Multi-signal models that read supplier emails, shipping data, and commodity prices outperform spreadsheets by 22%
The 2024 chip shortage exposed their forecasting model as a spreadsheet with delusions of grandeur. They'd overordered $60M in networking equipment that sat in warehouses for 8 months while running out of the storage drives their customers actually needed. The CFO called it 'the most expensive guess in company history.'
Managing 2M+ SKUs across 4,000 vendors and 40 warehouses globally. Their demand forecasting relied on 18-month historical averages — which worked fine in stable markets but collapsed during disruptions. A 5% forecast error at their scale meant $200M in misallocated capital. They needed a system that could see disruptions coming before they showed up in the order book.
We built a multi-signal AI forecasting system that ingests historical sales, real-time shipping data (port congestion, carrier delays), supplier email sentiment via NLP, commodity prices, and customer order velocity. Time-series transformer models update forecasts continuously. A custom dashboard surfaces the 50 highest-risk SKUs daily. Our data engineering team manages the pipeline 24/7 — because supply chains don't keep business hours.
We used to find out about shortages when our customers called to complain. Now we find out when our supplier's tone changes in an email three weeks before the official notice.— SVP of Supply Chain Operations
Let's talk about what AI + a supplemental engineering team can do for your business.
Talk to a Dev Lead →