Canada
Puerto Rico
USA
Belgium
France
UK
India
Client Background
The client is a leading global steel producer, recycling scrap metal into high-quality products. Operating across North America, they run an extensive network of recycling locations, steel mills, and downstream facilities. Their portfolio includes merchant bar, structural steel, grinding balls, rail spikes, and rebar, serving three customer segments: manufacturing, construction, and distribution.
Despite its focus on operational excellence, the company faced order-to-shipment inefficiencies that tied up capital and impacted on customer satisfaction.
Client Need
Order inefficiencies & cancellations
- Orders were cancelled without advance notice—sometimes after manufacturing
- Shipments were refused by customers, causing blocked inventory and disrupted manufacturing planning
- Orders were not closed on time, complicating inventory planning and manufacturing schedules
Inefficient demand forecasting
- Leadership spent a large amount of time in roundtable meetings to predict upcoming demand
- Manual forecasting accuracy did not meet business needs
Suboptimal shipment utilization
- Truck and rail shipments were not optimized; capacity was under-utilized
- Leadership wanted an algorithm to maximize tonnage per shipment while satisfying customer and business constraints
Order fulfillment delays & communications gaps
- Orders nearing shipping deadlines were at risk of delay without proactive customer engagement
- No automated notifications existed to nudge customers or operations to act
Need for speed, reliability, & constraint-aware planning
- Shipment creation and scheduling were manual
- Customer-specific constraints were not consistently enforced in planning
Solution
Predictive Order Inefficiency Model
- Analyzed 3 million+ historical records (2016–2019) across order entry, manufacturing, and shipment
- Parsed and cleaned data from three perspectives: Order Entry, Order Manufacturing, and Order Shipment
- Developed a classification model to predict inefficient orders as soon as an order is entered
- Engineered important features, derived new ones to augment performance, and interpreted drivers of inefficiency within business context
- Integrated model outputs into order management and manufacturing planning processes to proactively prevent blocked inventory and unnecessary production
Automated Shipment Forecasting
- Built a shipment time-series forecasting model using historical shipment records to capture true business seasonality and patterns
- Produced shipment forecasts across multiple hierarchies: segment, “sold-to,” customer division, and block resources
- Deployed as an automated job using AWS SageMaker pipelines to run monthly and generate 12-month rolling forecasts at multiple hierarchies
- Delivered a data-driven forecasting tool that replaced manual forecasting and provided reliable month-ahead and year-ahead shipment projections
Shipment Optimization Algorithms
- Developed an optimization algorithm using the pulp package to build truck/rail-bound shipments that respect customer and business constraints
- Maximized tons loaded on each shipment subject to constraints (truck/rail capacity, customer limits, business rules)
- Integrated optimized shipment creation with SEP for truck-based order fulfillment and provided continuous support to extend functionality
Smart Shipment Creation
- Created a linear programming solution to group items into trucks and rail shipments while honoring capacity and customer-specific constraints
- Integrated the solution with existing scheduling applications, and extensible across plants and customer groups
Automated Delay Notification System
- Implemented a Spring Batch job with multithreading to retrieve order data from AWS S3, persist via Athena → PostgreSQL, and trigger automated customer email notifications for orders approaching critical shipping dates
- Built a flexible template management application for creating personalized notification templates
- Exposed Nudge REST endpoints to support integration and programmatic control of notifications
- Integrated these notification features into existing order management systems with ongoing bug fixes and enhancements
Realized Benefits
Order Efficiency & Inventory / Manufacturing Planning
- Model predicted inefficient orders with 81% ROC-AUC, enabling early intervention and preventing unnecessary production
- Approx. $20 million saved by preventing inefficient orders and cleaning up blocked inventory
- Model outputs were fed into manufacturing planning—helping plan production quantities to meet real demand more efficiently
Forecasting Accuracy & Executive Time Savings
- Forecasting model delivered forecasts with MAPE ≤ 10%, significantly better than manual methods
- Automated forecasting process saved ~30 leadership hours per month previously spent in manual roundtable forecasting
Shipment Optimization & Cost Savings
- Optimization identified 10 customers for pilot testing and saved 62 shipments week-on-week across them
- Pilot savings of $51,000 across 10 customers translated into ~$2.4M in annual shipment cost savings, driven by optimized truck loading and routing that also cut logistics overhead
Operational Throughput & Plant Efficiency
- Number of shipments increased by 15% with better planning and order management
- Shipments decreased by 8% in volume where optimization consolidated loads, and items shipped increased by 11%, reflecting better utilization of capacity and fewer partial/failed shipments
- Plant efficiency improved because fewer shipments were required to move the same or greater tonnage
Customer Communication & Fulfillment
- Automated nudge notifications reduced delayed shipments by prompting customers to act before deadlines
- Improved inventory turnover and higher on-time fulfillment rates
- Personalized templates increased transparency and trust with customers
Trending Success Stories
Unlocking Healthcare Sales Insights with Conversational AI
Accelerating QA Transformation Through Intelligent Automation
Enabling Transformation of Order-to-Ship Life Cycle Through Advanced Forecasting
Transforming Medical Affairs with a Generative AI Research Assistant
Ready to Innovate with Us?
Let’s Talk!
Connect with us on social media
Write to us at
[email protected]