At Tightly, we don't just guess your future sales, we calculate them using a variety of sophisticated forecasting models. Because every product has a unique sales pattern (some are seasonal, some are steady, and some are trending), Tightly automatically assigns the best-fitting model to each individual product to ensure maximum accuracy.
You can access the Forecast Models page by clicking the tab under the Demand menu or by clicking "Explore Forecast Models" on your Demand Planning page.
1. Model Distribution Overview
At the top of the page, you will find a bar graph representing all active forecasting models currently in use across your catalog.
What it shows: Each bar represents a specific mathematical model (e.g., ARIMA, Prophet, or Exponential Smoothing).
Interactive Data: Hover over any bar to see exactly how many products are currently being forecasted using that specific model (e.g., "Active model for 3 products").
2. Model Insights
Directly below the graph, the Model Insights cards highlight specific examples of how Tightly has optimized your forecasting.
These cards showcase why a particular model was selected for a specific product, such as "Prophet selected for complex demand patterns" or "ARIMA selected for trending demand".
3. Recent Model Switches
Tightly’s engine is constantly testing. If our system discovers that a different mathematical model provides a more accurate result than the current one, it will automatically switch them.
Model Switch Column: See the transition from the "Initial model" to the new "Current model".
Switch Reason: Click the icon in the Switch Reason column to see the data-driven justification for the change, such as improved performance during testing phases.
4. How Each Forecasting Model Works
For users who want a deep dive into the "back-end" logic, this section provides user-friendly explanations of our five core models:
Model | Best Used For... | Logic Summary |
ARIMA | Regular patterns and steady trends. | Combines past values and recent changes to predict the next step. |
Prophet | Complex demand with heavy seasonality or holidays. | Focuses on long-term trends and calendar effects. |
Exponential Smoothing (ETS) | Steady, continuous demand that changes gradually. | Gives more weight to recent data while still considering older points. |
Non-stockout Moving Average | Stable demand with no clear seasonality. | Averages recent periods where stockouts did not occur to find true demand. |
TSB (Croston with Decay) | Intermittent or "sparse" demand (items that sell occasionally). | Separately tracks how often sales happen and the size of those sales. |