03 · What You Need to Know
Why an Important Problem Can Still Be the Wrong Study
Importance, Feasibility, and Ethics Are Separate Tests
Researchers sometimes treat significance as though it overrides every other consideration. If a problem is important enough, they assume, there must be a way to study it.
That does not follow.
The widely used FINER framework evaluates a research question across several dimensions: feasible, interesting, novel, ethical, and relevant. Feasibility includes practical considerations such as participant availability, technical expertise, time, money, and manageable scope. A question can therefore be highly relevant while failing the feasibility or ethical test for a particular project.
This distinction is useful because it prevents two opposite errors. You should not dismiss an important problem merely because it is difficult, but you also should not use importance to justify research that cannot produce credible evidence.
Important problem
Resolving the uncertainty could make a meaningful difference to knowledge, decisions, practice, policy, methods, or affected people.
Researchable problem
The relevant uncertainty can be investigated credibly and ethically using evidence, methods, access, expertise, time, and resources that can realistically be obtained.
First Diagnose Why the Problem Is Not Researchable
“Not researchable” is too broad to guide your next move. Identify the actual constraint.
The problem may be:
- too broad for one study;
- too vague to determine what evidence would answer it;
- dependent on inaccessible data;
- dependent on participants you cannot recruit adequately;
- dependent on events too rare to observe with the available design;
- beyond your technical or disciplinary expertise;
- too expensive or time-consuming for the project;
- impossible to investigate using a design capable of supporting the intended inference;
- ethically unacceptable to investigate in the proposed way; or
- not currently measurable or observable with adequate validity.
These are different problems and require different responses. A question that is merely too broad may be narrowed. A question requiring expertise you lack may be solved through collaboration. An unethical experiment cannot be repaired simply by increasing the budget.
If the Problem Is Too Large, Narrow the Contribution
One of the most common forms of non-researchability is excessive scope.
You want to explain poverty, eliminate educational inequality, understand all causes of employee burnout, determine why misinformation spreads, or identify every factor responsible for unequal health outcomes.
These may be important problems, but no single study can provide the complete answer.
Keep the larger problem as the context and identify a consequential uncertainty inside it. Narrow by population, context, outcome, mechanism, process stage, intervention, or type of question when there is a substantive reason for doing so.
The goal is not to make the problem artificially small. It is to find the part that one study can investigate credibly. When scope is the main obstacle, the appropriate response is to define a study-sized part of the larger research problem.
If the Data Are Inaccessible, Ask Whether Another Evidence Source Can Answer the Question
Suppose your ideal study requires confidential administrative records that you cannot access. That does not automatically end the project.
Ask what the records were supposed to tell you. Could another dataset provide the relevant information? Could you obtain aggregated rather than individual-level data? Could prospective data collection answer a narrower version of the question? Could interviews, observations, public records, archival material, remote sensing, laboratory measurements, or another evidence source address part of the uncertainty?
The alternative must actually fit the inference.
If the original question requires objective longitudinal records, replacing them with participants' retrospective recollections merely because interviews are accessible may fundamentally change what the study can establish. That may still produce a worthwhile new question, but it is not the same study.
If the Population Is Inaccessible, Do Not Quietly Substitute a Convenient One
A similar problem occurs with recruitment.
Suppose your problem concerns a difficult-to-reach population, but you cannot recruit enough relevant participants. It may be tempting to study a convenient population instead and retain the original significance claims.
That can break the connection between problem and evidence.
A different population may be appropriate if there is a defensible reason it can answer the research question. Otherwise, the substitution changes the problem. You may need multiple recruitment sites, community partnerships, a longer recruitment period, another design, existing data, or a narrower question.
Feasibility should shape the study openly rather than being hidden inside sampling choices.
If the Necessary Expertise Is Missing, Collaboration May Be the Answer
Some research problems are not beyond investigation; they are beyond one researcher's capabilities.
A project may require advanced statistical modeling, specialized laboratory techniques, qualitative expertise, clinical knowledge, engineering, economics, community engagement, data security, or another competency that the original researcher does not possess.
The correct response is not necessarily to abandon the question or attempt the method without adequate expertise.
Collaboration can change what is feasible. So can methodological consultation, training, shared infrastructure, multicenter research, or access to specialized facilities.
This is an important distinction: not feasible for me alone is not the same as not researchable.
If the Required Study Is Unethical, Change the Question or Evidence
Some questions cannot ethically be answered through the most direct imaginable design.
You cannot deliberately expose people to serious harm merely because doing so would provide a clean causal test. You cannot disregard informed consent, privacy, equitable participant selection, or other protections because the research question is important.
The Belmont Report's principles for research involving human participants include respect for persons, beneficence, and justice. Its treatment of beneficence requires attention to both anticipated benefits and the probability and magnitude of possible harms, with risks reduced where possible and justified by the knowledge or benefits reasonably expected from the research.
This means that an important research question does not make any method ethically acceptable.
Instead, researchers may need observational evidence, natural experiments, existing records, simulations, animal or laboratory models where appropriate, retrospective designs, or other ethically acceptable approaches. The alternative may provide weaker evidence for some inferences, and that limitation should be acknowledged rather than hidden.
Scientific Value Matters Because Participants Should Not Bear Burdens for an Uninformative Study
Research ethics is not only about avoiding obvious physical harm. Poorly designed research can also be ethically problematic when participants accept burdens or risks but the study is incapable of producing useful knowledge.
CIOMS guidance for health-related research involving humans emphasizes both scientific and social value. Scientific value concerns the ability of research to generate reliable, valid information capable of achieving its stated objectives, while social value concerns the importance of the information a study is expected to produce.
This creates an important connection between researchability and ethics: if your design cannot credibly answer the question, the importance of the problem is not sufficient justification for exposing participants to avoidable burdens.
Watch Out
Do not use “this problem is extremely important” as an argument for proceeding with a study that cannot answer the question credibly or cannot satisfy applicable ethical requirements. The greater the claimed importance, the more important it is that the design can actually produce informative evidence.
If You Cannot Measure the Concept, Reconsider What You Are Claiming to Study
Some problems become unresearchable because the central concept is too vague or because the proposed measure does not represent it adequately.
Imagine wanting to study whether a program “empowers communities,” whether workers are “truly fulfilled,” or whether a policy creates “social harmony.” These ideas may be meaningful, but research requires a sufficiently clear conceptualization to determine what evidence would count.
You may need to define dimensions of the construct, use established measures, develop and validate a new measure, investigate people's interpretations qualitatively, or reformulate the question around something that can be observed credibly.
Do not operationalize the concept as whatever happens to be easiest to measure. A convenient variable does not become a valid representation simply because it exists in your dataset.
If the Desired Inference Is Too Strong, Weaken the Claim Rather Than Overstate the Evidence
Sometimes the data and design are adequate for one question but not for the question you originally wanted to answer.
Suppose you want to know whether workplace flexibility causes lower employee turnover, but you have access only to cross-sectional observational data. Those data may support useful descriptive or associational questions, but they may not provide a credible basis for the causal claim you originally wanted.
You then have choices. Find a design better suited to the causal question, or change the research question to match what the available evidence can support.
The wrong response is to retain the causal question and simply write stronger conclusions than the design warrants.
If the Event Is Rare, Change the Design Rather Than Pretending You Have More Information
Rare events create distinctive feasibility problems. A single site may observe too few cases for a precise estimate or meaningful comparison. Recruiting a conventional sample may take many years.
Possible responses depend on the question: combine sites, extend the observation period, use registries or administrative databases, adopt an appropriate case-control or other efficient design, synthesize existing evidence, or focus on a different but related question.
The correct strategy is field- and question-specific. The general principle is that rarity can require a different research architecture rather than making the underlying problem unimportant.
Sometimes You Need a Feasibility or Pilot Study First
Your immediate research question may be answerable in principle, but you may not yet know whether the study procedures will work.
Can the population be recruited? Will participants complete the measurements? Can an intervention be delivered as planned? Can data be collected at the required frequency? Is the measurement procedure acceptable?
A pilot or feasibility study can investigate these uncertainties before committing to a larger study. Guidance on research-question development specifically identifies pilot or proof-of-concept work as one possible response when feasibility is uncertain.
But keep the distinction clear: a pilot study should answer feasibility questions. It should not be presented as though a small preliminary study definitively answers the substantive question the later full study was designed to address.
Sometimes the Best Study Is a Prerequisite Study
A large question can depend on knowledge that does not yet exist.
Perhaps you want to evaluate an intervention, but no valid outcome measure exists. You want to test a mechanism, but the phenomenon has not yet been described adequately. You want to compare two policies, but basic implementation data are missing.
Instead of forcing the final study prematurely, investigate the prerequisite.
The first project might develop or validate a measure, characterize the population, estimate baseline rates, assess feasibility, map a process, or identify plausible mechanisms. That study may appear less ambitious, but it can create the conditions necessary for stronger research later.
Some Problems Require a Program of Research
Complex problems often cannot be reduced to one decisive study without losing what makes them important.
Research may need to proceed sequentially: establish the problem, understand experiences, identify mechanisms, develop measures, test an intervention, evaluate implementation, examine longer-term outcomes, and synthesize findings.
In such cases, asking one project to answer the whole problem is the mistake.
Your task is to identify the next informative study in the sequence. A useful question is not “How can my study solve the whole problem?” but “What does the evidence base need to know next before progress on the larger problem becomes possible?”
Researchability Can Change Over Time
A problem that cannot be investigated well today may become researchable later.
New measurement technologies can make previously inaccessible phenomena observable. New datasets may become available. Policy changes may create natural experiments. Larger collaborations can provide sufficient samples. Analytical methods can improve. New ethical approaches or privacy-preserving systems may enable forms of research that were previously impractical.
This means postponing a study is not equivalent to declaring the problem permanently unresearchable.
Sometimes intellectual discipline means recognizing that the right question is ahead of the available evidence.
Do Not Distort the Problem Merely to Make a Study Easy
Feasibility can also be taken too far.
Suppose the important problem concerns why vulnerable residents cannot access a public service. The relevant population is difficult to recruit, so the researcher surveys easily accessible university students about their opinions of public services instead.
The new project may be feasible, but it no longer answers the important problem.
A feasible study that is disconnected from the motivating problem is not necessarily preferable to an important question that requires a different research strategy.
The goal is to find the overlap between importance and researchability.
| What Makes the Problem Unresearchable? |
Possible Response |
What to Avoid |
| Scope is too large |
Narrow to a consequential uncertainty. |
Trying to answer several major problems superficially. |
| Required data are inaccessible |
Seek another valid evidence source, partnership, or narrower question. |
Using convenient data that cannot answer the intended question. |
| Population is difficult to recruit |
Use partnerships, multiple sites, longer recruitment, existing data, or another appropriate design. |
Substituting an irrelevant convenient population. |
| Required expertise is unavailable |
Collaborate, consult, train, or use appropriate shared infrastructure. |
Using methods you cannot implement competently. |
| Direct study would be unethical |
Use an ethical alternative design or change the question. |
Arguing that importance overrides participant protections. |
| Construct cannot be measured adequately |
Clarify, operationalize, validate, or study a prerequisite question. |
Calling an easy proxy the construct without justification. |
| Available design cannot support the desired inference |
Improve the design or reduce the strength of the research question. |
Making causal or broad claims unsupported by the evidence. |
Sometimes You Should Walk Away From the Proposed Study
Researchers invest emotionally in ideas. Once a problem feels important, abandoning a proposed project can feel like abandoning the problem itself.
They are not the same thing.
If no ethical, feasible, and scientifically credible version of the study can produce useful evidence, stopping is a legitimate research decision. You can preserve the problem as a future research priority, identify what capability would make it researchable, or choose another question where your current resources can make a stronger contribution.
Research restraint can prevent wasted resources, misleading conclusions, and unnecessary participant burden.
