Stop guessing. Along with traditional statistical methods, DemandTraX™ uses the newest predictive models bolstered by Generative AI to automate demand forecasting, secure a more precise, self-healing inventory, and unlock trapped operating cash.
Per-SKU
Method Tournament
NB-Quantile
Safety Stock Sizing
<30 Days
Guided Onboarding
$1M–$50M
Inventory Range Built For
Built on the same forecasting science used by Fortune 500 supply chain teams — configured for mid-market businesses.
* Capability claims reflect the system's methodology and internal testing. Results depend on your own data.
From automated Genkit-powered executive briefings to dynamic algorithmic forecasting tournaments, our AI handles the complex math so you can handle the strategy.
Stops you from applying one-size-fits-all math to a wildly mixed catalog. DemandTraX™ tests multiple forecasting methods against your own sales history and selects the most accurate one — per part, automatically.
Set a capital ceiling and DemandTraX™ automatically reallocates your inventory budget to protect the parts that matter most — without a spreadsheet or manual review.
Instantly identify which parts haven't sold in 90, 180, or 365+ days — and get a recommended action: vendor return, markdown, or write-off. Stop letting obsolete stock silently drain your balance sheet.
If you run multiple warehouses or branches, each location is probably holding more safety stock than it needs. DemandTraX™ pools risk centrally so your branches stay stocked without the duplication.
Import directly from your ERP's CSV or Excel export — purpose-built for Microsoft Dynamics 365 Business Central, and compatible with any standard inventory and sales export. No lengthy IT project required; direct API connectors are on the roadmap.
Click through the dashboard views below to tour our interface and learn how we manage inventory at enterprise scale.
1,248
Across 3 warehouses
$4.82M
Active deployment
Your team starts every morning knowing exactly what's at risk. Low-stock alerts, capital tied up in slow movers, supplier status, and reorder actions — all on one screen, no report-building required.
We're onboarding a select group of founding design partners to configure DemandTraX™ around their specific industry, catalog, and ERP environment. Partners get extended early access, preferential pricing, and direct input on the product roadmap.
Extended Early Access
Free or heavily discounted for your first year
Roadmap Influence
Direct input on features built for your industry
Dedicated Onboarding
White-glove setup configured around your data
DemandTraX™ isn't a thin wrapper around a chat model. It runs a multi-stage pipeline that classifies every SKU by its demand pattern and routes it to the forecasting method that actually fits — proven statistical models for most parts, an LLM-assisted review for the hardest.
Raw historical demand is ingested and scrubbed. Anomalies, stockouts, and drop-ships are isolated so they don't corrupt the baseline. The system mathematically isolates structural breaks in demand.
For SKUs with stable, predictable history, an optional BigQuery path runs Google's TimesFM foundation model — zero-shot, no training required — scaling to thousands of parts. When that path isn't enabled, the same parts fall back to our deterministic math engine, so a forecast is never lost.
The system calculates the CV² (Coefficient of Variation) and MASE (Mean Absolute Scaled Error) for every SKU. Stable, low-variability parts stay on the fast deterministic path; intermittent, lumpy, and volatile parts are routed to the per-SKU method tournament and, where it earns its keep, an LLM-assisted review.
Each part is forecast by a field of proven intermittent-demand and time-series models — Croston, bias-corrected SBA and TSB, ETS, Theta, and trend models — each backtested with rolling origins over the full deployed horizon, not just one step ahead. The winner is picked per SKU on real out-of-sample error. Safety stock is then sized from a Negative-Binomial demand distribution, which captures the fat tails of intermittent demand that a normal-curve min/max under-sizes.
For the hardest, least predictable items (e.g., BOM-dependent assemblies, future manual spikes), the math engines pass their priors to a high-context LLM within the secure Google Cloud Vertex AI environment. The AI audits the math and generates a final, human-readable rationale for the recommended safety stock.
We employ state-of-the-art serverless architecture and enterprise database isolation to provide blistering analytics at extreme scale.
Render-optimized application code ensuring extremely fast visual loads, responsive layout layouts, and modular dynamic widgets.
Firestore rules enforce tenant isolation and an append-only audit trail — immutable to every customer role — that records overrides, imports, and key account actions.
For stable, high-volume parts, forecasts can run in BigQuery using Google's TimesFM foundation model (zero-shot) — handling large catalogs beyond browser limits.
Secure integration of Gemini generative parameters inside analytical solvers to output transparent forecast and restock rationales.
Join a small group of founding design partners and get the platform configured around your data, your ERP, and your team — before we open broadly. Questions? Email sales@demandtrax.app.