Under the Hood
Transparent algorithms you can trust. Not a black box — here's exactly how invent.sale AI works with your data.
Stochastic Demand Forecasting
Four model candidates — polynomial, sinusoidal, exponential, and logarithmic — compete for each product. The best fit wins, forming a stochastic flow. We generate P10/P50/P90 confidence intervals so you can plan for optimistic, expected, and pessimistic scenarios.
- 52-week seasonality classification (periodic, aperiodic, flat)
- Automatic model selection per SKU
- Confidence intervals: P10 / P50 / P90
- Weekly and monthly seasonality decomposition
Intelligent Dead Stock Detection
A multi-criteria filter identifies truly frozen inventory: no recent sales + not off-season + above stock percentile + has physical inventory. Each criterion must pass — eliminating false positives from seasonal dips or new arrivals.
Smart Markdown Engine
Calculates optimal discounts to maximize recovery from slow-moving inventory. Distribution: ~50% dead stock, ~25% overstock, ~15% seasonal items. Each candidate gets a priority score based on urgency, margin, and shelf life.
- Optimal discount calculation per item
- Urgency scoring: days on shelf × frozen capital
- Category-level markdown strategy
- Projected revenue from clearance
AI-Powered Purchase Recommendations
Six recommendation types generated automatically: purchase, capital release, markdown, demand growth alert, demand drop alert, and seasonal. Each is ranked by margin impact to prioritize what matters most.
- Pipeline: Forecast → Margin Priority → Stock Coverage → Order
- Purchase recommendations for high-margin, low-stock items
- Capital release alerts for frozen inventory
- Seasonal and demand shift notifications
How data flows into the platform
Connect any accounting system — 1C, SAP, Oracle, MySklad, Excel or API. Data is automatically cleaned, normalized and AI-enriched. You get ready-to-act recommendations — no manual work required.
- Supports any ERP: 1C, SAP, Oracle, MySklad and more
- Import via API, files (Excel/CSV) or direct connection
- Automatic data cleaning and normalization
- AI enrichment: forecasts, classification, recommendations