When a student takes six years to finish a four-year degree, the institutional conversation is usually about the student — motivation, preparation, circumstances. Sometimes that is right. Often the student did everything asked of them and the required course was full in the only two terms they could take it.
We wanted to know how much of the delay is structural rather than individual, because the two call for completely different responses and only one of them is cheap.
Delay concentrates in very few courses
Across 11 programmes, the four highest-impact courses per programme explained a median 34% of all delayed completions. In the most extreme case a single second-year mathematics course accounted for 19% of delay in its programme by itself. This concentration is the good news: a structural problem with four courses is tractable in a way that a diffuse student-readiness problem is not.
Capacity, not failure
We decomposed each bottleneck into three mechanisms: students failing and repeating, students unable to register because the course was full, and students blocked by prerequisite sequencing that made the course available in only one term per year.
In 7 of 11 programmes, capacity shortfall contributed more delay than failure did. This surprised us and it surprised most of the institutions. Failure is visible — it generates a grade, a conversation, a repeat enrolment. A student who could not register generates a null: an absence in the enrolment table that no report was looking for.
| Mechanism | Median share of delay | Range | Typical remedy | Cost |
|---|---|---|---|---|
| Capacity shortfall | 41% | 18–62% | More sections, or better forecasting | Moderate |
| Failure and repeat | 33% | 14–58% | Academic support, curriculum review | High |
| Prerequisite sequencing | 19% | 6–34% | Offer in both terms; relax chains | Low |
| Administrative hold | 7% | 1–21% | Earlier fee intervention | Low |
A student who fails a course leaves a record. A student who could not register leaves an absence, and nobody’s dashboard was watching for it.
Forecasting demand is the cheap fix
Most timetables we saw were built from last year’s enrolment, which systematically under-provisions courses whose demand is growing and over-provisions the reverse. Demand for a gateway course in a given term is fairly predictable from the cohort progressing toward it, the repeat population, and the declared-major mix — all of which the institution already knows.
A cohort-progression forecast reduced median absolute error from 14.2% to 5.6% against the last-year baseline. That is enough to change a sectioning decision, which is the only reason the number matters.
The intervention
Two institutions acted on the analysis for the 2025–26 intake, adding sections for their top two bottleneck courses and offering one previously annual course in both terms. On-time completion for affected cohorts rose 7.8 points against a matched comparison drawn from adjacent cohorts and comparable programmes.
We should be careful about that number. Two institutions is not a study, the intervention was chosen by them rather than assigned, and it coincided with other changes we could only partially control for. We report it as encouraging evidence for a mechanism, not as a settled effect size.
Method
Limitations
Attribution by counterfactual simulation assumes a student would otherwise have progressed normally, which over-attributes delay to a bottleneck course for students who had several difficulties at once. Six institutions, all partners, all with reasonably complete enrolment data — institutions with worse records probably have worse bottlenecks and are invisible here. We cannot see courses students wanted and never attempted, which likely understates the capacity mechanism further. And the intervention result rests on two self-selected institutions.
The sequence-tracing and forecasting code is available, along with a simulated corpus that reproduces the concentration finding: team@eduplatter.com.
Related work
Bailey, Jaggars and Jenkins [1] argue that curricular structure, not student deficiency, explains much of the completion problem — our decomposition is an attempt to put numbers on that claim in an Indian and international context. Attewell et al. [2] and Complete College America [3] document how remediation and prerequisite chains extend time to degree. Where we differ from most of this literature is in separating capacity shortfall from failure, which requires enrolment-attempt data that studies working from transcripts alone cannot see.
Artefacts
Everything below is published or available on request. A number nobody can reproduce is an advertisement, not a result. Real institutional records are never shareable under our processor obligations, so where that applies we release a simulated corpus that reproduces the qualitative finding.
References
- Bailey, T., Jaggars, S. S., & Jenkins, D. (2015). Redesigning America's Community Colleges. Harvard University Press.
- Attewell, P., Lavin, D., Domina, T., & Levey, T. (2006). New Evidence on College Remediation. Journal of Higher Education, 77(5).
- Complete College America (2012). Remediation: Higher Education's Bridge to Nowhere.
- Tinto, V. (1993). Leaving College: Rethinking the Causes and Cures of Student Attrition. University of Chicago Press.
- Rosenbaum, P. R., & Rubin, D. B. (1983). The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika, 70(1).
Bhardwaj, M. (2026). Course bottlenecks and on-time completion: where students actually get stuck. EduPlatter Research, EPR-2026-01.
