Abstract

Purpose: This study aims to extract strategic insights into startup success by applying interpretable machine learning techniques within the context of Management Technology and entrepreneurial strategy. It addresses the challenge of balancing predictive accuracy with transparency by incorporating explainable artificial intelligence (XAI) into the model development process. Research design, data and methodology: Utilizing data from 923 startups listed on Crunchbase, the study focuses on key features such as total funding, team size, investor relationships, investment stages, industry sector, and geographic distribution. Three machine learning models—Logistic Regression, Random Forest, and XGBoost—were employed to classify startup success. To ensure interpretability, SHAP (Shapley Additive Explanations) was used for both global and local explanations of model predictions. Results: Among the models, XGBoost demonstrated superior predictive performance with an accuracy of 84% and an AUC-ROC score of 0.90. SHAP analysis revealed that total funding, professional relationships, and number of funding rounds were the most significant predictors of success, while industry type and location had a marginal influence. Conclusions: This research presents a replicable, data-driven framework that integrates predictive analytics with interpretability. The results offer actionable implications for founders, investors, and policymakers involved in startup incubation, venture capital, and entrepreneurial ecosystem development.