03 · What You Need to Know
A Research Idea Should Have Conditions Under Which You Would Change Course
Scientific reasoning depends partly on allowing evidence to change decisions. That principle should apply not only to hypotheses but also to the studies researchers choose to conduct.
A research proposal is necessarily based on incomplete information. You estimate whether participants can be recruited, whether data will be usable, whether previous evidence leaves an important uncertainty, whether an intervention can be implemented, and whether the available design can answer the question. As planning proceeds, some of those estimates become better informed.
The rational response to new information is not automatically to preserve the original study. It is to update the decision.
This reasoning is closely related to the use of progression criteria in pilot and feasibility research. Methodological guidance recommends specifying criteria that help determine whether a future definitive study should proceed unchanged, proceed with modifications, or not proceed. The same logic can be applied more broadly when stress-testing a research idea before substantial investment.
Abandoning the Study Is Not the Same as Abandoning the Topic
This distinction matters.
Suppose you want to understand how generative AI affects university students' learning. Your first idea is to analyze historical institutional records, but you discover that AI use was never recorded reliably.
That may be a fatal problem for the proposed secondary-data study. It is not evidence that the broader research topic is worthless.
You might collect primary data, ask a different question, investigate another population, use another design, or postpone the question until appropriate evidence becomes available.
Abandoning a study
Deciding that a particular question-design-data combination should not proceed in its current form.
Abandoning a research area
Deciding that the broader phenomenon or topic is no longer worth investigating.
The first decision is far more common than the second.
The Research Question May No Longer Be Sufficiently Unresolved
A research idea can lose its justification because the evidence changes before you conduct the study.
Perhaps you initially found only a handful of relevant papers. A more systematic search later reveals several rigorous studies and a recent meta-analysis that already answer the central question with considerable consistency and precision.
Or perhaps another research team publishes a large, well-designed study while you are still developing your proposal.
That does not automatically make your work redundant. Replication, generalizability, methodological improvement, and resolving remaining uncertainty can still justify another study. But the rationale must be reconsidered against the evidence that exists now.
If the existing evidence already answers the important question sufficiently well, the responsible response may be to change the project rather than manufacture a smaller gap around the original idea.
The Expected Contribution May Become Too Small
A topic can remain important while a particular study becomes low value.
Suppose your proposed research asks whether a widely studied teaching strategy is associated with student satisfaction in another university. During planning, you discover numerous rigorous studies across comparable settings, with little reason to expect the relationship to operate differently in yours.
You can still conduct the study. The harder question is what anyone will know afterward that they did not know well enough before.
If the answer is very little, the problem is not feasibility. It is marginal contribution.
Research time is finite. A study that is possible is not automatically the study most worth doing.
The Central Assumption May Turn Out to Be False
Research ideas often depend on assumptions that initially seem reasonable.
You may assume that a database contains the required outcome, that an intervention can be delivered consistently, that a population can be accessed, that a measurement tool is suitable, or that a particular theoretical mechanism is plausible.
During planning, one of those assumptions may fail.
The important question is whether the assumption was merely convenient or structurally necessary. If the study can be redesigned without losing its central purpose, continue with the redesign. If the entire research question depends on the assumption being true, its failure may be a stop condition.
This is why explicitly identifying the assumptions on which a research idea depends is useful before substantial investment.
The Required Population May Be Inaccessible
Recruitment feasibility can turn an excellent question into an unrealistic study.
You may need participants with a rare characteristic, access through institutions unwilling to participate, or repeated follow-up from a population known to be difficult to retain. Preliminary recruitment information may show that the accessible population is much smaller than expected.
Low recruitment does not always require abandonment. You might add sites, extend recruitment, reduce unnecessary participant burden, or reconsider overly restrictive eligibility criteria where scientifically and ethically appropriate.
But there is a point at which realistic recruitment cannot produce the sample the study requires. If no defensible modification resolves that problem, continuing to optimize recruitment materials will not rescue the design.
The Data May Not Exist in a Form Capable of Answering the Question
Secondary-data projects are particularly vulnerable to this discovery.
A dataset can exist without containing the evidence your study requires. Essential variables may be absent, definitions may have changed, missingness may be extensive, measurements may be weak proxies, or temporal information may be insufficient for the intended inference.
Some problems can be cleaned, harmonized, modeled, or supplemented. Others cannot.
If the available data are too poor to support the central research question, the choices are to obtain better data, change the question, change the method, or stop the proposed study.
The fact that you have already spent months obtaining the dataset does not create information that the dataset does not contain.
The Method Required by the Question May Be Unavailable
Suppose your question requires evidence about change over time, but only cross-sectional information can realistically be obtained. Perhaps the causal inference you want depends on a comparison that cannot ethically or practically be created. Maybe the required instrument is unsuitable for the population and no defensible alternative exists.
A different method may solve the problem.
But if every feasible alternative changes the evidence enough that the original question can no longer be answered, the methodological constraint becomes a reason to change the question or abandon the current design.
Methodological flexibility is useful. Calling a different question the same question is not.
The Study May Be Ethically Unacceptable
Ethical constraints are not inconveniences to be engineered around until the original design survives.
A study may require withholding something participants should receive, collecting information whose risks cannot be justified, exposing vulnerable participants to excessive burden, or using data in ways inconsistent with consent, governance, privacy, or applicable requirements.
Sometimes ethical concerns can be addressed through redesign. Procedures can change, unnecessary data can be removed, safeguards can be strengthened, or another population can be considered when scientifically appropriate.
But if the scientific question can be answered only through procedures that are ethically unacceptable, the study should not proceed in that form.
Scientific value is part of ethical justification, not a license to override participant protection.
The Study May No Longer Be Feasible Within Available Resources
Time, funding, expertise, equipment, access, staffing, software, and institutional support can all determine whether a project can be executed rigorously.
Researchers sometimes respond to resource constraints by preserving the original scope and reducing quality elsewhere: fewer participants than required, abbreviated measurements, incomplete follow-up, inadequately trained data collectors, or analyses beyond the available expertise.
A better response may be to narrow the study.
If narrowing preserves the central question, the project may become stronger. If every feasible reduction removes the evidence necessary to answer the question, feasibility may have become a stop condition.
A Pilot or Feasibility Study May Tell You Not to Proceed
A feasibility study is not successful only when it gives permission for a larger study.
Its purpose is to resolve uncertainty about whether and how the future study can be conducted. Recruitment may prove inadequate. Procedures may be unacceptable. Intervention fidelity may be poor. Data collection may fail. The design may require changes substantial enough that the definitive study no longer makes sense.
Methodological guidance for pilot and feasibility trials recommends using progression criteria to inform decisions about whether to proceed, proceed with modifications, or not proceed. Importantly, such criteria should be interpreted with judgment rather than as simplistic mechanical thresholds.
A feasibility study that demonstrates that a larger study should not proceed may have prevented a much larger waste of participants, time, and resources.
Your Stop Condition Should Be About the Study, Not the Result You Prefer
Do not confuse a scientific stop condition with an unfavorable substantive result.
“I will abandon the study if the intervention does not work” is usually not an appropriate pre-study stop condition for a project designed to discover whether the intervention works. The uncertainty of the result is the reason for conducting the research.
Likewise, the possibility that the relationship you expect may not exist is not itself a reason to avoid the study if a credible null finding would still be informative.
A stop condition concerns whether the study can produce useful, ethical, interpretable evidence, not whether that evidence will agree with your hopes.
Some Problems Call for Repair, Others for Redesign, and Others for Stopping
| New information |
Possible response |
When abandonment becomes reasonable |
| Recruitment is slower than expected |
Investigate barriers, revise realistic timelines, or use justified additional recruitment strategies |
The accessible population cannot realistically produce an informative sample |
| Some data are missing |
Characterize missingness and use appropriate design or analytical responses |
Essential information is unavailable to an extent that prevents credible analysis |
| Preferred measure is unavailable |
Use a defensible alternative and reconsider the scope of the claim |
No available measure adequately represents the construct required by the question |
| Original method is infeasible |
Use another design capable of answering the question |
Every feasible alternative answers a materially different or insufficient question |
| Similar evidence is published |
Reassess the remaining uncertainty and value of replication |
The proposed study would add negligible information to an already strong evidence base |
| Project scope exceeds resources |
Narrow objectives, measures, sites, or procedures |
Feasible reductions remove the ability to answer the central question |
| Ethical concern emerges |
Redesign procedures or safeguards where possible |
The required research cannot be conducted with acceptable participant protections |
The decision is therefore rarely “persist or quit.” There is an important middle category: change the study.
06 · What This Means for You
Write Your Stop Conditions Before You Need Them
Before investing heavily in a research idea, ask what evidence would make you change course.
Do this while you are still relatively detached from the project. Once you have written half the thesis, recruited participants, built the dataset, learned specialized software, and explained the project repeatedly to colleagues, changing direction becomes psychologically harder even when the scientific evidence has not improved.
Your stop conditions do not need to be rigid numerical thresholds. They should identify the circumstances that would require serious reassessment.
A simple decision framework
If the problem is inconvenient but does not threaten the central question
Manage it without redesigning the entire project.
If the problem threatens one component but the central question can be preserved
Redesign that component and reassess the assumptions introduced by the change.
If a different design can answer the important question more defensibly
Change the study rather than protecting the original method.
If new evidence substantially reduces the expected contribution
Reassess whether replication, extension, or another unresolved question still justifies the project.
If the required evidence cannot be obtained ethically, feasibly, or with sufficient quality
Abandon or postpone the current study rather than lowering the evidential standard.
If only previous investment argues for continuing
Evaluate the project using future costs and expected future value instead of what has already been spent.
Use a Red-Amber-Green Approach When Exact Thresholds Would Be Artificial
Formal feasibility research sometimes uses progression criteria to support decisions about whether a larger study should proceed. These criteria can be adapted conceptually when evaluating an early research idea.
You might classify a critical dependency as green when available evidence supports proceeding, amber when modification or further investigation is required, and red when the problem is serious enough that the current design should not proceed.
| Status |
Meaning |
Typical response |
| Green |
The assumption or feasibility condition is sufficiently supported |
Proceed while continuing proportionate monitoring |
| Amber |
Important uncertainty or difficulty remains, but a credible remedy may exist |
Investigate, modify, obtain preliminary evidence, or create a contingency plan |
| Red |
The problem prevents a credible, ethical, feasible, or sufficiently valuable study under realistic conditions |
Fundamentally redesign, postpone, or abandon the current study |
The categories should be defined for the actual project rather than populated with arbitrary universal thresholds. Recruitment of 60% of target participants could be manageable in one study and fatal in another.
Ask Whether the Idea Can Be Saved Without Becoming a Different Idea
Research redesign often happens gradually.
The population changes because recruitment is difficult. The outcome changes because data are missing. The design changes because the preferred method is unavailable. The research question is softened to match what remains feasible.
Each modification may be defensible individually. Eventually, however, you should ask whether the resulting study still addresses the problem that originally justified the project.
If the contribution has disappeared during rescue, stopping may be more sensible than preserving the project's title while replacing everything underneath it.
Ask Someone With Less Investment in the Project
Researchers are not always ideal judges of when their own projects should stop.
A supervisor, collaborator, methodologist, statistician, subject specialist, or other informed colleague may identify options or problems that are difficult to see from inside the project.
Ask them not merely how to fix the study, but whether they think it should still be done.
A useful question is:
“Knowing what we know now, would you choose to start this project today?”
If the answer is no, ask what changed.
Separate the Decision From Your Identity as a Researcher
Researchers can become identified with a project surprisingly quickly. It becomes “my topic,” “my model,” “my intervention,” or “my framework.” Once that happens, evidence against the study can feel like criticism of the researcher.
It is not.
A research idea is a proposed allocation of future effort under uncertainty. It should remain revisable when the evidence changes.
The next question is therefore whether you are continuing because the project still has strong prospective value or because you have already invested too much to feel comfortable letting it go.
Watch Out
Do not define stop conditions so aggressively that every ordinary research difficulty becomes a reason to quit. Research contains uncertainty, imperfect information, recruitment challenges, unexpected findings, and methodological trade-offs. Stop conditions are most useful for problems that threaten the study's justification, ethical acceptability, feasibility, interpretability, or meaningful contribution, not for eliminating all inconvenience.
