Data-Driven Estimation of Heavy-Truck Residual Value at the Buy-Back

Cerquitelli, Tania; Regalia, Andrea; Manfredi, Emanuele; Conicella, Fabrizio; Bethaz, Paolo; Liore, Elena · 2020 · Crossref

DOI: 10.1109/access.2020.2998940

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of accurately estimating the residual value of heavy trucks at the end of leasing contracts (buy-back). Traditional methods rely on coarse benchmarks like mileage and age, which fail to capture the actual wear and tear of high-cost commercial vehicles. Motivated by the rise of connected vehicles and the availability of rich telematics data, the authors propose a data-driven, unsupervised methodology to assess vehicle degradation. This approach aims to provide leasing companies with objective, interpretable insights to optimize marketing strategies, maintenance decisions, and pricing for re-sold vehicles. The study utilizes a real-world dataset comprising more than 10,000 heavy trucks monitored over three years by a multinational corporation. The data includes 38 features, combining internal parameters (e.g., fuel consumption, engine braking, hours of usage) and external conditions (e.g., altimetry). The methodology consists of five steps: data exploration using boxplots to identify asymptotic limit values; correlation analysis and feature selection using Pearson correlation coefficients to reduce the dataset to nine significant variables; residual value estimation using two distinct formulas; outlier detection using the DBSCAN algorithm to identify trucks requiring immediate maintenance; and visualization via an interactive Power BI dashboard. The core contribution is the development of two estimation formulas. The "data-driven" formula calculates a score based on the ratio of measured parameter values to their mechanical asymptotic limits, amplifying the impact of degradation. The "business-driven" formula applies expert-defined weights to parameters based on their perceived impact on vehicle usury (e.g., prioritizing fuel consumption and travel hours). Both formulas account for depreciation during periods of inactivity. Experimental evaluation demonstrated that the proposed scores effectively differentiate trucks that traditional metrics treat as identical, revealing significant variations in actual usage and wear. The DBSCAN algorithm successfully identified outliers corresponding to the most degraded vehicles. The significance of this work lies in its deployment within a private company’s data analytics services, validating its practical utility. The results confirm that leveraging fine-grained telematics data provides more accurate residual value estimates than traditional benchmarks. This enables leasing companies to categorize trucks into basic, comfort, or premium classes, facilitating targeted marketing and differentiated leasing pricing. The study highlights the value of integrating data science with domain expertise to transform raw vehicle data into actionable business intelligence, improving revenue generation and operational efficiency in the automotive industry.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• exposure measurement