Introduction
Good research does not begin with a questionnaire, a dataset or a statistical model. It begins with a clear understanding of the problem that needs to be investigated.
Organisations invest considerable resources in research, monitoring, evaluation and data analysis to understand complex challenges and make better decisions. Yet even well-funded studies can produce findings that are technically sound but of limited practical value.
A survey may generate thousands of observations without explaining why a programme is underperforming. An evaluation may document activities and outputs without establishing whether meaningful change has occurred. A statistical model may identify associations without answering the decision-maker’s underlying question.
In many cases, the problem is not the method itself. It is that the method was selected before the problem was sufficiently understood.
Effective research requires more than choosing an appropriate technique. It requires defining the problem, understanding its context, clarifying what needs to be known and identifying how the resulting evidence will be used. These decisions shape the research design, determine the quality of the evidence and influence whether the findings can inform action.
1. The difference between a research topic and a research problem
Research often begins with a broad topic: youth employment, agricultural productivity, public health, programme effectiveness or access to public services. These topics are important, but they are not research problems in themselves.
A research problem identifies a specific gap in understanding, an unresolved issue or a decision that requires evidence.
Consider a development organisation implementing a youth employment programme. Its initial question might be:
How effective is the programme?
Although reasonable, this question is too broad to guide a complete research design. Effectiveness could refer to participation, skills development, job placement, employment retention, earnings or the distribution of benefits across different groups.
The organisation may actually need to understand why programme participation is high but employment outcomes remain low. Alternatively, it may need to establish whether participants achieve better employment outcomes than they would have achieved without the intervention.
These are different problems. They require different evidence and may call for different research designs.
The first step, therefore, is to move from a general topic to a clearly defined problem that identifies what is unknown, why the uncertainty matters and what the research needs to establish.
2. Start with the decision, not the data
One of the most useful questions to ask at the beginning of a research assignment is:
What decision will this evidence help someone make?
Research may be intended to inform whether a programme should continue, how resources should be allocated, which implementation barriers need attention or whether an intervention should be adapted or expanded.
Understanding the intended decision helps distinguish information that is merely interesting from evidence that is necessary.
For example, suppose a programme manager wants to understand why school attendance has declined in a particular area. Several explanations may be plausible: household financial pressures, distance to school, safety concerns, teacher absenteeism or changes in school-related costs.
A survey could describe attendance patterns and collect information about household circumstances. Interviews with parents, learners and teachers could help explain local experiences. Administrative records could reveal when and where attendance declined.
However, the appropriate combination depends on what the manager needs to decide. If the immediate concern is identifying where attendance has fallen, administrative data may provide a useful starting point. If the concern is understanding the reasons behind the decline, additional investigation may be necessary.
Starting with the decision does not mean designing research to produce a predetermined conclusion. It means ensuring that the investigation addresses a meaningful question while remaining open to findings that challenge initial assumptions.
3. Investigate the problem before committing to an explanation
A clearly stated research question can still be based on an incomplete understanding of the problem.
Organisations often approach research with an assumed explanation. A programme may be underperforming because staff need more training. Agricultural productivity may be low because farmers lack access to information. A service may be underused because communities are insufficiently aware of its availability.
These explanations may be correct, partly correct or entirely mistaken.
If the research design assumes the explanation in advance, it can overlook other important factors. A training intervention may not address poor programme performance if the underlying constraint is inadequate funding, weak supervision or an unrealistic implementation model.
Problem definition should therefore involve examining what is already known, identifying competing explanations and determining which uncertainties require further investigation.
Useful activities may include reviewing existing studies and administrative records, consulting relevant stakeholders, examining programme documentation and exploring the institutional, economic or social context.
This initial investigation does not always require a lengthy preliminary study. Its purpose is to establish a sufficiently grounded understanding of the problem to make informed choices about the main research.
4. Let the problem determine the method
Research methods are tools for answering questions. They should follow from the nature of the problem, the evidence required and the constraints under which the study will be conducted.
Different questions require different approaches.
Descriptive questions seek to establish what is happening, how frequently it occurs or how outcomes vary across populations. Surveys, administrative records and descriptive statistical analysis may be appropriate.
Explanatory questions seek to understand why a pattern or outcome occurs. Depending on the question, qualitative inquiry, statistical modelling, mixed-methods research or comparative analysis may help examine possible explanations.
Causal questions ask whether an intervention produced a change, rather than merely being associated with it. These questions require a credible strategy for estimating what would have happened in the absence of the intervention. Depending on context and feasibility, this may involve randomised designs or suitable quasi-experimental approaches.
Decision and resource-allocation questions examine which options are preferable under particular objectives and constraints. Economic analysis, cost-effectiveness assessment, budget analysis and structured comparison of alternatives may be useful.
These categories can overlap. A single assignment may require several methods to answer related questions.
The important principle is that methodological choices should be justified by the research problem rather than by familiarity with a particular tool, the availability of a preferred dataset or the assumption that more sophisticated techniques automatically produce better evidence.
5. Problem definition shapes evidence quality
Evidence quality is often associated with sample size, data accuracy, statistical significance or the sophistication of an analytical model. These considerations matter, but they cannot compensate for a poorly defined problem.
A large survey can estimate the wrong outcome precisely. A well-fitted model can answer a question that is irrelevant to the decision at hand. A carefully conducted interview study can provide rich descriptions without addressing the uncertainty the organisation needs to resolve.
A useful research design should establish a clear connection between the problem, research questions, concepts being measured, data collection methods, analytical strategy and intended use of the findings.
This alignment affects several aspects of quality.
First, it supports relevance by ensuring that the study investigates the issue that matters.
Second, it strengthens measurement by clarifying what needs to be observed and how key concepts should be defined.
Third, it improves analytical coherence by ensuring that the planned analysis can answer the questions being asked.
Finally, it strengthens interpretation by making the limits of the evidence clearer. Researchers can explain what the findings establish, which conclusions remain uncertain and what additional evidence may be required.
Research quality is therefore not simply a property of the method. It also depends on whether the entire investigation is logically connected to the problem it seeks to address.
6. A practical framework for defining a research problem
Before selecting a research method, researchers and commissioning organisations can work through six questions.
1. What is happening?
Describe the observed issue, pattern or uncertainty as precisely as the available information allows. Separate established facts from assumptions.
2. Why does it matter?
Identify the consequences of the problem, the people or institutions affected and the decisions that may be influenced by the findings.
3. What do we already know?
Review existing research, programme records, administrative data and relevant stakeholder knowledge. Identify where evidence agrees, conflicts or remains incomplete.
4. What do we still need to understand?
Specify the knowledge gap. Is the uncertainty about the scale of a problem, its causes, the effectiveness of an intervention, differences between groups or the relative merits of alternative options?
5. What decision must the evidence inform?
Clarify the intended users of the findings and the decisions they face. Consider what different possible findings would mean for those decisions.
6. What evidence would credibly answer the question?
Only now should the research team compare possible methods, data sources, analytical approaches, ethical requirements, resources and practical constraints.
This sequence is not a rigid formula. Research is often iterative, and early data exploration may reveal that the original problem needs to be reframed. The purpose is to make methodological choices deliberate and defensible.
7. From better questions to more useful research
Defining a research problem well does not guarantee that a study will produce definitive answers. Data may be incomplete, causal attribution may be difficult, stakeholder perspectives may differ and some uncertainties may remain unresolved.
However, a well-defined problem gives the research a stronger foundation. It helps researchers select appropriate methods, commissioning organisations specify realistic expectations and decision-makers understand how findings should be interpreted.
It also makes research more efficient. Rather than collecting data simply because it is available, teams can prioritise information that addresses the central uncertainty. Rather than producing reports that end with generic recommendations, they can connect findings to the decisions that prompted the investigation.
The most useful research is not necessarily the largest study or the one with the most advanced analytical technique. It is research whose questions, methods and interpretation are aligned with the problem that needs to be understood.
Conclusion: Define the problem before choosing the method
When research begins with a preferred tool, a familiar dataset or an assumed explanation, it risks producing technically competent work that does not resolve the underlying uncertainty.
A stronger approach begins by clarifying what is happening, why it matters, what remains unknown and which decisions depend on the answer. Methodological choices then follow from the evidence required to address those questions credibly.
For researchers, evaluators, programme leaders and policymakers, this discipline is fundamental to producing evidence that is not only rigorous, but also relevant and useful.
At DAREL Consulting, we believe that better evidence begins with better questions. By connecting research, evaluation, data analytics, economics and policy, we seek to help organisations make sense of complex problems and use evidence to inform better decisions.