03 · What You Need to Know
Manageability begins with knowing what the study cannot afford to lose
When a project becomes too large, researchers often begin cutting whatever looks expensive.
Fewer participants. Fewer variables. One less site. Remove the interviews. Shorten the follow-up. Drop the comparison group.
Some of these changes may be entirely sensible. Others may destroy the logic that made the study capable of answering its question.
The first step is therefore not to ask:
What can I remove?
Ask:
What must remain true for this study to answer the research question credibly?
That scientific core becomes the boundary within which simplification can occur.
Distinguish scientific necessity from project complexity
Some complexity is demanded by the research question.
If the question asks whether an intervention performs differently from usual practice, some defensible comparison is central to the question. If the question concerns change over time, temporal evidence is not decorative. If the phenomenon is explicitly about differences between two populations, removing one population changes the question.
Other complexity is optional.
You may have five secondary outcomes because all five seem interesting. You may have added interviews because mixed methods sounds richer. You may be recruiting from six sites even though the research question does not concern site variation. You may have seven research questions because the dataset appears capable of answering all of them.
Necessary complexity
Complexity required to answer the research question, support the intended comparison or inference, protect participants, or preserve methodological integrity.
Optional complexity
Additional populations, measures, methods, sites, analyses, or outputs that may be interesting but are not essential to the central research purpose.
Manageability usually improves by reducing the second category before compromising the first.
Start with one central research problem
A project often becomes unmanageable because several worthwhile studies have been combined into one.
Suppose the original project asks:
- How often do university students use generative AI?
- What predicts their use?
- Does AI use improve academic performance?
- How do students perceive institutional policies?
- How do instructors perceive those policies?
- How do policies differ across universities?
- What ethical concerns arise from AI use?
These questions are related, but related does not mean they belong in one study.
Ask which question represents the contribution you most need the project to make. The others can then be evaluated as supporting questions, future studies, or questions to remove.
Narrowing a project is not necessarily intellectual retreat. Sometimes it is what allows the research to say something precise rather than many things superficially.
Reduce the number of research questions before weakening each one
When time is limited, researchers sometimes preserve every research question but reduce the quality of evidence available for each.
A better strategy may be to answer fewer questions properly.
Suppose a mixed-methods project has four quantitative questions and three qualitative questions. If resources cannot support all seven, ask which are central to the study's purpose.
Removing peripheral questions can reduce:
- instrument length;
- participant burden;
- sample-size requirements for particular analyses;
- coding workload;
- analytical complexity;
- multiple-testing concerns;
- writing demands; and
- the number of conclusions that need defensible evidence.
This can make the study substantially smaller without reducing the quality of the evidence supporting its central question.
Do not collect variables simply because they might become interesting later
Questionnaires and data extraction forms tend to grow.
Someone suggests another demographic variable. A collaborator recommends an additional scale. You find a construct in the literature that “might be useful.” Soon a ten-minute survey requires forty minutes.
For each variable or measure, ask:
What research question, analytical requirement, confounding concern, descriptive need, or methodological purpose does this serve?
If the answer is merely “we might analyze it,” consider whether the additional burden is justified.
This is not an argument for collecting the absolute minimum. Important contextual variables, confounders, quality indicators, or descriptive characteristics may be necessary even when they are not the headline outcome. The point is to require a reason for inclusion.
Reduce secondary outcomes before compromising the primary outcome
Studies can accumulate outcomes because measuring more seems to increase value.
It also increases instrument length, analytical workload, interpretation, multiplicity, and reporting complexity.
If the study has a clearly defined primary outcome or central phenomenon, protect its measurement first. Secondary outcomes should earn their place by contributing meaningfully to the study.
This principle is especially important in confirmatory research, where ambiguity about which outcomes matter most can complicate interpretation.
A narrower outcome set can make the project more focused while leaving the primary scientific question intact.
Reduce sites when site diversity is not essential to the question
Every additional site can add access negotiations, ethics or governance requirements, recruitment coordination, training, data transfer, communication, monitoring, and site-level variation.
If the research question explicitly concerns differences among institutions, reducing sites may undermine the study.
If the question concerns a phenomenon that can reasonably be studied within one setting, however, a single-site project may be more appropriate for the available resources.
The tradeoff must then be reflected in interpretation. Evidence from one university does not automatically support claims about all universities.
This illustrates a recurring rule:
When you narrow the evidence base, narrow the claim accordingly.
Reduce populations when the comparison is not scientifically essential
Researchers sometimes include students, instructors, administrators, employers, and policymakers because each perspective is interesting.
Each additional population can require different recruitment channels, eligibility criteria, instruments, sampling strategies, consent materials, analyses, and interpretations.
If the study's central question concerns students' experiences, instructor interviews may be useful context but not necessarily essential evidence. They could become a separate study.
If the research question explicitly asks how student and instructor perspectives differ, however, removing one group changes the question itself.
The decision should therefore come from the research logic rather than workload alone.
Do not add a method merely because multiple methods appear more rigorous
Mixed-methods research can answer questions that require meaningful integration of quantitative and qualitative evidence. It also creates substantial additional work.
You may need two sampling strategies, two forms of data collection, two analytical processes, and a defensible integration strategy.
If the research question can be answered adequately through one methodological approach, adding another method does not automatically improve the study.
Ask:
What can the second method tell me that the first method cannot, and does the research question require that information?
If you cannot articulate the added inferential or explanatory value, the second method may be optional complexity.
Mixed methods should be selected because the question requires integrated evidence, not because one method feels insufficiently ambitious for a thesis.
Likewise, do not remove a method if the question genuinely requires it
The reverse mistake also occurs.
Suppose a study asks both whether an intervention changes an outcome and how participants experience the intervention. Quantitative evidence may address the first question while qualitative evidence addresses the second.
Removing one component can still produce a coherent study, but only if the research question is revised accordingly.
Do not retain the original multi-dimensional claims after removing the evidence needed to support them.
Simplify the design before simplifying the measurement
When a study feels too large, it may be tempting to replace strong measures with whatever is easiest to administer.
That can be a poor trade.
A shorter project using appropriate measures may be more defensible than a broader project measuring its central constructs poorly.
For example, if a validated instrument is too burdensome because the survey contains six other optional constructs, consider removing peripheral constructs before replacing the central measure with one untested question.
Manageability should not be achieved by weakening the evidentiary link between the construct and its measurement.
Simplify procedures without changing what they measure
Operational simplification can be particularly valuable because it may reduce workload without changing the scientific question.
Depending on the study, you might:
- use one standardized data collection mode rather than several;
- centralize recruitment;
- reduce unnecessary participant visits;
- collect several measures during one session;
- automate appropriate administrative processes;
- use a consistent file and data-management workflow;
- remove duplicate documentation; or
- standardize procedures across researchers.
The key is to verify that the simplified procedure still generates equivalent or appropriate evidence and remains consistent with applicable ethical and institutional requirements.
Use existing data when they genuinely fit the question
Secondary data can remove recruitment and primary data collection from the project, potentially reducing time, cost, and participant burden.
But existing data are not automatically a shortcut.
The dataset must contain appropriate variables, population coverage, measurement quality, time periods, documentation, and access rights. Researchers may also face substantial cleaning, harmonization, governance, or missing-data problems.
Do not change the research question merely to fit whichever dataset is easiest to obtain unless the revised question is itself worthwhile and scientifically coherent.
The data source should serve the question, not quietly become the question.
Use existing instruments when they are appropriate, not merely convenient
Developing a new instrument can require substantial conceptual work, testing, validation, and revision.
An existing measure may reduce that burden when it is appropriate for the construct, population, language, context, administration mode, and intended interpretation.
However, adopting an existing instrument does not eliminate the need to evaluate its suitability. Permissions or licensing may also apply.
The manageable option is not necessarily the shortest questionnaire. It is the instrument strategy that provides adequate evidence without unnecessary development work.
Reduce time points only if the temporal structure remains meaningful
Longitudinal studies can become difficult because every additional measurement occasion increases scheduling, retention, data management, and analysis demands.
Some time points may be removable.
But if the question concerns change, trajectory, delayed effects, sustainability, or temporal ordering, reducing follow-up may alter what the study can establish.
Suppose a study asks whether an educational intervention produces improvements that persist for six months. Removing the six-month follow-up because the thesis deadline is approaching does not merely simplify the study. It removes the evidence needed to address persistence.
The scientifically coherent alternatives may be to revise the question to immediate outcomes, change the deadline, or redesign the project.
Reduce follow-up burden before reducing essential follow-up duration
If longitudinal follow-up is scientifically necessary, look first for ways to make it operationally lighter.
Could follow-up use a shorter instrument focused on the necessary outcomes? Can visits be conducted remotely where appropriate? Can redundant measurements be removed? Can scheduling be improved?
Any change should still preserve measurement validity, participant protections, and protocol requirements.
The point is to simplify how the essential evidence is collected before eliminating the evidence itself.
Do not solve an unrealistic sample requirement by choosing an arbitrary smaller number
Sample size is an especially tempting place to simplify.
If the original plan requires 400 participants and only 100 appear feasible, choosing 100 because that is what can be recruited does not automatically make the revised study scientifically adequate.
For quantitative research, reassess the statistical objectives, expected precision, power where applicable, analysis complexity, effect sizes, design, and intended claims.
For qualitative research, avoid treating sample adequacy as a numerical target detached from the methodology. Sampling decisions may depend on the study purpose, heterogeneity, information needs, analytic approach, and methodological tradition.
If the feasible sample cannot support the original question, revise the question or design rather than preserving both and hoping the limitation section will repair the mismatch.
Reduce analytical ambition before forcing unstable analysis from inadequate data
A modest dataset can inspire surprisingly ambitious models.
Researchers may plan many predictors, interactions, subgroup analyses, mediation models, latent variables, machine-learning comparisons, or multilevel structures because the software makes them technically possible.
The relevant question is whether the data and research design support those analyses credibly.
If not, simplifying the analysis can improve rather than weaken the study.
Focus on analyses that directly address the research question. Secondary and exploratory analyses can remain secondary and exploratory rather than becoming additional obligations the project must satisfy.
Do not confuse sophisticated analysis with scientific sophistication
A complex statistical method does not compensate for a weak design, inappropriate measurement, biased sampling, or poorly defined question.
Conversely, a straightforward analysis can provide strong evidence when the design and data fit the question well.
Choose the analytical complexity required by the inferential problem rather than by the researcher's desire for the methods section to look impressive.
A method should earn its complexity.
Reduce outputs when the research itself is becoming overloaded
A project may be scientifically manageable but become operationally unmanageable because too many outputs have been attached to it.
A thesis, three journal manuscripts, two conference papers, a policy brief, public dataset, website, workshop, and infographic may all be worthwhile.
They do not all necessarily need to be completed before the central project deadline.
Separate:
Project-critical outputs
Outputs required to satisfy the research objective, funder, degree, contractual obligation, or other binding commitment.
Follow-on outputs
Useful dissemination products that can be developed after the core research project is completed.
This protects research time from being consumed by dissemination ambitions before the evidence itself is ready.
Reduce precision in scheduling, not rigor in science
Sometimes the project appears unmanageable because the plan itself has become unnecessarily detailed.
You may not need exact dates for every interview six months in advance. You may not need to predict every analysis task before seeing the structure of the cleaned data. You may not need a contingency for every imaginable minor inconvenience.
A plan can remain operationally flexible while the scientific core stays stable.
This is one reason to distinguish what can remain flexible early in a research project from decisions that need to be fixed.
Protect participant safety and ethical requirements from scope pressure
Ethical safeguards are not optional complexity.
Do not simplify a project by reducing informed-consent information below what is required, weakening privacy protections, skipping necessary monitoring, bypassing ethics review, or using an unauthorized recruitment or data-access route.
WHO's recommended research protocol format integrates ethical considerations into study planning and expects research involving participants to address issues such as informed consent and relevant risks. The Declaration of Helsinki likewise emphasizes that the importance of the research objective does not override the rights and interests of individual research participants.
Where an ethical or institutional requirement makes the project infeasible within the available resources, change the project rather than removing the protection.
Protect data quality from workload pressure
Quality-control procedures can look like administrative overhead when deadlines tighten.
Yet eliminating checking, calibration, interviewer training, coding verification, data documentation, or other necessary controls may save time by creating less trustworthy evidence.
Instead, simplify the data-generating process so that the required quality procedures remain feasible.
For example, collecting fewer well-documented variables may be preferable to collecting hundreds of poorly managed ones.
Protect the connection between evidence and claims
The most important scientific boundary is the relationship between what the study observes and what the researcher eventually claims.
Suppose a project originally planned five universities but can realistically recruit from one. Conducting the one-site study may still be entirely defensible.
What changes is the scope of inference.
Instead of claiming:
“University students use generative AI in this way...”
the study may need a more bounded interpretation tied to the particular setting and sample.
A narrower study does not need to pretend to be broad.
This principle often allows projects to become considerably more manageable without compromising integrity: reduce the evidence base when justified, then reduce the breadth of the claim to match it.
Use feasibility constraints to redesign, not merely shrink
Sometimes removing components one by one produces a study that is smaller but incoherent.
A better response is to redesign around the constraint.
Suppose a researcher planned a large longitudinal survey across several universities but has only six months.
Possible redesigns might include:
- a focused cross-sectional study addressing a narrower question;
- a smaller feasibility study explicitly examining whether the larger project is workable;
- a secondary-data analysis if an appropriate dataset exists;
- a qualitative study investigating a question that does not require population-level estimation; or
- a single-site study with claims appropriately bounded to that context.
These are not interchangeable substitutes. Each answers a different question.
The researcher should therefore choose the revised scientific question and design together rather than retaining the original question while progressively removing the evidence required to answer it.
Do not allow the deadline to dictate scientifically incoherent simplification
A fixed deadline is a legitimate constraint. It is not a methodological argument.
If a six-month follow-up is essential to the question but only four months remain, “we only had four months” does not make a four-month observation equivalent.
The appropriate response is to reconsider the question, design, resources, or deadline where possible.
Use backward planning from the final deadline to discover this mismatch before the study begins.
Watch Out
If making the project fit requires removing the comparison, measurement, follow-up, sampling structure, participant protection, or analytical evidence needed to answer the research question, you have crossed from simplification into redesign. Either redesign the question with the study or reconsider the constraint.
Prioritize what cannot be recovered later
When resources are limited, protect decisions and activities that become irreversible.
You can shorten a manuscript later. You can decide not to conduct an optional exploratory analysis. You can postpone a conference presentation.
You cannot easily recover a variable that was never measured, a follow-up that never occurred, a participant group that was excluded from a comparison the question requires, or a consent process that should have happened before data collection.
This provides a useful ordering principle:
Protect irreversible scientific and ethical requirements first. Simplify reversible and optional work second.
Ask whether each component earns its workload
Every component of a research project creates costs.
A new site creates coordination. A new measure creates participant burden and analysis. A new method creates another data collection and analytical workflow. A new population creates recruitment and interpretation. A new time point creates retention and scheduling. A new research question creates another claim that requires evidence.
For each component, ask:
Contribution What does this component add to the answer?
Necessity Can the central question still be answered without it?
Cost What time, participants, money, expertise, coordination, analysis, and documentation does it require?
Dependency What else becomes harder or slower because this component exists?
Consequence of removal Would removing it merely narrow the study, or would it invalidate the intended claim?
This is essentially a scientific value-to-complexity test.
Think in terms of a minimum defensible study, not a minimum possible study
The smallest study you can physically conduct is not necessarily worth conducting.
A more useful concept is the minimum defensible study: the smallest version of the project that can still answer a worthwhile research question with appropriate evidence, methodological coherence, ethical integrity, and claims proportionate to what was actually observed.
That version may still be demanding.
But once it is identified, everything beyond it becomes easier to classify as either valuable enhancement or unnecessary complexity.
Build outward from the minimum defensible study only when resources allow
Instead of starting with the most ambitious imaginable design and cutting components under pressure, begin with the scientific core.
Then ask what additions materially strengthen the project.
Perhaps a second site meaningfully improves variation in context. Perhaps a qualitative component explains a mechanism that the quantitative data cannot address. Perhaps an additional follow-up captures an outcome central to the theory.
Add those components when their scientific value justifies their cost and the timeline can absorb them.
This approach tends to produce cleaner contingency planning because optional enhancements are distinguishable from the project's essential structure.
Use a scope hierarchy when you need to cut quickly
When a project is already too large, reductions can be prioritized rather than improvised.
| Consider reducing first |
Reduce cautiously |
Protect unless the question or design changes |
| Optional outputs |
Number of sites |
Participant protections and required approvals |
| Peripheral research questions |
Secondary outcomes |
Core construct measurement |
| Exploratory analyses |
Additional populations |
Essential comparison structure |
| Nonessential variables |
Follow-up occasions |
Evidence needed for the primary question |
| Duplicative procedures |
Methodological components |
Minimum sample or sampling logic needed for defensible analysis |
| Premature dissemination products |
Geographic or contextual breadth |
Necessary data quality and management procedures |
This is not a universal ranking. A secondary outcome could be central in one study, and multiple sites may be indispensable in another. The table is a prompt to ask what is essential before making cuts.
Recheck feasibility after every major simplification
A scope reduction should solve something measurable.
If removing two optional outcomes reduces the survey by only three minutes but recruitment remains impossible, the main feasibility problem remains.
If dropping one site saves months of access negotiations and makes the timeline workable while leaving the central question intact, the simplification has materially improved feasibility.
After each substantial change, recalculate:
- timeline;
- recruitment;
- workload;
- analysis;
- budget;
- dependencies;
- required approvals; and
- what the resulting study can claim.
If the project has already been formally reviewed or data collection has begun, also follow the appropriate process for managing changes to the research plan.
Stop simplifying when the remaining study becomes both feasible and worthwhile
The goal is not the smallest project.
It is a project that fits the available resources and still produces evidence worth interpreting.
If further reductions would remove the central comparison, make the primary construct poorly measured, eliminate necessary follow-up, undermine sampling, or leave the study unable to answer a meaningful question, you have reached the boundary.
At that point, the remaining options are not further simplification. They are additional resources, more time, a different design, or a different research question.
