Evaluating the Effect of Human Capital on Startup Scalability Using Random Forest Regression and Variable Importance Analysis

Abstract

The objective of this study was to evaluate the effect of multidimensional human capital components on startup scalability by applying Random Forest regression modeling and variable importance analysis within the startup ecosystem of Tehran. This applied quantitative study employed a cross-sectional explanatory–predictive design. The statistical population consisted of technology-oriented startups operating in Tehran, from which 214 startups were selected through purposive sampling based on operational maturity and measurable growth indicators. Data were collected using a structured questionnaire measuring human capital dimensions including innovation competence, technical expertise, managerial capability, learning orientation, workforce skill intensity, professional experience, entrepreneurial experience, team diversity, and educational background, alongside scalability indicators such as market expansion capability, operational flexibility, and scalable business model readiness. Content validity was confirmed through expert review, and reliability indices exceeded acceptable thresholds. Data analysis was conducted using Random Forest regression to model nonlinear relationships between predictors and scalability outcomes. Model performance was evaluated using cross-validation procedures, prediction accuracy metrics, and error estimation, while variable importance and partial dependence analyses were employed to identify the relative contribution and marginal effects of human capital factors. Results demonstrated strong predictive performance of the Random Forest model, explaining a substantial proportion of variance in startup scalability. Innovation competence emerged as the most influential predictor, followed by technical expertise, managerial capability, learning orientation, and workforce skill intensity. Human capital variables collectively showed significant nonlinear and interaction effects on scalability outcomes. Practical competencies and experiential knowledge exerted stronger predictive power than formal education level. Partial dependence analysis indicated threshold effects whereby increases in innovation and managerial capability significantly accelerated scalability after moderate competency levels were reached, confirming the multidimensional and synergistic nature of human capital influences on startup growth potential. The findings indicate that startup scalability is primarily driven by dynamic human capital configurations integrating innovation capacity, technical knowledge, managerial leadership, and organizational learning capability. Machine learning analysis demonstrates that scalable growth emerges from complex interactions among human competencies rather than isolated factors, highlighting the necessity of human-centered entrepreneurial strategies and talent development policies to enhance sustainable startup expansion.