We built Supplytrx because demand planning was stuck reading yesterday's data to make tomorrow's decisions.
Founded in Atlanta in 2022. One product, one problem: shifting the replenishment signal from reactive to predictive.
Diana spent 7 years watching the same problem repeat itself every quarter.
She spent seven years running demand planning and supply chain analytics at a mid-size food manufacturer in Atlanta. Quarter after quarter, the team reacted to stockouts that — reviewed afterward — had been entirely predictable from NOAA weather data, freight market indices, and social trend signals available weeks before demand moved.
The problem wasn't data access. Every signal existed. Weather forecasts were public. Port congestion indices were published daily. Social listening tools reported ingredient trends in near real time. The problem was translation: nothing connected those signals to the specific SKU-DC forecast adjustments a demand planner could act on within the replenishment window. The insights arrived as a slide deck three weeks after the stockout.
Diana left in 2022 to build that translation layer. She recruited Marcus Webb, who had spent five years building real-time freight data pipelines at a logistics technology company and brought the ingestion infrastructure Supplytrx needed. Priya Nair, a PhD in operations research who had spent years on CPG forecasting models, built the signal correlation engine. James Kowalski — who had run demand planning at a regional grocery retailer for a decade — joined to run customer success and make sure every output passed the Monday morning planner test: is this a number I can act on, or just a signal I still have to interpret myself?
"The data to predict demand has existed for years. It's been sitting in weather APIs and shipping reports while demand planners were stuck reading POS output from 3 weeks ago." Diana Guerrero, CEO & Co-Founder
Signal before reaction
Every demand event leaves a footprint in external data before it registers in POS. Our job is to read those footprints and translate them into replenishment-ready forecast adjustments.
Specific, not generic
A generic "weather will affect demand" output is useless to a planner with a Monday order deadline. A "SKU 42381 at DC-07 needs +28% replenishment before Friday based on the Gulf Coast storm track" is actionable. Every output we produce has a specific SKU, a specific DC, and a specific recommended adjustment — not a directional nudge.
Planner-grade, not consultant-grade
Output goes directly into your planning system via API or flat-file — not into a consulting report that lands after the replenishment decision is already made. We measure success by fill rate improvement, safety stock reduction, and fewer emergency orders. Not by NPS surveys.