What Parts of a Research Project Usually Take Longer Than Expected?

Which parts of research are easiest to underestimate?

A research timeline can look perfectly reasonable on paper. Two weeks to finalize the instrument. A month for approval. Six weeks for recruitment and data collection. Three weeks for analysis. Another month to write everything up.

Then the project begins.

The instrument needs another round of revision. Approval requires clarification. The organization that agreed informally to provide access now needs a formal request. Recruitment produces fewer eligible participants than expected. Several interviews are rescheduled. The dataset contains inconsistencies. Analysis raises questions that require additional checking. Your supervisor returns the manuscript with comments that cannot be addressed in an afternoon.

There is no universal ranking of the slowest research tasks because studies differ enormously. However, certain activities are especially easy to underestimate because their duration depends on uncertainty, iteration, other people, or problems that become visible only after the work begins.

Expect uncertainty around approvals, access, recruitment, data, analysis, and revision

The slowest part of research is often not the part that looks largest

Researchers naturally estimate visible work. If you plan 30 interviews, you can estimate how long each interview will take. If you have 500 survey responses, you can imagine the analysis. If you need a 20,000-word thesis, the writing workload is obvious.

Less visible work is easier to omit.

An interview requires recruitment, scheduling, possible rescheduling, consent, recording, file handling, transcription or preparation, checking, organization, and eventually analysis. A survey requires more than distributing a link. A statistical analysis requires more than clicking the command that produces the final model. A thesis requires more than generating its first complete draft.

This distinction helps explain why research schedules can be optimistic even when no individual estimate appears absurd. The project contains more elapsed time, iteration, coordination, and recovery work than the initial task list represented.

Ethics review may involve more than waiting for a decision

Where ethics review is required, researchers sometimes estimate it as a single waiting period: submit application, wait a few weeks, receive approval.

The actual process may contain additional steps. The application must first be sufficiently complete for review. Reviewers may request clarification or modifications. Revised documents may need to be submitted and checked. Depending on the applicable process, further review may be required before approval becomes effective.

Under U.S. Department of Health and Human Services regulations, for example, an Institutional Review Board can approve research, require modifications to secure approval, or disapprove it. Official HHS guidance also explains that when an IRB cannot make the determinations necessary for approval, it may require revisions or additional information, and the research cannot proceed until the required review and approval have occurred.

This is one regulatory system, not a universal timeline. Ethics procedures differ across institutions, jurisdictions, and types of research. The planning lesson is broader: do not estimate only the nominal review interval. Consider the time required to prepare a reviewable application and the possibility of revision before the relevant research activities can begin.

Institutional access can become a separate approval process

Ethics approval and permission to enter a research setting are not necessarily the same thing.

A study involving a school, hospital, company, government agency, community organization, archive, laboratory, or other institution may require authorization from one or more gatekeepers. A secondary-data project may require a data-use agreement, security review, application, payment, or approval from a data custodian.

Informal interest can also be misleading. Someone saying, “That sounds like a useful study” is not necessarily equivalent to having authority to provide access.

The time required may depend on organizational procedures that are invisible to the researcher when the project is first planned. This is why ethics approval, recruitment, data access, and other dependencies should be investigated early rather than placed on the timeline as generic administrative tasks.

Developing the instrument may take several iterations

“Prepare questionnaire” can look like a one-week task until the researcher begins deciding what each construct means, whether an existing measure is suitable, whether permission is required, how items should be worded, whether translation is necessary, how the survey should flow, and what pilot testing reveals.

The same applies to interview guides, observation protocols, extraction forms, laboratory procedures, coding manuals, and other data collection tools.

Instrument preparation can involve:

• reviewing existing measures or procedures;
• checking appropriateness for the population and setting;
• obtaining permissions where required;
• drafting or adapting content;
• expert or stakeholder review where appropriate;
• translation or cultural adaptation where relevant;
• pilot or pretesting;
• technical configuration; and
• revision after problems are identified.

Not every instrument requires all of these steps. The important planning error is assuming that the first version is automatically the final version.

Recruitment is frequently governed by a rate, not a target

A sample-size calculation or qualitative sampling plan tells you approximately how many participants you intend to recruit. It does not tell you how quickly those participants will appear.

Suppose you need 150 participants. If you can enroll 15 eligible participants each week, recruitment requires about ten weeks. If the actual rate is eight per week, the same target takes almost nineteen weeks.

The difference can arise from eligibility criteria, response rates, gatekeeper access, competing demands on participants, inconvenient scheduling, seasonal variation, study burden, geographic limitations, or recruitment channels that perform less effectively than expected.

For this reason, clinical research guidance from the U.S. National Institute of Diabetes and Digestive and Kidney Diseases uses interim enrollment targets as examples of project milestones. The particular percentages used in that context are not universal requirements. The useful principle is to monitor the recruitment rate before the scheduled recruitment period is almost over.

Scheduling participants creates elapsed time that the protocol may not show

An interview may last 60 minutes, but completing ten interviews does not necessarily require only ten hours.

Participants need to be contacted. Availability must align. Some will not respond. Others will cancel or reschedule. Interviews may need to occur outside the researcher's preferred working hours. Rooms, laboratories, equipment, interpreters, or other resources may also need coordination.

Longitudinal research adds another constraint. If participants must be followed for six months, the final participant recruited still requires the full follow-up period. Recruitment finishing late can therefore extend the project considerably beyond the date of the last enrollment.

Estimate elapsed calendar time, not merely participant-contact hours.

Data collection can be slower because of the procedures around each observation

The apparent unit of data collection often understates the actual workload.

A laboratory measurement may require preparation, calibration, cleaning, documentation, and quality checks. Field observations may involve travel and access windows. Interviews create recordings and notes that must be handled securely. Record extraction may require navigating multiple systems and resolving ambiguous entries.

Researchers should therefore estimate the complete collection workflow:

prepare → collect → document → transfer or store → check → resolve problems

If only the central collection activity is timed, the project plan excludes work that must happen for the resulting evidence to remain usable.

Data cleaning can take much longer than expected

“Clean data” often appears as one small task between collection and analysis. Depending on the dataset, it can become a substantial stage.

Potential issues include duplicate records, impossible values, inconsistent coding, incorrect dates, unexpected missingness, mismatched identifiers, variable-label errors, problems merging files, survey-branching mistakes, transcription errors, inconsistent qualitative metadata, or discrepancies between source records and extracted data.

Some problems can be corrected quickly. Others require returning to original records, consulting the data collector, reconstructing processing decisions, or deciding how the issue should be handled analytically.

Cleaning also includes documentation. An analysis-ready dataset should not depend on one researcher's memory of what columns named q7_r2_final2 happen to mean.

For larger or more complex projects, it is useful to make “data ready for analysis” a separate research project milestone rather than assuming that analysis begins immediately when collection ends.

Transcription and qualitative data preparation can become major workloads

An hour-long interview does not necessarily create one hour of downstream work.

Audio may require transcription, checking, de-identification, formatting, organization, and linkage to field notes or participant information. Automated transcription can reduce some manual effort, but accuracy still depends on audio quality, speakers, accents, terminology, language, and the requirements of the study. Sensitive data may also restrict which services can appropriately be used.

Qualitative researchers should estimate the entire path from recorded conversation to material ready for analysis, not merely the interview duration.

Where analysis begins iteratively during data collection, some of this work can overlap. That can reduce the end-of-collection backlog, but it does not eliminate the workload.

Analysis is often underestimated because the final procedure looks short

Running a statistical model may take seconds. Reaching the point where that model is appropriate can take considerably longer.

Quantitative analysis may involve checking distributions, coding variables, constructing scores, examining missing data, testing assumptions, choosing defensible model specifications, diagnosing problems, conducting sensitivity or robustness analyses, producing tables and figures, and verifying that the results have been interpreted correctly.

Qualitative analysis may involve repeated reading, coding, memoing, comparison, category or theme development, attention to divergent evidence, discussion among researchers, and refinement of interpretations.

Mixed-methods research may add another layer: the quantitative and qualitative findings must be integrated in a way that addresses the study's mixed-methods purpose rather than merely reported in adjacent sections.

The less familiar the method, software, data structure, or analytical problem is to the research team, the less defensible it is to estimate only the time required for the final command or coding pass.

Learning a method is different from applying a method

A project timeline can quietly assume expertise that the researcher does not yet have.

If you need to learn multilevel modeling, qualitative comparative analysis, structural equation modeling, a new programming language, specialized laboratory procedures, geographic information systems, or another unfamiliar technique, the learning period belongs in the project schedule.

That time may include reading methodological literature, completing training, practicing on sample data, consulting an expert, troubleshooting code, and revising the analysis after feedback.

This is particularly important for graduate research. “Analyze data for two weeks” may be plausible for a familiar procedure and wholly unrealistic when the first several days are spent discovering what the software expects the data to look like.

Writing usually involves more than producing the first draft

Researchers frequently underestimate writing because they estimate drafting rather than completion.

A research output may pass through:

These steps can overlap, but they consume time. A first draft is a milestone worth celebrating, particularly after a long thesis, but it is not the same milestone as “ready to submit.”

Feedback takes both waiting time and response time

Researchers often schedule the time another person needs to review something but forget the time required to respond to the review.

Suppose a supervisor needs ten days to review a thesis chapter. The student receives detailed comments and requires another week to revise it. A second review follows. What looked like a ten-day review has become a multi-week feedback cycle.

The same can happen with coauthors, statisticians, methodological advisers, editors, institutional offices, and project partners.

Plan the cycle:

send → wait for review → interpret feedback → revise → verify or resubmit

If the reviewer is essential to a later milestone, their turnaround time becomes part of the project's dependency structure rather than merely an inconvenience.

Collaborative decisions can take longer than individual decisions

A research team may need to agree on analytical choices, coding interpretations, manuscript framing, authorship responsibilities, revisions, or responses to external feedback.

Collaboration can strengthen research, but coordination requires time. Different schedules, disciplinary perspectives, institutional responsibilities, and interpretations can turn what appears to be a short decision into several rounds of discussion.

For multi-investigator projects, include coordination time explicitly rather than assuming that every decision occurs instantly because everyone has access to email.

Administrative completion can outlast the scientific work

A thesis may be scientifically complete but still require examination, revisions, signatures, formatting checks, repository deposit, clearance, or other institutional procedures.

A funded project may require final reports, data documentation, archiving, or other closeout obligations. A manuscript may need reporting checklists, author declarations, supplementary files, figure preparation, and final coauthor approval before submission.

The relevant procedures vary among institutions, funders, and journals. Verify them early enough that “research complete” and “requirement complete” do not unexpectedly become two different dates.

Unexpected findings can create legitimate additional work

Not every delay is a planning failure.

Analysis may reveal an anomaly requiring investigation. A participant-safety issue may require action. A measurement problem may need to be understood. Qualitative findings may indicate that additional sampling is appropriate under the methodology. A robustness analysis may reveal that the original interpretation needs reconsideration.

The objective is not to prevent all additional work. It is to avoid a timeline so tight that any scientifically responsible response to an unexpected problem makes completion impossible.

Some delays arise because an earlier decision was never really settled

A task can appear slow when the underlying problem is actually indecision.

Instrument development may stall because the construct has not been defined clearly. Analysis may stall because the research question and intended comparisons are still changing. Ethics preparation may stall because recruitment procedures remain unresolved. Writing may stall because the project has accumulated analyses without a clear connection to the original question.

Before concluding that a task simply “takes a long time,” ask whether it is waiting on a decision that should have been made earlier.

This is why settling consequential decisions before the study begins can save substantial time later even when the planning stage initially feels slower.

The tasks most likely to delay the project share recognizable characteristics

Instead of memorizing a list of supposedly slow activities, look for the properties that make duration difficult to predict.

These characteristics provide a better basis for schedule caution than assuming that every project will experience the same bottleneck.

The activities that take longest are not always the activities that matter most to the deadline

Duration and schedule impact are different.

A six-month literature review that overlaps with other activities may not determine the completion date. A two-week delay in data access might prevent the entire study from beginning.

When assessing risk, ask both:

How long might this take?

and:

What cannot happen until it is finished?

This dependency perspective is central when working backward from a submission or graduation deadline. The schedule should protect the activities capable of moving the final date, not simply the activities with the largest nominal duration.

Past projects are often your best source of better estimates

If you have completed similar work before, compare the original estimate with the actual duration.

How long did recruitment really take? How many rounds of instrument revision occurred? How much time passed between sending a chapter and completing the resulting revision? How long did cleaning take relative to analysis?

Teams can maintain similar records across projects. Even simple actual-versus-planned comparisons can improve future estimates.

If you lack personal data, ask supervisors, research offices, collaborators, laboratories, service providers, or other people familiar with the local process. Their estimates may still be uncertain, but they are usually more informative than choosing a duration because it fits neatly into the calendar.

Use uncertainty to decide where buffer belongs

Knowing which tasks are likely to run long should change the schedule.

Do not merely note that recruitment is uncertain and then schedule the last participant on the latest date analysis could possibly begin. Protect the milestone. Do not assume that supervisor review will be instantaneous because no writing is occurring during the waiting period. Include the elapsed time.

The appropriate response is not necessarily to add the same number of weeks everywhere. Instead, build buffer into the research timeline around uncertain, externally controlled, difficult-to-recover, and dependency-critical activities.

See how a six-week data collection plan becomes a much longer project stage

The original six-week estimate was not wrong because an interview takes longer than an hour. It was incomplete because “conduct 20 interviews” represented only the visible center of a larger workflow.

A better timeline would estimate recruitment, scheduling, interviewing, data preparation, and analysis-related work separately, identify which can overlap, and leave contingency around participant availability.

Optimistic timelines usually omit work rather than miscalculate arithmetic

Estimate the workflow around the task, not just the visible task itself

When reviewing your timeline, look for activities represented by deceptively simple labels such as “get approval,” “recruit,” “collect data,” “clean data,” “analyze,” and “write.” Expand each one long enough to see the actual workflow and dependencies.

Then compare these risks with your major project milestones. A task deserves particular attention when its delay threatens a milestone that controls everything downstream.

If the timeline becomes much longer after making these hidden activities visible, do not immediately compress them again. First ask whether the project can be simplified without damaging the research question, whether additional resources are available, whether activities can legitimately overlap, or whether the schedule itself needs revision.

A realistic timeline can feel disappointingly long. That disappointment is still considerably cheaper than discovering the same information two weeks before submission.

Check the tasks most likely to undermine an optimistic timeline

Common questions about research activities that run late

Look for uncertainty, iteration, and dependencies when estimating research time

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.