Research Summary: Predicting student churn in subscription EdTech: explainable machine learning for improving retention

Research in Educational Technology (EdTech) demonstrates that machine learning models can effectively predict student churn in subscription-based platforms, offering a strategic approach to improve retention. An optimized XGBoost model achieved high accuracy and recall in identifying at-risk students, leveraging activity, engagement, and financial data.

Why it matters

Predicting customer churn is strategically vital for subscription-based business models across various sectors, as it directly impacts revenue stability and growth. Proactive retention strategies, informed by such predictions, can significantly reduce costs associated with customer acquisition and enhance overall profitability.

Key insights

• Student churn represents a significant revenue loss for subscription-based EdTech companies, making retention a critical factor.
• Retaining engaged learners is more cost-effective than acquiring new ones, highlighting the financial and operational value of churn prediction.
• A predictive churn model was developed using anonymized student data, encompassing activity, engagement, and financial features.
• Multiple machine learning algorithms (logistic regression, random forest, neural networks, XGBoost) were evaluated.
• The XGBoost (XGB) model demonstrated the best performance, achieving approximately 0.84 accuracy, 0.63 F1 score, and 0.85 recall.
• The study followed a CRISP-DM process for model development and evaluation.

Source

Educational Technology Research and Development — https://link.springer.com/article/10.1007/s11423-026-10695-y