THE ROLE OF CURRICULUM IN STUDENT PERFORMANCE ANALYSIS USING KPI AND Z-SCORE

Abstract

After the 2022 reform of academic programs, IT majors in universities across Kazakhstan started to reveal a pattern that was not immediately obvious. On paper, the curricula were updated, but in practice their internal structure did not always keep up with the growing complexity of courses, especially those related to AI. In the “Information Systems” program at Khoja Akhmet Yassawi International Kazakh-Turkish University, this became noticeable over time: students who were doing fine in the early semesters began to lose consistency later on. Not abruptly, but in a way that repeated often enough to raise concern. This led to a more detailed look at student performance data extracted from the Platonus system. The data itself turned out to be far from clean – there were gaps, inconsistencies between semesters, and a fair amount of noise. It took additional effort to bring everything into a usable form, and without that step, any conclusions would have been questionable. After applying Z-score normalization and calculating KPIs for key subjects, the picture became clearer. Across different cohorts, roughly 14–18% of students consistently appeared in what can be described as a higher-risk group. Importantly, this was not a one-time fluctuation but a pattern that kept repeating. It suggests that the issue is not tied to individual performance alone, but to the way the learning trajectory is structured. To address the delay between the emergence of problems and their detection, a small tool was built using Streamlit. Unlike static reports, it allows changes to be tracked over time and makes it easier to notice early deviations. In practice, this gave academic advisors in the Computer Engineering department more time to respond – often before the situation escalated to retakes. It does not solve the problem entirely, but it changes when it becomes visible, which turns out to matter a lot. Overall, the results point to a broader conclusion: predictive analytics in education is no longer just an additional layer. In this context, relying only on end-of-semester averages misses too much of what is actually happening.