A Machine Learning-Based Untapped Potential Index for Regional Photovoltaic Sales in Poland

The rapid growth of the photovoltaic (PV) market in Poland has increased the importance of data-driven sales management and strategic resource allocation in renewable energy companies. Despite dynamic market expansion, significant regional differences in PV adoption indicate the existence of underutilized sales opportunities. This study aims to develop a decision-support framework for identifying company-specific regional gaps between actual and model-predicted photovoltaic contract volumes. The analysis is based on more than 15,000 real contracts obtained from a large PV installation company and enriched with regional demographic, economic, energy-related, housing, and regulatory data. Several machine learning models were evaluated, including Random Forest, XGBoost (XGB), HistGradientBoosting, and CatBoost. To improve transparency, explainable artificial intelligence (XAI) techniques based on SHapley Additive exPlanations (SHAP) were applied. Additionally, a model-based Untapped Potential Index (UPI) was developed as an operational indicator of the company-specific gap between actual and expected regional sales performance. The results indicate that CatBoost achieved the highest predictive accuracy and enabled the identification of regions with the largest positive company-specific sales gaps. The proposed UPI supports the optimization of marketing activities, sales resource allocation, and regional expansion strategies within the analyzed company. The broader methodological framework may be extended to regional market assessment through recalibration and external validation using independent market data.