Research Publications

The complete record.

Everything the group has published since 2019. Report numbers are permanent: when we revise a paper the number stays and the version increments, so a citation never breaks.

Full papers11With abstract, method, figures, limitations and artefacts.

Technical notes2Shorter write-ups of a single decision.

Other outputs4A preprint, a dataset, a position paper and a talk.

Corrections4Papers revised after a reader found an error.

Record

Everything, by year.

Full papers link through. Notes, preprints and talks are listed with venue and date — write to us for a copy of anything not linked.

2026

EPR-2026-04 Early-warning model calibration across 14 institutional contexts EduPlatter Modelling team · EduPlatter Research Vendors sell one retention model to every institution. We tested whether that can work by fitting on one context and predicting in another, fourteen… Paper16 July 2026

EPR-2026-03 What actually worked: efficacy of 220 student success initiatives EduPlatter Efficacy team · EduPlatter Research We assessed 220 initiatives across partner institutions using matched comparison. Fewer than half produced a measurable effect. The ones that did… Paper24 June 2026

EPR-2026-02 Equity gaps in early-warning systems: who gets flagged, and who gets missed EduPlatter Responsible Analytics team · EduPlatter Research Our own first-generation model under-flagged first-generation students by 11 points and we did not notice for two terms. This is how we found it,… Paper19 May 2026

EPR-2026-01 Course bottlenecks and on-time completion: where students actually get stuck EduPlatter Curriculum Analytics team · EduPlatter Research Institutions treat non-completion as a student problem. In our data it is more often a timetable problem. Four gateway courses accounted for a third… Paper12 April 2026

EPT-2026-06 Warm-starting a persistence model with one term of history EduPlatter Modelling team · EduPlatter Technical Note Practical procedure behind the warm-start result in EPR-2026-04. Technical note4 July 2026

— Why we stopped selling card-swipe integration EduPlatter Data Engineering team · Indian Higher Education Analytics Forum, Bengaluru Slides and transcript published. 22 minutes. Conference talkMay 2026

— simulated-contexts-v2 EduPlatter Modelling team · Public release Fourteen simulated institutional contexts for transfer research. DatasetMarch 2026

2025

EPR-2025-02 Data quality is the binding constraint, not model choice EduPlatter Data Engineering team · EduPlatter Research We compared six model families against six levels of data completeness on the same task. Moving from the worst model to the best gained 4 points.… Paper14 October 2025

EPR-2025-01 What students say against what they do: topic analysis on 340,000 feedback responses EduPlatter Student Voice team · EduPlatter Research We ran topic modelling over eight years of course feedback and compared what students complained about with what actually predicted their outcomes.… Paper20 August 2025

EPT-2025-04 A data dictionary template for Indian higher education EduPlatter implementation team · EduPlatter Technical Note The definitions we ask registrars and IR offices to sign, and why each one is contested. Technical noteNovember 2025

— Against automated academic consequence EduPlatter Responsible Analytics team · EduPlatter Research Why we refuse to wire a risk score to probation or funding withdrawal. Position paperSeptember 2025

— Suppressed demand: estimating courses students never attempted to take EduPlatter Curriculum Analytics team · Preprint, under review Method for recovering registration attempts that leave no record. Comments wanted. PreprintJuly 2025

2022

EPR-2022-02 Small cells, big leaks: reconstruction attacks on institutional returns, and what differential privacy costs M. Bhardwaj · EduPlatter Research Colleges publish the same students many times over: by programme, year, category, gender and result, in statutory returns, ranking submissions and… Paper8 December 2022

EPR-2022-01 Small sections, loud numbers: empirical Bayes shrinkage for course-level result analysis M. Bhardwaj · EduPlatter Research Colleges rank course sections, and often teachers, by pass percentage after every examination. At the enrolments most sections actually have, those… Paper17 March 2022

2021

EPR-2021-01 Doubly robust estimates of programme effects when uptake clusters by college M. Bhardwaj · EduPlatter Research A university rolls out an advising programme across its affiliated colleges, and some colleges take it up far more than others. The usual adjustments… Paper24 June 2021

2020

EPR-2020-01 When the term moved online: detecting and correcting dataset shift in early-warning models M. Bhardwaj · EduPlatter Research Every persistence model trained before March 2020 now scores students whose attendance means something different, whose fees are under new strain and… Paper18 September 2020

2019

EPR-2019-01 Time to departure: discrete-time hazard models for student attrition M. Bhardwaj · EduPlatter Research Most attrition models ask whether a student will ever leave. Advisors need to know who is likely to leave this term, while there is still time to do… Paper22 August 2019

Citing this work

BibTeX for the full papers.

@techreport{epr04,
  title  = {Early-warning model calibration across 14 institutional contexts},
  author = {{EduPlatter Modelling Team}},
  year   = {2026},
  number = {EPR-2026-04},
  institution = {EduPlatter Research}
}
@techreport{epr03,
  title  = {What actually worked: efficacy of 220 student success initiatives},
  author = {{EduPlatter Efficacy Team}},
  year   = {2026},
  number = {EPR-2026-03},
  institution = {EduPlatter Research}
}
@techreport{epr02,
  title  = {Equity gaps in early-warning systems: who gets flagged, and who gets missed},
  author = {{EduPlatter Responsible Analytics Team}},
  year   = {2026},
  number = {EPR-2026-02},
  institution = {EduPlatter Research}
}
@techreport{epr01,
  title  = {Course bottlenecks and on-time completion: where students actually get stuck},
  author = {{EduPlatter Curriculum Analytics Team}},
  year   = {2026},
  number = {EPR-2026-01},
  institution = {EduPlatter Research}
}
@techreport{epr02,
  title  = {Data quality is the binding constraint, not model choice},
  author = {{EduPlatter Data Engineering Team}},
  year   = {2025},
  number = {EPR-2025-02},
  institution = {EduPlatter Research}
}
@techreport{epr01,
  title  = {What students say against what they do: topic analysis on 340,000 feedback responses},
  author = {{EduPlatter Student Voice Team}},
  year   = {2025},
  number = {EPR-2025-01},
  institution = {EduPlatter Research}
}
@techreport{epr02,
  title  = {Small cells, big leaks: reconstruction attacks on institutional returns, and what differential privacy costs},
  author = {Bhardwaj, Mohit},
  year   = {2022},
  number = {EPR-2022-02},
  institution = {EduPlatter Research}
}
@techreport{epr01,
  title  = {Small sections, loud numbers: empirical Bayes shrinkage for course-level result analysis},
  author = {Bhardwaj, Mohit},
  year   = {2022},
  number = {EPR-2022-01},
  institution = {EduPlatter Research}
}
@techreport{epr01,
  title  = {Doubly robust estimates of programme effects when uptake clusters by college},
  author = {Bhardwaj, Mohit},
  year   = {2021},
  number = {EPR-2021-01},
  institution = {EduPlatter Research}
}
@techreport{epr01,
  title  = {When the term moved online: detecting and correcting dataset shift in early-warning models},
  author = {Bhardwaj, Mohit},
  year   = {2020},
  number = {EPR-2020-01},
  institution = {EduPlatter Research}
}
@techreport{epr01,
  title  = {Time to departure: discrete-time hazard models for student attrition},
  author = {Bhardwaj, Mohit},
  year   = {2019},
  number = {EPR-2019-01},
  institution = {EduPlatter Research}
}