may be exactly right.
5. Measurement
Do your variables or interview questions actually observe the concept?
Suppose your objective refers to:
quality of care
but your questionnaire measures only:
- waiting time;
- facility cleanliness.
Those may be dimensions of care.
They may not adequately represent the broader construct.
Concepts need operational definitions.
Qualitative studies have the same issue.
An interview guide about “satisfaction” may not adequately explore “agency.”
6. Analysis
Will the planned analysis answer the objective?
Objective:
Estimate prevalence.
Analysis:
Report a mean score.
Mismatch.
Objective:
Compare two groups.
Analysis:
Report only an overall percentage.
Mismatch.
Objective:
Explore decision-making.
Analysis:
Count how many participants mentioned each code and rank the codes.
Possibly a mismatch, depending on the qualitative methodology and question.
7. Conclusion
Does the conclusion stay inside the evidence?
Cross-sectional association:
“X was associated with Y.”
Potentially appropriate.
Cross-sectional association:
“X caused Y.”
Usually much harder to defend.
Purposive qualitative sample:
“Participants described...”
Appropriate.
Purposive qualitative sample:
“80% of the national population believes...”
Not supported.
The conclusion is part of the design.
It should not suddenly become more ambitious than everything before it.
one complete alignment chain
Research problem:
University students may have access to SRH services but still avoid them because seeking care is perceived as socially unsafe.
Research question:
How do university students describe confidentiality concerns when deciding whether to seek SRH services?
Objective:
To explore how students describe confidentiality concerns and how these concerns influence decisions about seeking care.
Design:
Qualitative study using in-depth interviews.
Population:
Students with relevant experience navigating or considering university-linked SRH services.
Sampling:
Purposive selection to capture variation in gender, year of study and service-use experience.
Measurement/data collection:
Open-ended interviews exploring confidentiality, recognition, provider interactions, social visibility and decision-making.
Analysis:
Qualitative thematic analysis appropriate to the chosen methodology.
Conclusion:
Describes the ways confidentiality concerns operate in students' decision-making.
Every arrow is doing approximately the same job.
That is alignment.
Bench Check: choosing a method first
You say:
“I want to do a mixed-methods study. What question should I use?”
Wrong order.
A method is not a research topic.
Start with the problem and question.
Then ask whether one kind of evidence is sufficient.
Mixed methods is appropriate when the question genuinely benefits from integrated quantitative and qualitative evidence.
It is not methodological seasoning.
What do I actually write in a proposal?
A useful design-justification paragraph might say:
The study will use a cross-sectional survey because the primary objective is to estimate current service utilization and examine associations with participant and service characteristics during the same study period. The design is not intended to establish causal effects.
Or:
A qualitative design using in-depth interviews was selected because the study seeks to understand how participants experience and interpret confidentiality concerns during decisions about seeking care. Purposive sampling will be used to recruit participants with direct experience of the phenomenon.
Or:
A sequential mixed-methods design will first estimate the prevalence and distribution of service use using a survey, followed by qualitative interviews to investigate mechanisms underlying the quantitative patterns.
The justification should explain why the design fits the question.
Not simply name the design.
The red-flag phrases
Check your protocol if you see:
“A cross-sectional design was used because it is quick and cheap.”
That explains feasibility.
Not methodological fit.
“Purposive sampling was used because the researcher chose participants.”
That describes the word purposive.
Not the sampling rationale.
“Mixed methods was chosen to obtain comprehensive data.”
Comprehensive how?
For which parts of the question?
“Regression will be used to determine the causes of...”
A regression model does not automatically establish causation.
Draw this chain on one page:
QUESTION
↓
OBJECTIVES
↓
DESIGN
↓
POPULATION + SAMPLE
↓
MEASUREMENT
↓
ANALYSIS
↓
CONCLUSION
Then write one sentence in every box.
Read from top to bottom.
Where do you have to explain away a jump?
That is probably where the study needs revision.
Frequently asked questions
- Should the research question come before the methodology?
Usually yes. The question and problem should determine what evidence is needed, which then guides methodological choices.
- Can I change my research question after choosing the method?
Research develops iteratively, especially early in proposal development. But do not distort the question merely to defend a method you had already decided to use.
- How do I know if my objectives match my analysis?
For each objective, write the exact data and analysis needed to answer it. If you cannot do that, alignment is incomplete.
- Can one study have quantitative and qualitative objectives?
Yes, particularly in mixed-methods research, if the components fit the overall question and there is a coherent plan for integration.
- Does a more complex method make a study stronger?
No. Complexity is not rigor. The strongest design is one that answers the question credibly and feasibly.
This article should become the foundation for a future Methods Bench Study Alignment Checker.
For now, use the Study Design Decision Guide and the Objective-to-Analysis Table from Day 14 together.
References and further reading
- 1.World Health Organization. Recommended format for a research protocol.
- 2.World Health Organization. A Practical Guide for Health Researchers.
- 3.CDC Field Epidemiology Manual. Guidance on designing analytic studies and matching design to the epidemiologic question.
- 4.STROBE Statement. Guidance for reporting major observational study designs. --- # Cross-linking instructions for Lovable, Batch 3
Take it further
Better research, once a week.
Practical methods guidance, research tools and funding opportunities from Methods Bench.
Get practical research notes and new opportunities
Short, useful emails for researchers in Uganda and East Africa. No spam.