Norman Müller, Peter Burggräf, Fabian Steinberg, Carl René Sauer, Maximilian Schütz
{"title":"An analytical review of predictive methods for delivery delays in supply chains","authors":"Norman Müller, Peter Burggräf, Fabian Steinberg, Carl René Sauer, Maximilian Schütz","doi":"10.1016/j.sca.2025.100130","DOIUrl":null,"url":null,"abstract":"<div><div>Predicting delivery delays is crucial for companies, especially in times of increasing global uncertainty and vulnerable supply chains. Machine learning (ML) offers significant potential to improve the forecast performance and quality of delivery delay prediction. Although various prediction approaches have been proposed in research, a structured and comprehensive overview is lacking. This paper addresses this gap by conducting a systematic literature review on the direct prediction of delivery delays. The objective is to identify applied prediction approaches and data sources, assess their readiness for real-world implementation, and derive a research agenda. The findings reveal that current research often focuses on marginal optimization of prediction performance while lacking practical applicability. Furthermore, most studies emphasize classifying deliveries as on time or delayed, rather than predicting the actual delay magnitude. Regarding the data used for prediction, combining enterprise resource planning (ERP) data with data from logistics improves prediction performance. However, environmental and location data, which could be easily integrated into ERP-based ML models, are rarely considered. This indicates a misalignment in current research, emphasizing the need for models combining practical applicability with predictive accuracy. Further research is required to address these identified deficits. Therefore, the present paper proposes a research agenda, to prioritize the most important deficits. These include, among others the industrial application, optimal prediction timing and ideal data combinations to achieve high prediction accuracy. It also highlights the need for integrated decision support systems that provide prediction-based recommendations, enhancing the practical value of predictive models in supply chain management.</div></div>","PeriodicalId":101186,"journal":{"name":"Supply Chain Analytics","volume":"11 ","pages":"Article 100130"},"PeriodicalIF":0.0000,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Supply Chain Analytics","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2949863525000305","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Predicting delivery delays is crucial for companies, especially in times of increasing global uncertainty and vulnerable supply chains. Machine learning (ML) offers significant potential to improve the forecast performance and quality of delivery delay prediction. Although various prediction approaches have been proposed in research, a structured and comprehensive overview is lacking. This paper addresses this gap by conducting a systematic literature review on the direct prediction of delivery delays. The objective is to identify applied prediction approaches and data sources, assess their readiness for real-world implementation, and derive a research agenda. The findings reveal that current research often focuses on marginal optimization of prediction performance while lacking practical applicability. Furthermore, most studies emphasize classifying deliveries as on time or delayed, rather than predicting the actual delay magnitude. Regarding the data used for prediction, combining enterprise resource planning (ERP) data with data from logistics improves prediction performance. However, environmental and location data, which could be easily integrated into ERP-based ML models, are rarely considered. This indicates a misalignment in current research, emphasizing the need for models combining practical applicability with predictive accuracy. Further research is required to address these identified deficits. Therefore, the present paper proposes a research agenda, to prioritize the most important deficits. These include, among others the industrial application, optimal prediction timing and ideal data combinations to achieve high prediction accuracy. It also highlights the need for integrated decision support systems that provide prediction-based recommendations, enhancing the practical value of predictive models in supply chain management.