Data Analytics Innovations in U.S. Higher Education Decision-Making: A Critical Review of Predictive Modelling, Institutional Governance and Policy Implications
Noah Dzotepe *
East Tennessee State University, Johnson City, Tennessee, USA.
*Author to whom correspondence should be addressed.
Abstract
Colleges and universities in the United States now rely on predictive models, dashboards, automated outreach and algorithmic optimisation to inform decisions about admissions, financial aid, advising, course support and institutional planning. These tools are frequently promoted as instruments for improving retention, completion and efficiency, yet their adoption has outpaced evidence on whether better prediction produces better decisions or fairer outcomes. This critical narrative review examines how data analytics has been used in United States higher education decision-making, with particular attention to predictive modelling of student success, the institutional governance of analytics and the public policy environment that shapes both. Peer-reviewed literature, policy evaluations and official documents published between January 2007 and September 2026 were identified through structured searches of education, multidisciplinary and citation databases, supplemented by citation chasing and authoritative institutional sources, and were appraised for design quality, relevance and transferability. The synthesis indicates that models built from routine administrative records predict persistence and completion with moderate to good discrimination, but that risk rankings are unstable across specifications, accuracy is often lower for racially minoritised students and statistical bias-mitigation methods have not reliably removed these disparities. The evidence that analytics-triggered interventions improve outcomes is considerably weaker than the evidence for prediction. Randomised evaluations of technology-enhanced advising and of large-scale behavioural nudges have mostly reported null or short-lived effects, whereas intensive, well-resourced human support has produced larger gains. Governance research documents low student awareness of data use, professional resistance among advisers, uneven institutional capacity and growing dependence on commercial platforms, while federal privacy law offers limited guidance on predictive uses. Recent federal changes, including programme-level earnings accountability, a proposed expansion of admissions data collection and the prohibition of race-conscious admissions, are reshaping the data on which institutional models depend. Analytics appears most defensible when it is treated as a decision-support infrastructure subject to local validation, equity auditing, transparent governance and rigorous impact evaluation, rather than as a substitute for human support or a self-justifying innovation.
Keywords: Learning analytics, predictive analytics, algorithmic fairness, student success, academic advising, data governance, enrolment management, higher education accountability