MyPertamina Transaction Increase Strategy Through Optimizing Application Use at Gas Stations Based on Machine Learning

Abstract

Background: The digital transformation in the energy sector is accelerating cashless payment adoption. MyPertamina, PT Pertamina's mobile payment platform, has been deployed across gas stations nationwide. However, only 5.65% of 1.3 million registered users (September 2024) are active, and only 11.5% qualify as loyal users (≥4 transactions/month). This performance gap, which exists between registered, active, and loyal users, is the focus of this study, as existing research has overlooked the role of gas station operational governance in shaping transaction behavior.

Objective: This study aims to (1) identify factors influencing MyPertamina usage at DKI Jakarta gas stations, (2) develop a machine learning-based prediction model to classify transaction behavior (MyPertamina vs. cash), and (3) create a G-STIC framework to increase adoption, usage intensity, and loyalty.

Methods: A quantitative case study using the CRISP-DM framework analyzed secondary POS transaction data from 8,000 transactions (5,200 MyPertamina; 2,800 cash) at DKI Jakarta gas stations (2024). Stratified sampling was used, and the models—Decision Tree, Gradient Boosted Trees, and Decision Stump—were evaluated based on accuracy, precision, and recall.

Results: Gradient Boosted Trees achieved the highest accuracy (97.75%). Gas Station Type and Class showed the strongest correlations with MyPertamina usage, suggesting further investigation of the Gas Station Code correlation.

Conclusion: MyPertamina adoption is influenced by operational governance and service standards. The G-STIC framework provides actionable strategies for increasing digital transaction adoption, contributing to both academic literature and managerial practice in the energy retail sector.