Enterprise AI for Defense Supply Network Risk

Challenges

A defense and aerospace conglomerate produces 70 unique product variations for their most technical program that receives over 4,000 inbound deliveries from 374 suppliers annually for highly technical components with low yield and minimal sourcing options for their unique parts. The company attempted to proactively identify suppliers at risk of delay by connecting their ERP source system to an external data lake, which allowed their business intelligence tools to generate descriptive analytics. However, they were only able to identify and successfully mitigate 1% of their inbound delays with an average lead time of 10 days. The company was not able to identify the risks with enough lead time to change the outcome and reduce production delays.

The following challenges continued to impact the company’s ability to mitigate inbound risk:

• Exporting ERP source data to the data lake removed key fields that defined the table relationships between goods receipts and the planned deliveries making it challenging to accurately assess supplier performance
• Their tools lacked predictive capabilities, limiting analysis to historic insights and causing key patterns to stay hidden
• Legacy tools failed to construct comprehensive timeseries analytics, failing to capture useful delay signals required for predictive algorithms

As a result of these challenges, the global manufacturer faced an inventory backlog at port and distribution centers, transportation fines due to emergency shipments, and higher labor costs due to over-staffing.

Approach

C3 AI configured the C3 AI Supply Network Risk application for 70 unique finished goods for the defense company’s technical program with one manufacturing location. The team leveraged C3 AI Supply Network Risk’s logistic regression algorithms to identify the probability of an individual order arriving more than seven days beyond the scheduled date. The predictions identified high-risk inbound deliveries more than 90 days prior to the scheduled date — a 9x increase in lead time compared to the previous approach. The team configured a risk aggregation pipeline to prioritize high-risk items, purchase orders, and suppliers.

The joint team also configured the C3 AI Supply Network Risk user interface to visualize AI-based insights and inbound delay risk scores — empowering supply chain teams to proactively address at-risk suppliers and orders. Supply chain teams can now evaluate suppliers and conduct predictive analytics with additional timeseries analytics and supplier performance metrics.

About the Company

• $60+ billion annual revenue in 2022
• 1500 unique materials in the program
• 374 unique suppliers in the program
• 70 in-scope finished goods in the program

Project Objectives

• Integrate and unify disparate data sources from across the enterprise (e.g., ERP systems data, weather, financials, and supplier data)
• Apply machine learning pipelines to predict and identify the risk of inbound orders arriving more than seven days past their scheduled delivery date
• Configure the C3 AI Supply Network Risk user interface and expose AI insights and analytics to end users

Project Highlights

• Twelve years of historical data integrated, comprised of 40 million rows from 13 enterprise data sources and three external data sources
• Built an extensible data model with 20 C3 AI logical objects
• Developed 50+ timeseries analytics for machine learning models
• Configured and tested 100+ machine learning model permutations to predict the risk of late deliveries and suppliers

By using the C3 AI Supply Network Risk application, the defense company is able to:

• Generate up to $9 million in accelerated operating income through reductions in inbound delays, production delays, and mitigating the impact of customer delays on profit margins
• Predict late inbound deliveries 90 days prior to the scheduled delivery date
• Increase the probability of successful inbound risk mitigation by 8x compared to current practices
• Reduce production delays and impact to outbound order fulfillment