From raw external signal to replenishment order in hours, not weeks.
Six nodes, one flow: signal ingestion to NLP parsing to demand driver mapping to forecast adjustment API to your planning system to replenishment trigger. Each step is auditable. Each output is a concrete number tied to a specific SKU and DC — not a directional forecast adjustment you still have to interpret.
Architecture
Six-node signal-to-order flow
Under the hood — three layers, no black box
Every demand planner who's been burned by a model they couldn't audit deserves to know exactly what's happening inside. Each layer does one thing and hands off to the next with a structured, inspectable output.
Supplytrx pulls from a mix of real-time API feeds and structured data files. Weather data refreshes hourly from NOAA and commercial forecast APIs. Freight index data updates daily from major shipping intelligence providers. Social listening feeds are monitored continuously with a 15-minute sampling window. Commodity price feeds refresh throughout the trading day.
All data arrives as raw text or semi-structured JSON. Supplytrx normalizes each feed into a common signal schema before it reaches the NLP layer — one consistent format regardless of source.
The NLP layer classifies each ingested text fragment by signal type (weather event, port disruption, social trend, commodity movement, news event), extracts key attributes (magnitude, geographic scope, affected product categories, timeline), and scores relevance against your configured SKU catalog.
Category-to-SKU mapping is maintained by your onboarding team during setup. When a signal fires — say, a hurricane forecast for the Gulf Coast — the system queries your SKU catalog for pantry-load categories in affected DC service areas and outputs a demand adjustment recommendation with confidence score.
Supplytrx outputs adjusted demand numbers via REST API or flat-file export. The API follows a simple demand-adjustment schema: SKU, DC, forecast period, adjustment factor, confidence score, signal source. Most planning systems accept an external demand adjustment input through their standard API layer.
For systems without a live API (or for teams that prefer to review before committing), we support flat-file output: a CSV or Excel file formatted to your planning system's import template. The file lands in a configured SFTP location or S3 bucket on your designated refresh schedule. Both paths are supported simultaneously — teams can run live API for high-confidence signals and flat-file review for lower-confidence or high-impact signals.
From contract to first improved forecast: under 2 weeks.
Onboarding is a two-week sprint, not a six-month implementation. We don't run consulting engagements or charge for professional services discovery. You connect your data, we configure the signal mapping, and your planning team reviews the first adjusted forecast output before week two is done.
API credentials, data feed configuration, SFTP or S3 setup. Your team handles one environment file. We handle the rest.
We ingest your SKU catalog and DC list. Category-to-signal correlations are configured from your product taxonomy and historical demand patterns.
Signal thresholds and sensitivity settings are tuned against your historical data. Replenishment trigger rules configured per your operating policies.
First signal-adjusted forecast output delivered. Your team reviews alongside baseline. Adjustments made. Second cycle runs clean.
30 minutes, your actual SKU categories — not a generic demo.
Bring your top 5 categories and we'll walk through which signals have historically moved demand in those categories, what the lead time advantage looks like, and how the forecast integration would work with your current planning tool.
Request Access