03 · What You Need to Know
Research planning is iterative, but it should not be circular
A perfectly linear research plan would be convenient. You would settle the question, choose the design, determine the sample, select the measures, specify the analysis, obtain approval, and execute the study. Each decision would remain untouched while the next one followed neatly behind it.
Actual research rarely behaves that way.
A proposed sampling strategy may reveal that the intended population is inaccessible. That may require reconsidering the setting. A measurement problem may reveal that the research question is broader than the available evidence can support. An analysis requirement may show that an important variable has been omitted. An ethics concern may require changing recruitment. A feasibility problem may force reconsideration of the scope.
Iteration is therefore normal. The objective is not to prevent earlier decisions from ever being reconsidered. It is to make the dependencies explicit enough that revision becomes purposeful rather than repetitive.
First, distinguish an unresolved decision from an unresolved task
Suppose your project plan says:
Finalize questionnaire.
You have revised it repeatedly, but it still does not feel finished.
The problem may not be the questionnaire. Perhaps the team has not decided whether “academic engagement” is a primary outcome, secondary outcome, or contextual variable. Until that conceptual decision is made, the instrument cannot be finalized intelligently.
Blocked task
The visible piece of work that cannot be completed, such as finalizing a questionnaire.
Unresolved decision
The upstream choice preventing completion, such as deciding which construct the questionnaire actually needs to measure.
Working harder on the blocked task may produce another draft without resolving the problem.
When work repeatedly stalls, ask: What decision would need to be made for this task to become straightforward?
Write down the decisions that are actually unresolved
Uncertainty becomes difficult to manage when it remains implicit.
Instead of writing “methodology still needs work,” identify the specific open decisions:
- Will the study compare two groups or examine one population descriptively?
- Which outcome will be primary?
- Will recruitment occur through one institution or several?
- Will the study use an existing measure or develop a new instrument?
- Will interviews be conducted before or after the survey?
- Will identifiable information need to be retained?
- Which dataset can provide the required variables?
A project with seven explicit open decisions is easier to manage than a project whose entire methodology is simply described as “not final.”
Map which decisions depend on which others
Once the open decisions are visible, draw the relationships among them.
Suppose a project has these unresolved choices:
- sample size;
- sampling strategy;
- primary outcome;
- statistical analysis;
- questionnaire content.
These are not five independent problems.
The primary outcome may influence the questionnaire and sample-size reasoning. The research design and outcome structure influence the analysis. The population and sampling strategy affect recruitment feasibility. The intended analysis affects what information needs to be collected.
A simplified decision map might therefore look like this:
Research question and intended claim What exactly must the study be able to conclude?
Required evidence and primary outcome What information would support that conclusion?
Study design and population How and from whom or what can that evidence be obtained?
Sampling and measurement Which cases and measures will generate the required data?
Analysis and sample-size implications How will the resulting evidence be interpreted, and what quantity or structure of data is needed?
The actual sequence will differ among methodologies. The point is to discover which choices are upstream rather than assuming that every unresolved item has equal priority.
Look for the decision with the greatest downstream influence
When several choices are open, prioritize the one whose resolution would clarify the largest number of other decisions.
Suppose you are debating survey length, sample size, statistical analysis, and recruitment strategy, but you have not yet decided whether the study is primarily descriptive or intended to estimate a particular association.
That purpose is likely more consequential than the survey length. Resolving it may clarify which variables are necessary, what analytical approach is appropriate, and what sampling considerations matter.
This is one reason the first decisions after settling on a research question should concern the evidence and broad design rather than immediately committing to downstream procedures.
Do not finalize downstream details while their foundations are unstable
Premature detail can create the appearance of progress.
You might spend several days perfecting questionnaire wording before deciding whether the construct belongs in the study. You might calculate an exact sample size before the primary analysis is defined. You might build an elaborate timeline around a data source whose accessibility has not been confirmed.
These activities are not necessarily wasted, but they carry a high risk of rework.
A useful rule is:
Keep downstream decisions only as detailed as their upstream assumptions justify.
If the study design is provisional, the sampling plan may also need to remain provisional. If the population is settled but recruitment access is uncertain, eligibility criteria can perhaps be developed while exact recruitment channels remain open.
This is controlled provisionality rather than indecision.
Use assumptions explicitly when you need to keep planning
Sometimes you cannot wait for every decision before doing any downstream work.
In that case, state the assumption under which the work is proceeding.
For example:
Working assumption: recruitment will occur through two universities. If only one site is available, the sampling and timeline will need to be reassessed.
Or:
Working assumption: the primary outcome will be the validated writing self-efficacy measure. Instrument length and sample-size planning remain provisional until this choice is confirmed.
Explicit assumptions are useful because they reveal what would need to change if the assumption fails. Hidden assumptions tend to emerge much later, usually during meetings in which everyone discovers they had been planning a slightly different study.
Separate reversible decisions from expensive commitments
Not every unresolved choice deserves immediate resolution.
Some decisions are easy to change later. Others become costly once resources are committed, participants are involved, data are generated, or approvals are obtained.
| Decision situation |
Planning response |
Example |
| Easy to reverse and low consequence |
May remain provisional |
Internal meeting frequency |
| Needed to resolve several downstream choices |
Prioritize early |
Primary outcome or evidence source |
| Requires information you do not yet have |
Identify how to obtain that information |
Choosing a site when actual recruitment capacity is unknown |
| Creates substantial cost or commitment |
Resolve before committing resources |
Purchasing specialized equipment |
| Affects what data will exist |
Resolve before the affected data are collected |
Whether a variable required for the main analysis will be measured |
| Affects participants or approved procedures |
Resolve according to applicable ethics requirements |
Recruitment, consent, or collection of sensitive identifiable information |
This prevents researchers from spending equal amounts of decision-making effort on choices with very different consequences.
If a decision cannot be made, ask what information is missing
“We cannot decide yet” should lead to another question:
Why not?
Perhaps you need:
- a more focused literature search;
- information from the data custodian;
- a feasibility estimate from a participating site;
- feedback from a statistician or qualitative methodologist;
- information about instrument licensing;
- a pilot or pretest;
- clarification from the ethics office;
- a budget estimate;
- a discussion with a community partner; or
- actual information about how many eligible participants are available.
The missing information becomes the next task.
This transforms:
“We still cannot decide the recruitment strategy.”
into:
“We need estimated eligible participant numbers and access requirements from the two candidate sites before choosing the recruitment strategy.”
The second statement is actionable.
Use feasibility work to resolve decisions that cannot be answered from the desk
Some planning questions require empirical information.
You may not know whether participants can be recruited at the necessary rate, whether a procedure is tolerable, whether an instrument works as intended, or whether data can be extracted reliably until you test the process.
A pilot, feasibility assessment, pretest, technical test, or small-scale procedural exercise may therefore be appropriate, depending on the research and applicable ethics requirements.
The purpose should be explicit. If the unresolved decision is whether a recruitment pathway can support the main study, collect information that informs recruitment feasibility. If the uncertainty concerns survey comprehension, pretesting should address that issue rather than becoming an unfocused miniature version of the entire study.
Feasibility work is most useful when it is linked to a decision that will actually change based on what is learned.
Consult expertise before making a decision that will constrain the entire study
Some dependencies exist because a methodological decision requires expertise the team does not currently possess.
For example, the appropriate sample-size approach may depend on the intended statistical model. A complex qualitative design may require methodological guidance before recruitment and analysis procedures can be finalized. Data-security requirements may need input from an institutional information-security office.
Consulting relevant expertise early can prevent a downstream specialist from later discovering that the study did not collect what their analysis requires.
The aim is not to outsource scientific judgment. It is to obtain the information necessary to make a defensible decision while that decision is still inexpensive to change.
Resolve circular dependencies by identifying the minimum decision needed to proceed
Sometimes two choices genuinely influence each other.
Suppose the sample size depends on the analysis, while the feasible analysis partly depends on the sample size you can realistically recruit.
This does not necessarily have a one-directional solution.
Instead, define the feasible range and iterate:
Start with the scientific requirement What analysis would appropriately address the research question?
Estimate its data requirements What sample structure or quantity would the intended analysis reasonably require?
Check feasibility Can the required participants or observations realistically be obtained?
Reconcile the mismatch If not, reconsider the analysis, design, scope, resources, or question rather than pretending either constraint does not exist.
Iteration is productive when each cycle incorporates new information and narrows the options. It becomes circular when the same unresolved questions are repeatedly discussed without obtaining the information needed to choose among them.
Time-box exploratory planning when appropriate
Some decisions benefit from additional investigation, but research planning can also expand indefinitely.
If two defensible options remain, decide how much additional information is worth obtaining before choosing. You might give the team one week to compare two instruments, request methodological advice, or obtain access information from candidate sites.
At the end of that period, ask whether the new information materially distinguishes the options.
If not, the project may need to choose based on the best available evidence and document the rationale rather than waiting for certainty that is unlikely to arrive.
This helps prevent unresolved decisions from becoming a permanent reason not to start. Eventually, researchers need to determine when planning has become sufficient and the project needs to move forward.
Use decision deadlines, not only task deadlines
Research timelines usually contain dates for activities but fewer dates for choices.
Yet an unresolved decision can block several activities simultaneously.
Consider adding decision milestones such as:
- primary outcome confirmed by March 10;
- data source selected by March 15;
- participating sites confirmed by March 25;
- sampling strategy finalized by April 1; or
- analysis approach agreed before instrument finalization.
These dates should reflect genuine dependencies rather than arbitrary pressure. Their value is that they make decision latency visible.
A project can appear busy while a single unresolved upstream decision quietly prevents several major milestones from advancing.
Keep a decision log for consequential choices
When several decisions evolve together, it can become difficult to remember why a particular option was chosen.
A simple decision log can record:
- the decision;
- the options considered;
- the information available;
- the choice made;
- the rationale;
- the assumptions involved;
- the date;
- who was involved; and
- which downstream parts of the project are affected.
This is particularly useful when the project has several investigators, supervisors, sites, or methodological advisers.
The log does not need to become a second research protocol. Its purpose is to preserve the reasoning behind decisions so that the team does not repeatedly reopen settled questions without new evidence.
Know when a provisional decision needs to become fixed
Early planning can support considerable flexibility. That flexibility decreases as the project approaches consequential commitments.
WHO's recommended protocol format illustrates the range of decisions that eventually need explicit treatment in a developed study: objectives, design, population and sampling, instruments and procedures, data management and analysis, quality assurance, timeline, anticipated problems, project responsibilities, and ethical considerations.
For human-participant research within WHO's own system, the protocol, study instruments, informed-consent documents, and associated materials form part of the ethics-review submission. The exact requirements vary across institutions, but the broader point is that provisional planning eventually has to become sufficiently concrete for review and implementation.
By the time the study approaches data collection, revisit what needs to be fixed before data collection starts. Choices that determine what evidence will exist, how participants are treated, or how the study will be interpreted should not remain indefinitely unresolved.
Use the protocol to consolidate decisions once enough of them are stable
A protocol is useful partly because it forces interconnected decisions into one coherent account.
WHO recommends that a research protocol describe the rationale and objectives, study design, methodology, data management and analysis, quality assurance, project duration, anticipated problems, responsibilities, and ethical considerations. NIH likewise notes that writing a protocol can help researchers solidify the study design, assess feasibility, and plan the analysis.
If drafting the protocol reveals that the sampling section cannot be written because the population is unresolved, or that the analysis section cannot be written because the outcome remains undefined, that is useful information. The protocol is exposing a dependency that still needs attention.
This is one reason creating a research protocol can be more than a documentation exercise. It can function as a diagnostic test of whether the study's decisions actually fit together.
Do not confuse iteration with permission to redesign the study indefinitely
Research planning should permit revision when new information justifies it. However, repeatedly reopening foundational decisions without a clear reason creates instability.
Before revisiting a settled choice, ask:
What new evidence, constraint, requirement, or methodological insight has emerged?
If the answer is substantive, reconsideration may be appropriate.
If nothing has changed and the team is simply uncomfortable committing, additional discussion may produce diminishing returns.
Watch Out
Do not keep every methodological option open until data collection forces a choice. Some uncertainty is legitimate, but unresolved decisions that affect participants, measurement, sampling, data availability, analysis, or required approvals can become far more expensive to correct once the study has begun.
When a decision changes, propagate the change through the dependency map
Suppose the team decides to add another participant group.
That may affect more than the sampling section. It could require different recruitment materials, another consent form, revised interview questions, a larger workload, additional analysis, a longer timeline, new site permission, and possibly an amendment to ethics documentation depending on the study and applicable requirements.
Or suppose the team changes the primary outcome. That may affect measurement, sample-size reasoning, analysis, data collection, and the interpretation of the research question.
When an upstream decision changes, do not update only the paragraph where that decision appears. Trace every downstream element that depends on it.
If those changes become substantial, revisit how the research plan itself should be updated.
