Skip to Content

An automotive company used AI to forecast logistics lead time

The onset of the global pandemic has disrupted every form of business. Artificial Intelligence (AI) has emerged as a strong transformation driver of all industries, especially the automotive industry. AI is being effectively implemented in the automotive value chain. It is insightful to learn how AI can improve the accuracy of lead time forecasting for an efficient production and delivery processes, and eventually, a great customer experience.

Problem Statement

An automotive manufacturing company envisioned higher customer satisfaction through on-time deliveries of vehicles to customers. The company experienced delays in logistical deliveries to end customers. It was partly because of lower accuracy in forecasting the logistics lead time, resulting in missed customer deliveries as per promised schedules. The basis of the forecasts was pre-defined lead time, which was static. The low accuracy demonstrated the ineffectiveness in delivery operations and a poor customer experience. Moreover, using pre-defined lead time made it challenging to renegotiate the lead time contracts with the transport carriers.

The company wanted to understand how AI could help improve the lead time forecasts and consider various logistical dynamics. These included — vehicle batching, interim distribution centers, multiple transport modes (such as road, rail, and sea), and geographical dynamics. The company had historical data records describing the logistics details of every vehicle.  We, together with the company, set the ambition to design, build, and establish an AI model to forecast the logistics lead time for every vehicle shipped from the factory to its destination.

Solution

We came up with a transformative and transparent solution design. It combined the knowledge of logistics and AI algorithmic domains as well as assessed and selected suitable technology for easy implementation within the client’s technological environment. It had two parts — building an AI model and facilitating the consumption of AI insights.

1.Building an AI model

The model consisted of two steps:

  • A robust feature engineering pipeline: After a rigorous exploratory analysis, our findings concluded that eight out of 39 features were influential in forecasting the lead time accuracies. For example, the transport carrier dwell time feature had a greater weight in the forecast.
  • AI model: After extensive experimentation, the final AI model had an accuracy of more than 80% in forecasting the lead time of the vehicles (measured as the number of times the forecast date is equal to the actual delivery date, calculated by backtesting the AI model on the transport data).

2.Facilitating the consumption of AI insights

We facilitated and accelerated the consumption of AI insights through two steps:

  • Dashboard: Built a customizable front-end dashboard to visualize the AI results for logistics planners to facilitate decision-making. As a result, the logistics planners could proactively identify several vehicles that required interventions and risked not being delivered on time.
  • Enhancing adoption: Accelerated the AI model adoption through a learning journey, ensuring the logistics planners have the skills and right mindsets to help build trust in the AI models.

Business Impact

The AI model was a game-changer as it enabled the following:

  • A 24% annualized reduction in delayed vehicle deliveries seemed possible. With an estimate obtained by backtesting and identifying the number of vehicles that had the risk of running late, the logistics planners could intervene proactively to ensure timely deliveries.
  • Led to higher productivity in the logistics chain by empowering logistics planners to identify vehicles requiring intervention.
  • Led to improved customer satisfaction through reduced delivery delays.
  • New insights gained within the contracted lead times space, opened new possibilities for renegotiating the lead times’ contracts with transport carriers, enabling cost savings.

Through this unique approach of using AI in the logistics chain, the client offered a powerful example of the potential of AI-assisted logistics leads time predictions. They now have concrete validation that an AI approach may improve other aspects of the logistics chain as well. We also learnt the importance of using the logistics domain knowledge while developing AI models. Data revealed the historical patterns and showed the relationships among different variables. However, when we integrated the logistics experts’ knowledge with the technology development process, the resultant model was way more impactful.

About Author

Mukund Subramaniyan

Email | Linkedin | Twitter

Mukund works in the manufacturing sector, mainly advising clients on topics related to digital manufacturing, data analytics, artificial intelligence, and operations transformation. He has extensive experience working with major automotive companies in India and Sweden including original equipment manufacturers (OEMs) and suppliers. His educational and professional background integrates technical expertise as computer scientists with the manufacturing expertise of engineers. This ensures that the deep insights generated through AI translate into real measurable impact in an organization. He is passionate about transforming manufacturing operations using IIoT, AI, data, insights, and actions.