<![CDATA[PT Adhi Surya Amanah (ASA Trans), a domestic B2B freight forwarding company in Indonesia, faces persistent inefficiencies in its pricing process that hinder operational speed, accuracy, and competitiveness. The company relies heavily on manual quotation preparation using fragmented information spread across spreadsheets, emails, and handwritten records. This results in inconsistent pricing decisions, long quote response times, and a heavy reliance on staff interpretation, limiting the company's ability to respond efficiently to customer demand in an increasingly competitive logistics environment. These inefficiencies pose both scientific and managerial challenges, as accurate and timely pricing is a key factor in determining profitability and market performance in freight forwarding. Therefore, this study aims to develop, test, and implement a data-driven price forecasting infrastructure that can increase efficiency, reduce human error, and improve strategic decision-making at ASA Trans. The research begins by identifying the root causes of inefficiency through a detailed assessment of ASA Trans’s operational workflow. This evaluation identified four key challenges: unstructured and fragmented pricing information, manual verification, lack of access to real-time cost and market data, and over-reliance on subjective judgment. All of these issues lead to decreased operational reliability and limit the company's ability to grow. To address these challenges, this paper proposes a forecasting model to be developed based on the premise that ASA Trans' historical records of freight, cost, and pricing information are sufficient to provide the operational trends necessary for accurate forecasting. The study also assumes that market factors such as fuel and toll prices are subject to specific, predictable trends. The objectives of this study include determining the data requirements for price prediction, comparing forecasting techniques, creating a practical prototype of a price forecasting system, and integrating the system into ASA Trans' workflow. To achieve these objectives, the researchers employed a multi-stage methodology. Primary data was collected using internal freight data, operational cost records, and historical prices, while secondary data was collected in the form of freight benchmarks and competitor rate cards. Analysis Phase: The machine learning process involved is structured and includes data cleaning, standardization, exploratory data analysis, model development, and validation. Supervised learning technology was applied to three models: Multiple Linear Regression, Random Forest, and LightGBM to determine which approach provided the most accurate and operational predictions. Model evaluation is conducted using MAE, RMSE, and MAPE, supported by business-oriented validation criteria that measure quotation speed, usability, and alignment with financial objectives. The findings indicated that the best model was the Random Forest model, which had the lowest error and the most consistent results across various shipping situations. The fact that this model can capture nonlinear relationships makes it a favorable solution for the multifaceted and cost-based structure found in the Indonesian domestic freight market. The implementation of the Random Forest model was realized in a web-based Price Forecasting System created at ASA Trans. The implementation phase transformed the model into a working tool that can input estimated shipping costs, estimated operational costs, and estimated lead times with minimal user input. Test results showed that bid completion time was reduced from several hours to minutes, indicating significant improvements in operational efficiency. The fact that this system improves consistency, reduces administrative workload, and enhances the accuracy of pricing decisions was confirmed by user feedback during the training and socialization phases. This system is also capable of providing structured and continuous cost details, which increases transparency and enhances customer trust. This research contributes to scientific knowledge by demonstrating the practical application of machine learning in pricing optimization for domestic freight forwarding. It also establishes an empirical foundation for integrating predictive analytics into pricing workflows in small to mid-scale logistics companies in emerging markets. In the case of ASA Trans, the forecasting system is a strategic move towards digitalization, which will have long-term advantages with better pricing management, uniform decision-making, and preparedness to undertake an automation initiative in the future, like a more comprehensive Transportation Management System (TMS).]]>
