Which Uncertainties Must Be Resolved Before the Study Starts, and Which Can Remain Open?

Which unknowns are acceptable when a study begins?

Research rarely starts from complete certainty. Even a carefully planned project contains unknowns about recruitment, data quality, participant behavior, implementation, unexpected findings, and practical complications. If you waited until every uncertainty disappeared, many studies would never begin.

Yet some uncertainties cannot safely be carried forward. If you have not settled what your primary outcome means, whether you can ethically collect the required data, or whether the design can answer the research question, beginning the study may lock you into a problem that becomes difficult or impossible to repair.

The practical challenge is therefore not to eliminate uncertainty. It is to distinguish uncertainty that threatens the integrity of the study from uncertainty that can legitimately remain open and be resolved as the research develops.

Resolve uncertainties that could invalidate or fundamentally redirect the study

The real issue is the consequence of being wrong

An unresolved question is not automatically a flaw. The more useful question is: what happens if your current assumption turns out to be wrong?

Suppose you are uncertain whether recruitment will yield 80 or 100 participants. That may be manageable if both numbers support the planned study and you have appropriate recruitment monitoring. Now suppose you are uncertain whether the dataset contains the variable needed to measure your primary outcome. That uncertainty is qualitatively different. If the variable is absent or unusable, the intended study may no longer exist in the form you planned.

This distinction can be understood in terms of dependency and reversibility. Some unknowns sit upstream of many later decisions. Others concern details that can change without altering the study's underlying logic.

Resolve uncertainties that determine what question you are actually answering

A project should not begin while its central purpose is still shifting between substantially different questions. A study designed to estimate prevalence, for example, is not interchangeable with one intended to explain causes or evaluate an intervention.

Refinement is normal, but the core question should be stable enough that participants, variables, data collection, and analysis are being selected for a known purpose. If you are still uncertain about this foundation, you may not yet know the minimum needed to design the study responsibly.

Resolve uncertainties that determine what evidence must be collected

Before irreversible data collection begins, you generally need clarity about information that cannot later be reconstructed. If a variable is necessary for your intended analysis but is never collected, statistical sophistication will not manufacture it afterward.

This is especially important for primary outcomes, exposures, interventions, key covariates, timing of measurements, comparison conditions, and other information central to the intended inference. Exactly what must be fixed varies by methodology, but the principle is straightforward: information that must exist for the research question to be answered should not be left to chance.

Resolve uncertainties that affect participant rights, safety, or informed consent

Ethical uncertainty deserves a lower tolerance than many operational uncertainties. Before participants enter a study, researchers should understand the procedures participants will undergo, reasonably foreseeable risks and burdens, what information will be collected, how consent will operate where required, and how privacy and confidentiality will be protected.

Formal requirements depend on the jurisdiction, institution, study type, and population. The governing ethics body or institutional review process should therefore be treated as the authoritative source for the requirements applicable to a particular project.

Resolve uncertainties that determine who enters the study

Eligibility criteria, sampling logic, recruitment source, and unit of analysis can shape the evidence produced. If these decisions change after recruitment has begun, early and later participants may effectively have been selected under different rules.

Not every recruitment detail must be frozen. Advertisement wording or scheduling procedures, for instance, may sometimes be adjusted without changing the target population. The distinction is whether the change alters who has a realistic opportunity to enter the study or changes the population to which the findings are intended to apply.

Resolve uncertainties that affect comparability

When observations must be meaningfully compared, essential procedures should be sufficiently standardized. Changing an instrument, intervention, coding definition, observation schedule, or data collection mode halfway through a study may introduce systematic differences that become entangled with the phenomenon being studied.

Sometimes such changes are unavoidable and can be documented or modeled. That does not make them methodologically neutral. If a source of comparability can reasonably be secured before the study begins, doing so is usually preferable to repairing avoidable heterogeneity later.

Resolve uncertainties with large downstream dependencies

Some decisions behave like load-bearing walls. Sampling depends on the target population. Measurement depends on the construct. Analysis depends on the data structure. Recruitment may depend on site access. Ethics approval may depend on finalized procedures.

When many later decisions depend on one unresolved issue, carrying that uncertainty forward creates a cascade of provisional choices. In these situations, map the dependencies and resolve the upstream issue first. This is particularly useful when several parts of the study depend on decisions that have not yet been made.

Some operational uncertainty can remain open

You may not know exactly how quickly people will respond to recruitment, whether interview appointments will need rescheduling, how many reminder emails will be necessary, or which minor logistical problems will occur. These are often manageable uncertainties rather than reasons to postpone the study.

The important condition is that reasonable variation in these details should not redefine the study or compromise its ethical and methodological foundations.

Some analytical choices can remain open, but not without boundaries

How much analytical flexibility is acceptable depends heavily on the research design and purpose. Confirmatory research generally requires greater advance specification of hypotheses, outcomes, exclusions, transformations, and analyses because selecting among alternatives after seeing the data can distort statistical inference.

Exploratory research can legitimately be more adaptive. Qualitative analysis may also evolve iteratively as researchers engage with the data. The important issue is transparency about what was prespecified, what developed during the research, and why.

Emergent qualitative research requires a different kind of certainty

It would be inappropriate to apply a rigid pre-specification standard to every qualitative design. In some traditions, sampling, questioning, conceptual categories, and analysis deliberately evolve in response to what researchers learn.

What should be settled is the rationale for that flexibility. Researchers should know why adaptation is methodologically appropriate, what broad question or phenomenon guides the inquiry, how decisions will be documented, and what ethical boundaries constrain adaptation.

Uncertainty can sometimes be managed rather than eliminated

Some uncertainties cannot be answered confidently in advance. Recruitment rates, adherence, missing-data patterns, acceptability, and whether procedures work as intended may only become observable when tested in practice.

In methodological work on feasibility studies, this need for additional information before proceeding is treated as a central reason for conducting feasibility work. A pilot study goes further by implementing the future study, or part of it, on a smaller scale. When an uncertainty can materially affect the full study but cannot be settled through reasoning or existing evidence, preliminary empirical work may be the appropriate response.

Sorting uncertainties before a multi-site survey begins

The team does not need certainty about everything. It needs certainty about the decisions that define the evidence. For unavoidable unknowns such as response rates, it needs a contingency rather than a prediction pretending to be knowledge.

Common mistakes when managing unresolved decisions

Classify uncertainties by consequence, reversibility, and timing

Before starting, make a short uncertainty register. Write down the important things you still do not know rather than allowing them to remain implicit. Then ask three questions about each one: What depends on this? When must it be decided? What happens if we discover later that our assumption was wrong?

Timing matters because research decisions gradually become less reversible. An unresolved issue that is cheap to settle today may become expensive after recruitment and impossible to correct after data collection. Identifying those transition points helps you avoid crossing the project's practical point of no return with foundational questions still unanswered.

Before carrying an uncertainty into the study

Questions about uncertainty before research begins

You do not need certainty about everything

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.