03 · What You Need to Know
Detail should follow consequence, not a target page count
There is no universal number of pages, sections, tasks, variables, or procedural steps that makes a research plan sufficiently detailed.
Authoritative protocol guidance reflects this. The World Health Organization's recommended protocol format identifies the kinds of issues that a developed study may need to address, including rationale, objectives, design, population and sampling, methodology, instruments, data management and analysis, quality assurance, project duration, anticipated problems, responsibilities, and ethical considerations. NIH likewise provides different protocol templates for different forms of research rather than one template assumed to fit every study.
The practical implication is that detail should be proportional to what the study needs.
A one-researcher analysis of a well-documented public dataset may require a concise operational plan. A multisite intervention involving recruitment, several data collectors, repeated measurements, sensitive information, and multiple approvals may require extensive documentation.
Both plans can be sufficiently detailed. They simply have different jobs to do.
Begin by asking what the plan needs to accomplish
A research plan can serve several purposes simultaneously.
It can help you test whether the study is coherent. It can guide implementation. It can coordinate a team. It can support ethics or institutional review. It can document what was intended before results were observed. It can expose dependencies and provide a basis for monitoring progress.
The amount of detail needed depends partly on which of these functions matter for the project.
For example, if one researcher will conduct all interviews, a short description of interviewer responsibilities may be adequate. If twenty interviewers across five sites will conduct them, the procedures may need considerably greater standardization.
The research question should be clear enough to constrain the study
Before detailed implementation begins, the research question should do more than identify a topic.
Compare:
Generative AI in higher education
with:
How do university instructors decide whether generative AI is appropriate for preparing assessment materials?
The first identifies an area. The second begins to indicate what phenomenon needs to be investigated and what kind of evidence may be relevant.
You do not need to treat the wording as sacred. Early planning may still justify refinement. However, the question should be stable enough that the rest of the study is not being designed around a moving target.
If you are still unsure what evidence would answer the question, return to the first decisions that follow from settling a research question before adding procedural detail.
The design should be specific enough that you know what kind of evidence will exist
Writing “mixed methods,” “qualitative research,” or “quantitative survey” in the plan is not necessarily enough.
You should be able to explain how the broad design connects the question to evidence.
For example, if the study uses interviews, why are interviews appropriate? Who will be interviewed? What information are those interviews expected to provide? If the study uses a survey, what constructs or variables need to be represented? If an existing dataset will be analyzed, which records and variables make it suitable?
The plan does not necessarily need every downstream detail at the earliest stage. It does need enough specificity to establish that the proposed evidence can support the intended research claim.
Population and sampling detail should increase before recruitment or selection begins
Early in planning, “undergraduate students” may be sufficient to explore feasibility. Before recruitment, that description is usually inadequate.
You may need to clarify:
- which institution or setting;
- which academic level or program;
- what counts as current enrollment;
- inclusion and exclusion criteria;
- how potential participants will be identified;
- how they will be approached;
- how the sample or cases will be selected; and
- what determines the intended sample size or sampling adequacy.
Equivalent detail is needed for non-participant research. A document analysis may need eligibility criteria for documents. Secondary-data research may need record-selection rules. A systematic review requires explicit eligibility criteria. Laboratory research needs criteria governing specimens or experimental units.
The important boundary is the point at which selection begins. Once participants, cases, documents, records, or observations start entering the study, ambiguous selection rules can become difficult to repair.
Procedures should be detailed enough that implementation is not invented case by case
A useful plan should tell the research team what is supposed to happen.
For a simple interview study, that might include recruitment, consent where applicable, scheduling, approximate interview duration, use of the interview guide, recording, field notes, secure transfer of files, and procedures for unusual situations.
For an experiment, substantially more detail may be required about assignment, conditions, intervention delivery, timing, equipment, measurement, deviations, and quality control.
WHO recommends that study procedures be described in sufficient detail and notes that when multiple sites are involved in a specified protocol, methodology should be standardized and clearly defined.
A practical test is:
Could another appropriately trained person conduct the planned procedure consistently enough using this documentation?
If not, more operational detail may be needed.
But not every procedure needs word-for-word scripting
Consistency should not be confused with mechanical uniformity.
A semistructured qualitative interview can allow responsive probing. Ethnographic observation may intentionally follow emerging events. Adaptive designs may include prespecified mechanisms for modification. Exploratory analysis may generate new questions.
In such cases, the plan should specify the logic and boundaries of flexibility rather than pretending that no adaptation will occur.
For example, a qualitative protocol might state the core domains that all interviews should address while allowing follow-up questions based on participant responses. That is more useful than either scripting every possible sentence or saying only, “Participants will be interviewed about their experiences.”
This is the distinction between incomplete planning and methodologically appropriate flexibility at the beginning of a study.
Instruments should be ready for the role they are about to play
An instrument can remain a draft while the study is still exploring measurement options. It should not remain conceptually unfinished when it begins generating main-study evidence.
Before use, determine what level of preparation the instrument requires. Depending on the study, that may involve validity evidence, reliability considerations, permissions or licensing, translation, cultural adaptation, pilot or pretesting, technical configuration, scoring rules, interviewer guidance, or equipment calibration.
The appropriate standard is not “perfect instrument.” It is “sufficiently justified and prepared for its intended use.”
If important changes remain likely, consider whether you are still in development or piloting rather than main data collection.
Data management should be specific before the first data are created
“Store data securely” is an intention, not much of a plan.
Before data collection or acquisition, you should generally know:
- what files or records will be created;
- where they will be stored;
- how they will be named and organized;
- who will have access;
- how identifiers will be handled;
- how backups or other protections will work;
- what documentation will accompany the data;
- how versions will be managed; and
- what applicable retention, sharing, or disposal requirements exist.
Formal requirements vary. NIH's Data Management and Sharing Policy, for example, requires covered research to prospectively plan how scientific data will be managed and shared. NIH's guidance identifies elements such as data type, related tools or software, standards, preservation and access, distribution or reuse considerations, and oversight of data management and sharing.
Those requirements apply to covered NIH research rather than every study. The underlying planning principle is widely useful: decisions about data should precede the arrival of the data they govern.
Analysis should be detailed enough to test whether you are collecting the right evidence
You do not always need a fully locked statistical analysis plan or complete qualitative coding framework before research begins.
You do need enough analytical thinking to answer:
If I obtain the data exactly as planned, how will they help me answer the research question?
WHO's recommended protocol format asks researchers to describe data management and analysis, including proposed statistical methods and sample-size reasoning where relevant, and calls for sufficient detail about qualitative analysis.
For quantitative studies, advance detail may include primary variables or outcomes, major comparisons, variable construction, analytical methods, and foreseeable missing-data issues.
For qualitative research, it may include the analytic approach, how material will be prepared and coded, how interpretation will develop, and how the approach fits the methodological tradition.
The amount of prespecification should reflect whether the study is exploratory, confirmatory, iterative, or otherwise governed by a particular methodological logic.
Confirmatory claims generally require more advance detail
If the study intends to test a specific hypothesis or estimate a prespecified primary effect, decisions that could be influenced by observing the results deserve greater advance specification.
For example, choosing the primary outcome, exclusion criteria, subgroup definitions, transformations, or analytical models only after seeing which choices produce the strongest result can undermine the distinction between confirmatory and exploratory evidence.
This does not make post hoc or exploratory analysis illegitimate. It means the reporting should distinguish what was specified before relevant results were observed from what was developed afterward.
Preregistration, registered reports, statistical analysis plans, trial registration, or other prospective documentation may impose additional requirements depending on the study and discipline.
Ethical procedures need enough detail to describe what will actually happen
Where research involves human participants, the plan should not rely on generic statements such as “ethical principles will be followed.”
The relevant detail may include recruitment, informed consent, foreseeable risks, participant burden, privacy, confidentiality, withdrawal, compensation or reimbursement where applicable, data access, and procedures for relevant adverse or unexpected events.
WHO's recommended protocol format requires ethical considerations to be addressed and treats informed consent as an integral component of the protocol for research involving participants.
Requirements vary across institutions and jurisdictions. Use the standards and templates of the responsible ethics body rather than assuming that a generic protocol is sufficient.
The more people implementing the study, the more ambiguity matters
A single researcher can sometimes compensate for incomplete documentation through memory. That is risky enough. A team cannot reliably depend on everyone remembering the same unwritten decision.
As the number of investigators, data collectors, analysts, sites, laboratories, or partners increases, the plan often needs greater detail about:
- responsibilities;
- standardized procedures;
- training;
- communication;
- version control;
- data transfer;
- quality checks; and
- who has authority to make or approve changes.
The purpose is not bureaucratic expansion. It is reducing variation created by different people interpreting the same vague instruction differently.
The more difficult a decision is to reverse, the earlier it deserves precision
Reversibility provides one of the strongest tests of necessary detail.
| If the decision... |
Then... |
| Can be changed cheaply without affecting the evidence |
It may remain relatively flexible. |
| Determines what data will be collected |
Specify it before the relevant data are generated. |
| Affects participant rights, consent, privacy, or risk |
Resolve it according to applicable ethics requirements before the relevant activity. |
| Creates a major financial or contractual commitment |
Resolve it before committing the resource. |
| Would make earlier and later data difficult to compare if changed |
Standardize it before collection unless adaptation is methodologically justified. |
| Could be influenced by seeing the results |
Prespecify it where necessary for the intended inferential claim. |
| Affects only low-consequence logistics |
Leave room for practical adjustment. |
This is why a research plan often becomes progressively more detailed. Early exploration tolerates many provisional choices. Approaching recruitment, data collection, intervention, or analysis creates commitment points at which particular decisions need greater precision.
Dependencies should be detailed enough to reveal what can block the project
A plan can describe the scientific method beautifully while remaining operationally unrealistic.
Suppose recruitment requires ethics approval, school authorization, an approved survey, a functioning data system, and trained research assistants. If those dependencies are not visible, “begin recruitment June 1” is more aspiration than schedule.
Identify the major prerequisites, who controls them, when they can begin, and when they must be resolved.
For complex projects, explicitly map ethics approval, recruitment, data access, and other dependencies rather than burying them inside broad timeline categories.
The timeline should be detailed enough to expose infeasibility
A useful timeline does not need every research activity broken into fifteen-minute blocks.
It should, however, show the major stages, milestones, dependencies, review periods, externally controlled processes, and fixed deadlines well enough to reveal whether the study fits the available time.
“January to June: research” is too broad to be useful.
At the other extreme, scheduling “open statistical software” and “create new project file” is unlikely to improve project control.
A good task is usually concrete enough to estimate and monitor and substantial enough to represent meaningful work.
If the project has a hard endpoint, work backward from the submission or graduation deadline to determine when major milestones actually need to occur.
Detail should increase around uncertainty, not only around certainty
Researchers sometimes document the parts they understand best in extraordinary detail while leaving the uncertain parts as vague placeholders.
For example:
Statistical analysis: six pages.
Recruitment: participants will be recruited from local universities.
If recruitment is the part most likely to determine whether the study is feasible, the imbalance is unfortunate.
Detail is particularly valuable where it exposes assumptions. How many universities? Who grants access? Approximately how many eligible participants exist? What recruitment channels are permitted? What happens if recruitment is slower than expected?
You may not know every answer yet. That is precisely why the uncertainty should be made explicit rather than hidden behind a broad sentence.
Some uncertainties should be documented rather than prematurely resolved
A detailed plan does not require pretending to know what cannot yet be known.
If the ethics review date is uncertain, record the dependency and planning range rather than inventing a precise approval date. If qualitative sampling will legitimately evolve based on emerging analysis, describe the sampling logic rather than manufacturing a final participant sequence. If an exploratory analysis may generate new hypotheses, distinguish that possibility from prespecified confirmatory analyses.
The plan should represent uncertainty accurately.
False precision is not the same as detail.
Use working assumptions when downstream planning cannot wait
Sometimes the project needs to move forward while an upstream detail remains unresolved.
State the assumption explicitly:
Working assumption: two participating schools will provide access to approximately 300 eligible students. Recruitment targets and timeline will be reassessed if confirmed access differs materially from this estimate.
This allows planning to continue without disguising uncertainty as a settled fact.
If several parts of the study are waiting on unresolved assumptions, use the principles for managing research decisions that depend on other unmade decisions.
Protocol templates are floors or structures, not substitutes for thinking
Templates can help researchers remember important areas, but completing every heading does not guarantee that the study is adequately planned.
NIH provides different protocol templates for behavioral and social science research, interventional trials, observational studies, prospective data collection, repositories, retrospective research, and secondary research.
The diversity of those templates reinforces an important point: the right detail depends on the study.
A template can prompt you to describe sampling. It cannot decide whether your sampling strategy makes sense. It can provide a heading for analysis. It cannot determine whether the proposed analysis answers the research question.
Use templates to structure reasoning, not replace it.
More detail is not always better
Excessive detail has costs.
It consumes time. It creates more material to update when the study changes. It can make important decisions harder to find. It may encourage premature commitment to details that should remain flexible. It can also create a false impression that uncertainty has been eliminated.
Consider a semistructured interview study. Writing 150 possible follow-up questions may create a more detailed document, but not necessarily a better interview process. A clearer plan might identify six core domains, the purpose of probing, boundaries around sensitive topics, and how interviewers should follow relevant participant responses.
The appropriate question is therefore:
What ambiguity would additional detail remove, and does removing that ambiguity matter?
If you cannot answer that, the additional detail may not be helping.
Too little detail has a different cost
Underplanning pushes decisions downstream.
The study begins, and researchers then discover they do not agree on eligibility. Interviewers use different procedures. Important variables were omitted. File names become inconsistent. The analysis requires information that was never collected. Consent documents and actual procedures diverge. Nobody knows which version of the questionnaire is current.
These are not merely administrative inconveniences. Some can affect the validity, ethics, reproducibility, or interpretability of the study.
Watch Out
If an unresolved detail could change who enters the study, what participants experience, what evidence is generated, how data are protected, or what conclusions the study can support, do not dismiss it as something that can simply be worked out later.
Use the next commitment point to decide how much detail you need now
You do not need the same level of detail throughout the entire project on day one.
Instead, ask what the next consequential commitment is.
Before choosing the broad design Clarify the question, evidence requirements, feasibility, and major methodological alternatives.
Before submitting for formal review Develop the protocol and associated materials to the level required by the reviewing body.
Before recruitment Clarify eligibility, recruitment, participant information, consent, responsibilities, and applicable permissions.
Before data collection Finalize the decisions that determine what evidence will be generated and how it will be handled.
Before confirmatory analysis Ensure that the analytical decisions requiring advance specification are documented to the degree appropriate for the design.
Before submission Clarify reporting, authorship, review, documentation, data-sharing, and other completion requirements that apply.
This progressive approach allows the plan to mature without requiring researchers to predict every downstream detail before they have enough information to make it intelligently.
A plan is detailed enough when another question becomes more important than “What are we doing?”
There is no magical moment at which the plan becomes complete.
But a useful readiness signal is that the major uncertainties have shifted.
Early in planning, you may ask:
What exactly is the study? Who are we studying? What evidence do we need? How will we obtain it?
Later, the remaining questions become more operational:
Which interview slot works for this participant? When should this independent task be scheduled? Which team meeting should review the first quality report?
When the remaining uncertainty is largely operational or methodologically legitimate, while consequential scientific, ethical, analytical, and dependency decisions are sufficiently resolved for the next stage, the plan may be detailed enough to proceed.
The final challenge is recognizing when you have planned enough and need to start.
