CNH, a leading manufacturer of agricultural machinery, revised their inventory management processes related to safety stock by implementing a predictive analytics tool designed to optimize safety stock levels. This innovative solution, developed by Superlinear in partnership with Flanders Make, replaces reactive, manual inventory practices with data-driven, automated optimization.
“This solution helps us to transition from a reactive to a more proactive approach. In this way, shortages of stock at the production line can be mitigated before they ever occur, rather than merely preventing a reoccurrence.” - Pieter De Pourcq, Operational Excellence and Business Transformation Manager at CNH
Executive Summary
Context
As a manufacturer of highly complex agricultural machinery, CNH needs to ensure the timely availability of over 30 000 different parts for each of their production sites. The Zedelgem plant experiences high costs related to uncertainties in the supply chain, such as supplier delays and varying demand. Ideally, safety stock forms a buffer to smooth out the different uncertainties, but at the moment this does not suffice. To ultimately let production continue as planned, expensive interventions are frequently needed: express deliveries of missing parts or delayed reworks of machines once the missing parts are available again.
Goal of the project
Show the potential value of replacing CNH Zedelgem’s current purely manual approach to safety stock management with an automated data-driven approach to optimizing inventory management.
Solution
Starting from an analysis of the high costs experienced today, Superlinear collaborated with CNH to create a solution that combines both forecasting and optimization technologies to automatically suggest safety stock settings for the wide variety of parts managed by CNH, balancing financial objectives with operational requirements. The end-result is a user-friendly interface that integrates seamlessly with CNH’s inventory management workflow.
Case Study
The challenge
For any manufacturing company, the goal is to stick to the production schedule at all times. Especially for factories with line production, if the line is stopped for any reason this results in big uncertainty on the impact and likely a very high cost to the company.
Producing highly complex agricultural machinery, CNH is what is called a high mix - low volume manufacturer. A high mix - low volume context comes with a number of specific challenges: for their supply of parts, this means their many different parts can not all be managed in the same way. More common parts might be used very regularly, while parts related to niche configurations might have very irregular usage. On the demand side, their low volumes and high variety in products means it is challenging to predict up front which exact configurations will be sold. Finally, in our global context factories can have suppliers all over the world, meaning long lead times. These can make it difficult to move quickly when the future need of specific parts is uncertain.
CNH Zedelgem runs up high inventory management costs due to costly interventions, such as express deliveries of missing parts and delayed reworks. These interventions are needed to deal with the uncertainties on both the supply and demand side described above, ultimately allowing the plant to stick to their production schedule.
The logistics team at CNH sees safety stock as the buffer to shield their production process from uncertainties in the supply of parts, internal production and market demand. However, today safety stock is managed manually. With a couple of hours per month to manage the safety stock for over 30 000 parts, the current safety stock positions are too static and unable to cover the needs of the moment.
The briefing
The leadership team at CNH Zedelgem sought an innovative partner to help transform their inventory management approach for safety stock. As a participant in the Flanders Make Accelerator Programme—designed to make manufacturing companies more resilient and sustainable at an accelerated pace—they proposed the project idea there. The project was followed up not only by CNH, but also regularly discussed with the other leading manufacturing companies involved. The programme received a ‘Flemish Resilience’ grant of 11 million euros, complementing the 50 million euros contributed by companies themselves—funding that helps companies become more resilient, remain competitive, and reduce their dependence on foreign suppliers.
After a thorough selection process, CNH decided to collaborate with Superlinear, a company specialized in AI, based on their long-term vision for applying predictive methods and optimization technologies in supply chains. This involved their experience with combining different AI technologies such as predictive modeling and optimization.
The work
To ensure we fully understood CNH’s context, we started by deep diving into their inventory analytics and processes for material planning, inventory management, and safety stock management today in a workshop with the logistics team. At this time, it was crucial to align on definitions of different concepts, ensuring that what was found in the data aligned with the understanding in the plant. A following data analysis exercise gave a view on where in the production process most drivers of uncertainties were located that might give rise to process disruption later on.
These exploratory exercises gave us the necessary insight to build a first proof-of-concept solution that could forecast future uncertainties in the plant’s inventory based on historical data & recommend safety stock settings accordingly. The first promising result led us to 2 more rounds of feedback and improvements together with the logistics team at CNH Zedelgem to take the solution from a proof-of-concept to a solution that can be used for dynamic inventory optimization in the daily operations of the plant.
Both the algorithm and the user interface were integrated into CNH’s cloud environment, ensuring access to the solution for the right people and access to the latest data for the algorithm. Next to that, additional data sources were integrated providing more insight into the impact on the operations of stockouts for specific parts. This data allowed further optimization of the recommended safety stock levels. Finally, a thorough data quality exercise was performed with the logistics team to remove irregularities from the solution.
Image 1: process of creating safety stock suggestions
In the current cloud application you can find safety stock suggestions which are a result of a 4-step inventory optimization method (depicted in image 1).
In the first step, we quantify the different drivers of uncertainty in the plant’s inventory based on various data streams from the plant. Examples of drivers are delivery delays and short-term demand increases.
In the second step, we estimate the future risk for these drivers of uncertainty based historical data available in the plant.
Next, we simulate the effect of the indicators on plant stock levels and how safety stock would mitigate stock shortages.
Finally, we use a mathematical model to optimize the suggested safety stock levels for the selected range of parts. The mathematical model balances the costs of holding safety stock with the value of holding safety. This results in the optimal allocation of budget over safety for different parts.
The impact
Based on a thorough analysis together with CNH Zedelgem we estimate that, while maintaining the same total value of safety stock, for inventory related issues we can:
Reduce express delivery costs by >10%
Reduce the frequency of missing parts on the line by >50%
With a runtime < 1 minute
“Superlinear truly partnered with us to find the best solution. They took our ideas and input seriously, actively engaging with them as we evaluated various potential enhancements. Throughout this process, they remained focused on our primary objective: reducing costs associated with missing parts and avoiding express transport expenses.” - Lies Pierloot - Manufacturing Value Chain Logistics Engineering
The future
With the projections of impact looking positive, we are currently working with the plant in Zedelgem on landing the application in the organization, allowing them to take future safety stock decisions in a data-driven manner.
The promising results at the plant in Zedelgem have sparked the interest of scaling the solution for data-driven safety management across CNH more widely. In parallel with the project in Zedelgem, we are working with the global CNH supply chain team to develop the next generation of this solution that is meant to be used by all of the CNH plants.
“Our ambition for the future is to further build trust in these complex algorithms within the company, not only in our Zedelgem plant but also in other plants.” - Pieter De Pourcq, Operational Excellence and Business Transformation Manager at CNH
For more information about what AI could mean for your logistics and supply chain, read this article.
