Inventory forecasting
Understand how historical data is transformed into actionable business insights.
Dema’s inventory forecasting engine transforms your historical sales data and product characteristics into precise, forward-looking demand forecasts for the remainder of your selling season. These intelligent insights empower you to optimize inventory levels, implement strategic pricing, and allocate marketing resources with confidence—preventing costly stockouts and reducing excess inventory before they impact your bottom line.
How to access
The forecasts are available in the Reports section of the platform as selectable metrics.
How our forecasting engine works
Core algorithm
The system is built on gradient boosted decision trees, an advanced algorithm architected for large-scale retail datasets, delivering speed and accuracy for complex inventory forecasts.
Continous model learning
Incorporate new sales patterns, seasonal shifts, and market changes to maintain forecasting precision throughout your selling cycle once a day and update your forecasts accordingly.
Core algorithm
The engine employs an ensemble of decision trees that systematically analyze your data through intelligent questioning patterns. For example: “Did this SKU experience strong performance in the European market during last spring’s campaign?” or “How did similar products in this category perform during comparable promotional periods?”
By aggregating insights from thousands of these decision trees, the model generates comprehensive end-of-season sales forecasts rather than traditional day-by-day forecasts. This approach provides more stable and actionable forecasts for strategic inventory planning.
Consumer preferences and market conditions evolve rapidly. Configure automatic model retraining to trigger after significant changes in your product assortment, pricing strategy, or distribution channels to maintain optimal forecasting accuracy.
Hierarchical architecture
The system employs a bottom-up approach to ensure that both granular inventory decisions and high-level strategic planning benefit from consistent, accurate data.
The resulting output is a collection of individual end-of-season sales forecasts, each on the smallest granular level, enabling you to make the most informed decisions.
Distribution Center → Product Category → Individual SKU/Variant
This approach enables accuracy from the ground up, ensuring that detailed insights aggregate correctly to provide reliable top-line forecasts.
Features
The system uses a combination of historical performance indicators, product attribute analysis, and market context variables to create a comprehensive feature set.
- Trend velocity and seasonal patterns
- Geographic variations
- Product consistency and demand metrics
- Price point positioning across product hierarchy
- Brand equity and collection performance history
- Marketing spend activity across available product hierarchies
Feature importance can differ greatly across categories, brands, warehouses, and regions. Our machine learning model dynamically adapts to these differences by identifying and prioritizing the most predictive features based on your specific business context and customer behavior.
Forecasted metrics
The system generates a comprehensive set of forecasted metrics that can be used to make informed inventory decisions which can be accessed in the Reports section of the platform.
| Metric | Description |
|---|---|
| Forecast incoming returns | Estimates the number of products expected to be returned on already sold orders for the rest of the season, helping anticipate inventory adjustments and returns impact. Based on historical trends and inputted return rates. |
| Forecast gross products sold | Forecasts the total quantity of products expected to sell during the rest of the season, aiding in demand planning and promotional strategies. |
| Forecast return percent | Estimates the percentage of sold products that are likely to be returned during the rest of the season, based on historical trends and inputted return rates. |
| Forecast net products sold | Estimates the number of products expected to sell during the rest of the season, adjusted for historical return rates. This metric helps in planning inventory and marketing strategies by forecasting future sales performance based on past trends. |
| Forecast plus actual net products sold | Combines forecasted net product sales with actual sales to date, providing a comprehensive view of seasonal sales performance. |
| Forecast sell-through | Calculates the ratio of forecasted net products sold to total available stock (current, incoming, and forecasted returns), providing insights into inventory efficiency. |