03 · What You Need to Know
Scope Creep Is Usually an Accumulation Problem, Not One Dramatic Decision
First, distinguish scope creep from legitimate scope change
Not every change to an ongoing study is scope creep.
Suppose recruitment reveals that an eligibility criterion is unintentionally excluding the population the study was designed to investigate. New safety information may require a procedural change. A planned data source may become unavailable. An instrument may prove unsuitable. In an iterative qualitative methodology, emerging analysis may appropriately shape later sampling or questioning.
Those situations can justify changing the study.
Scope creep is more usefully reserved for expansion that occurs without adequate control or reconsideration of what the addition does to the project. A question is added because it is interesting. Another outcome is measured because the questionnaire has room. Another subgroup analysis is planned because the demographic variable already exists. Each decision increases scope, but nobody pauses to ask whether the project as a whole remains coherent and feasible.
Legitimate scope change
A deliberate revision made for a scientific, methodological, ethical, or necessary practical reason, with its consequences evaluated and documented.
Scope creep
Incremental expansion of the project's questions, evidence, methods, analyses, populations, or expected outputs without proportionate reconsideration of purpose, capacity, and consequences.
This distinction matters because the solution to scope creep is not a prohibition on change. It is a disciplined process for deciding which changes belong.
You cannot control scope that was never clearly defined
A vague initial scope makes almost every addition appear compatible with the study.
If the project is described as "research on generative AI in higher education," student learning, faculty attitudes, academic integrity, institutional policy, assessment, accessibility, and technology adoption could all plausibly fit.
A more explicit question creates boundaries. If the study examines the relationship between first-year students' use of generative AI for academic writing and writing self-efficacy, proposals involving faculty policy attitudes or institutional expenditure are much easier to recognize as separate questions.
Before data collection, make sure the study establishes at least the central research question, relevant population or cases, phenomena or variables, setting and timeframe where relevant, primary outcomes or objectives where the design requires them, and the methodological approach necessary to answer the question.
These boundaries form the project's baseline scope. If they remain unclear, return to the task of defining the scope and its deliberate delimitations before attempting to police later changes.
Write down what is outside the project when the exclusion is consequential
A clear scope does not require an exhaustive inventory of everything the study will not investigate. Still, some exclusions are worth making explicit because they are especially likely to return later disguised as additions.
For example, a project might state that it examines student self-efficacy rather than objective writing performance, first-year students rather than faculty or postgraduate students, and one academic-writing context rather than all educational uses of generative AI.
That documentation creates a reference point when someone later proposes adding grades, faculty interviews, or general technology-acceptance measures.
The question becomes, "What has changed that now makes this necessary?" rather than, "Could this also be interesting?"
The most dangerous phrase may be “while we’re already collecting data...”
Many additions seem almost free.
While students are already completing the questionnaire, add another scale. While interviews are underway, ask another set of questions. While the dataset is open, test another association. While researchers are visiting the institution, recruit another population.
The marginal collection cost may indeed be small. The research cost can be much larger.
An additional construct may require conceptual justification, instrument validation, participant time, sample-size considerations, analytical decisions, interpretation, and reporting. A new population may require another sampling strategy and raise questions about comparability. Another outcome may create multiplicity concerns in quantitative work. Another interview topic may substantially enlarge the qualitative corpus and analytical task.
Cheap data are not necessarily cheap research.
Require every proposed addition to state what question it answers
When someone proposes adding a variable, population, method, or analysis, ask for the research question or objective that requires it.
If the answer is vague, the addition probably needs more justification.
Consider a proposal to add academic motivation to a study of AI use and writing self-efficacy. Why is motivation needed? Is it a hypothesized confounder, mediator, moderator, outcome, or explanatory construct? Does theory predict a particular role? Will the study test that role?
If the answer is merely "motivation is also important in education," the variable may be relevant to the topic without being necessary to the study.
This simple requirement prevents topical relevance from becoming automatic inclusion.
Make additions pay their methodological cost
A useful change-control question is not only "Would this be interesting?" but "What does this addition require?"
| Proposed addition |
Possible hidden obligations |
|
Another population
|
Eligibility criteria, recruitment, sampling, consent, subgroup adequacy, comparability, additional interpretation |
|
Another variable or construct
|
Conceptual rationale, measurement, participant burden, data quality, analytical role, interpretation |
|
Another outcome
|
Measurement, sample-size implications, multiplicity, prioritization, reporting |
|
Another research site
|
Access, approvals, contextual heterogeneity, site coordination, sampling and analytical implications |
|
Another method
|
Expertise, instruments or protocols, data management, analysis, integration with existing evidence |
|
Another subgroup analysis
|
Rationale, adequate subgroup evidence, statistical or interpretive complexity, multiplicity |
|
Longer follow-up
|
Retention, resources, participant burden, additional approvals, missing data, delayed completion |
If a proposed addition creates substantial new obligations, it deserves a correspondingly strong rationale.
Use a “one in, one out” question even when you do not enforce it literally
When a project is already at its practical capacity, ask: If we add this, what are we willing to remove, reduce, postpone, or resource differently?
This does not mean every new variable requires deleting another variable. The value of the question is that it exposes the fiction that research capacity is unlimited.
If adding interviews requires 80 additional researcher-hours, where will those hours come from? If another site extends recruitment by three months, can the project accommodate that extension? If another outcome requires a larger sample, can recruitment support it?
When nobody can identify how the new obligation will be absorbed, the proposal is not merely an intellectual addition. It is a resource decision.
Protect the primary question from secondary questions
Secondary questions can be valuable, but they should not compromise the project's ability to answer the question that justified the study.
One practical safeguard is to distinguish among primary, secondary, and exploratory questions or objectives when the methodology supports those categories.
The primary question drives the central design. Secondary questions are planned but subordinate. Exploratory questions can investigate additional patterns without being presented as though the study was originally designed to provide definitive answers to them.
The exact terminology varies by methodology, and not every study needs this hierarchy. The underlying principle is broadly useful: not every interesting question needs equal status.
If secondary questions begin requiring independent populations, literatures, methods, and analyses, reconsider whether the project is trying to answer more than one study can support.
Create a parking lot for good ideas
One reason scope creep is difficult to resist is psychological: rejecting an addition can feel like losing a potentially valuable idea.
You do not need to discard it.
Maintain a separate record of possible follow-up questions, secondary analyses, future variables, additional populations, alternative methods, and subsequent studies. When an idea appears, record its rationale and the evidence that motivated it.
This allows the team to say, "This is worth investigating, but not necessarily in the current study."
That distinction is central to a productive research program. A research program can be broad precisely because each individual study does not have to contain everything.
Do not allow the literature review to expand the study indefinitely
Literature review continues throughout many research projects. New papers will introduce additional constructs, theories, populations, and methods.
The discovery of another relevant factor does not automatically invalidate the existing scope.
Ask whether the new evidence reveals a genuine flaw in the current design. Perhaps a newly identified variable is necessary to address confounding. Perhaps recent evidence undermines the validity of the selected instrument. Those discoveries may justify revision.
But if the new literature simply reveals another interesting dimension of the topic, record it for interpretation or future research rather than automatically redesigning the project.
No literature review ends with the reassuring discovery that nobody has ever thought of another variable.
Control collaborator-driven expansion explicitly
Scope creep can emerge from perfectly reasonable collaboration. A statistician sees another analysis. A content expert suggests another construct. A partner institution wants its population represented. A supervisor sees an opportunity for a second objective.
These contributions can improve the study. They can also accumulate because each collaborator evaluates the addition from the perspective of their own expertise rather than the total project burden.
Use the same change criteria regardless of who proposes the idea:
- What research question requires it?
- What does it add to the study's contribution?
- What new methodological obligations does it create?
- Does the existing design support it?
- What resources will it consume?
- What existing priority, if any, should change?
This turns disagreements about scope from questions of personal preference into questions about research design.
Keep the protocol, preregistration, or analysis plan visible
Where a study uses a protocol, preregistration, statistical analysis plan, or another formal research plan, treat it as an active reference rather than a document filed away after approval.
Compare proposed changes with the current version. Does the new idea alter an objective, population, outcome, eligibility criterion, procedure, analysis, or other substantive feature?
For human-participant studies, this can also have governance consequences. NIH guidance, for example, treats changes to approved protocols and study materials as modifications requiring IRB review in its intramural system, and significant changes affecting scientific intent, study design, eligibility, sites, enrollment, or participant risk may require additional prior approval in relevant NIH-funded studies.
Requirements vary across institutions, funders, jurisdictions, and study types. The practical lesson is broader: before implementing a substantive addition, check whether the change requires formal review rather than treating it solely as a research-team decision.
Do not collect first and decide what the question was later
Large datasets make analytical scope creep particularly easy. Once data are available, researchers can test numerous additional associations, outcomes, transformations, subgroups, and model specifications.
Exploratory analysis can be scientifically useful. The problem arises when exploratory findings are retrospectively presented as though the analyses were planned to answer prespecified questions.
Maintain a distinction between analyses specified before examining the relevant results and those developed afterward. If an unexpected pattern produces a new hypothesis, label it appropriately and consider whether it warrants confirmation in independent data.
This protects the study from a subtler form of scope creep in which the data collection remains fixed but the number of questions silently multiplies during analysis.
Analysis can experience scope creep too
Suppose a study initially plans one primary analysis. During analysis, the team tries several outcome definitions, multiple subgroup splits, different covariate sets, alternative exclusion rules, several transformations, and numerous interaction terms.
Some sensitivity analyses may be methodologically necessary. Others may be exploratory. The concern is not the number of models itself but whether analytical expansion is driven by a clear purpose and reported transparently.
A useful analytical discipline is to ask what each additional analysis is intended to test:
Does it evaluate robustness? Address an assumption? Investigate a prespecified secondary question? Explore an unexpected pattern?
If the answer is simply "we kept trying analyses until something interesting appeared," the project has moved beyond ordinary scope refinement.
Qualitative research needs boundaries without pretending the design cannot evolve
Some qualitative methodologies are intentionally iterative. Emerging findings may shape later interviews, sampling, observation, or theoretical attention. Preventing scope creep does not mean imposing an inappropriate fixed-design model on such research.
Instead, distinguish methodologically expected iteration from unrelated expansion.
If theoretical sampling introduces participants because emerging categories require further development, that evolution may be intrinsic to the methodology. If the project suddenly begins examining an unrelated organizational issue merely because participants mentioned it, the addition needs a stronger rationale.
An audit trail can document how the inquiry evolved and why particular decisions followed from the methodological logic of the study.
Set decision points rather than renegotiating scope every day
Research teams can reduce reactive expansion by establishing planned moments at which scope is reviewed.
For example, the team might review scope after pilot testing, after an initial recruitment period, before closing data collection, and before finalizing the analysis plan. The appropriate points depend on the methodology.
Between those checkpoints, new ideas can be documented without immediately becoming study changes.
This approach provides room for adaptation while preventing every new observation, paper, or meeting from reopening the entire design.
Use explicit criteria for accepting a scope change
A simple decision rule can make change control more consistent.
| Question |
If the answer is weak... |
|
Is the addition necessary to answer or protect the central research question?
|
Consider leaving it outside the current study |
|
Does it address a meaningful new scientific, methodological, or ethical issue?
|
Do not add it merely because it is interesting |
|
Can the existing design generate adequate evidence for it?
|
Do not make claims the study was not designed to support |
|
Can the project absorb the added time, recruitment, measurement, analytical, and reporting burden?
|
Reduce another demand, obtain additional resources, or defer the addition |
|
Can required approvals be obtained before implementation?
|
Do not implement the change prematurely |
|
Can the change be documented transparently?
|
Reconsider why the project needs it |
A proposal does not need to satisfy every criterion in exactly the same way. The framework is meant to force consideration of the whole study rather than the attractiveness of one addition in isolation.
Some requests should become another study
The most productive response to scope creep is sometimes not "no" but "not in this project."
Suppose a student survey generates an important question about faculty policy. That could justify a subsequent faculty study. An unexpected quantitative pattern could motivate qualitative follow-up. A local finding could lead to a multi-institution replication.
Separating those questions can allow each one to receive the design, population, evidence, and analysis it deserves.
When a proposed expansion requires substantially independent research questions, populations, methods, or analytical structures, ask whether the project has reached the point where expansion should become several studies.
Preventing scope creep is ultimately about protecting the answer
Scope control can sound like project administration: deadlines, task lists, protocols, and change logs. Those things help, but the methodological purpose is more important.
Every study has finite capacity. When that capacity is spread across too many questions, measurements, populations, sites, and analyses, the project may become less capable of answering any one question rigorously.
The purpose of controlling scope is therefore not to protect the original proposal from all change. It is to protect the relationship among the research question, evidence, methodology, analysis, and claims.