Study on the use of machine learning in supply chain risk management

In cooperation with the company riskmethods, we have conducted a study on the use of machine learning in supply chain risk management. The results have been resented for the first time at the 56th BME Symposium on DIGITAL Purchasing and Logistics during the session “New ways in supply chain risk management: don’t be afraid of artificial intelligence! From analog to selflearning risk monitoring in the supply chain”.

We have shown how an automated SCRM system helps the company to create transparency in supply networks, and which potential success factors are compatible with the use of artificial intelligence in SCRM. Finally, I presented the study results.

In recent months, many companies have learned and adapted, and now can react faster to supply chain disruptions via preparation and extensive information provision. However, the results of the study on the use of machine learning in supply chain risk management show that 50% of respondents are still caught unprepared by a risk event, which can lead to financial and reputational damage as well as delays. 42% of respondents do not know the financial impact of a damaging event. Consequently, these companies are at high risk, as losses in the millions of dollars are not uncommon.

Only about half of the companies have a high or very high degree of risk
transparency for their own production facilities and direct suppliers (Tier-1). Subsuppliers (tier-2 or tier-n) are only monitored in a handful of cases, and, as such, SCRM only covers a small part of the supply chain. This increases the risk of production downtime, rising transportation costs, and the costly development of alternative procurement sources. One finding of the survey: looking at direct suppliers is no longer sufficient to identify critical nodes in the supply chain early enough to effectively implement preventative measures.

Consequently, the respondents see significant potential for improvement in their risk management, especially with regard to supply chain transparency and risk identification methods. For more than 80% of respondents, the early identification of risks is an added value which helps to initiate countermeasures in time and thus avoid disruptions in the supply chain. In many cases, this preventative time advantage can be achieved by automating risk identification (alerting), which generates analyzable real-time data from supply networks. Here, self-learning systems can offer significant contributions.

The study (in German language only) can be downloaded here.

A Systematic Investigation of the Integration of Machine Learning into Supply Chain Risk Management

Our article on “A Systematic Investigation of the Integration of Machine Learning into Supply Chain Risk Management” was published. The main objective of the paper is to analyze and synthesize existing scientific literature related to supply chain areas where machine learning (ML) has already been implemented within the supply chain risk management (SCRM) field, both in theory and in practice. Furthermore, we analyzed which risks were addressed in the use cases as well as how ML might shape SCRM. For this purpose, we conducted a systematic literature review. The results showed that the applied examples relate primarily to the early identification of production, transport, and supply risks in order to counteract potential supply chain problems quickly. Through the analyzed case studies, we were able to identify the added value that ML integration can bring to the SCRM (e.g., the integration of new data sources such as social media or weather data). From the systematic literature analysis results, we developed four propositions, which can be used as motivation for further research.

Schroeder, M. & Lodemann, S. (2021): A Systematic Investigation of the Integration of Machine Learning into Supply Chain Risk Management, in: Logistics 2021, 5(3), 62;

The Use of AI to Reduce Delivery Risks

The use of artificial intelligence, especially machine learning (ML), offers significant advantages for detecting delivery risks at an early stage.

The search for and contacting of suitable suppliers, as well as the time spent waiting on their respective answers, often consumes substantial working time and ultimately receives a low response rate. The use of machine learning can proactively reduce or avoid certain delivery risks (Cavalcante 2019; Hamdi et al. 2018). By analyzing historical data, recommendations for supplier efficiency and performance can be made, taking into account not only price, product quality, product quantity, and service but also the risk factors of uncertainty, vulnerability and possible supply disruptions. Non-qualified suppliers can thus be excluded in advance from bid solicitation and the process can be handled much more efficiently.

If the company is also able to foresee the failure of a supplier before its competitors do, the use of ML represents a competitive advantage. This is because the company can use the early information to look for an alternative supplier even before the competition becomes actively aware of the need to.

“If I tell you that we are already using AI, then I am triggering an avalanche and losing my competitive advantages.” (Deloitte 2020, p. 18)

It is therefore not surprising that only a few examples of the use of ML in supplier management can be found in the literature. Three examples are given below:

  • Volkswagen has developed a bidder-list-generator that uses machine learning to precisely determine possible suppliers
  • To be able to predict the reliability of deliveries, the automotive supplier Hella GmbH & Co. KGaA tested the use of ML techniques. Here, the receipt history of goods, purchasing and delivery data, supplier and material master data, etc. were used in ML in order to be able to predict the reliability of delivery quantities for a defined period of 30 days. (Huang 2020)
  • The LEAR Corporation, a globally active manufacturer of car seats and electrical distribution systems, also uses ML in five LEAR plants in order to be able to anticipate failures at an early stage. More than 30 influencing factors are included in the analysis to detect volume deviations during transport registration. All known transport orders are checked by the system for the following week. If the algorithm detects a risk, the persons concerned are automatically informed so that appropriate measures can be taken (Logistik heute 12/2019)

However, the current study by Deloitte (2020) once again makes it clear that the use of ML in supplier management is still in its infancy (p. 17).

Towards Supply Chain Risk Management 4.0 (SCRM 4.0)

What influence do the new technologies have on SCRM?
At our Institute for Logistics and Corporate Management (TU Hamburg), we are dealing with the question of what influences new technologies exert on SCRM, and specifically on the individual phases of the SCRM process: risk identification, analysis and assessment, management and risk control. The focus of the analysis is the application of artificial intelligence, in particular machine learning, as well as cloud technologies, sensors, and robotics. We categorize this topic as “Supply Chain Risk Management 4.0 (SCRM 4.0)“.

SCRM 4.0. can be defined as a building block of supply chain management which includes all strategies, processes, and related measures that are supported by IoT (Internet of things) and are both suitable for the reduction of risks within a supply chain and increase overall resilience.
If you are interested in our research, please contact us.

New applications for risk management through the use of Big Data

The increasing digitalization, combined with the generation of extensive amounts of data – also referred to as Big Data. Big Data presents new challenges to the traditional risk management implemented so far on the one hand, but on the other hand it also opens up potential for a broader practical application, especially with regard to risk identification and assessment, which has not been possible to this extent so far.

In processes such as machine learning – a subdomain of artificial intelligence – IT systems develop patterns and regularities for risks on the basis of these existing data sets and with the aid of algorithms, and make corresponding forecasts which are continuously adjusted (Beierle & Kern-Isberne 2003, p. 99f.). This enables more comprehensive early warning systems to be implemented and thus promotes the proactive handling of risks. In addition to accessibility, the validation of data plays a decisive role in this context.

Furthermore, the higher data availability is accompanied by improved transparency, which opens up new potentials for visualization (also called mapping). Thus, it is possible to visualize e.g. movements of persons or objects and to predict future changes in the process due to occurring risks by means of corresponding data sets.

The improvement in risk identification has also resulted in the development of automated analysis and evaluation procedures. The large volumes of data and the large number of sources of data supply require the use of algorithms and complex evaluation procedures in order to draw conclusions about the future development of risks (predictive analytics) and to be able to forecast the company’s optimum response to future risks (prescriptive analytics) (Krause & Pellens 2018, p. 51). As a result, strategies for dealing with risks can be decided at an earlier stage and precautionary measures can be initiated that are tailored to the specific situation. In addition, a better cost-benefit assessment can be made for individual measures, e.g. through better quantification of resources.

In our paper about New Applications for Risk Management through the Use of Big Data – Two Case Studies, we describe two practical examples of the use of Big Data in risk management. Firstly, the containment of epidemics through risk analysis using Big Data. Secondly, the reduction of the risk of pollution from ship hulls through the evaluation of satellite data on global sea conditions, as well as data sets on sea temperatures, cloud cover, biological factors, salinity, acidity and other minerals.

Extract from Schröder et al. 2019, Neue Anwendungsmöglichkeiten für das Risikomanagement durch den Einsatz von Big Data – Zwei Fallbeipspiele, in: Schröder, Meike; Wegner, Kirsten (Hrsg.) Logistik im Wandel der Zeit, Springer: Wiesbaden, pp. 121-136.


Suche in OpenEdition Search

Sie werden weitergeleitet zur OpenEdition Search