Quick Summary: The app generates forecasts for every stocked item using sales history, demand streams, and configuration settings. Understand these core mechanics to determine when to strategically adjust system recommendations.
How Forecasts Are Generated
At the beginning of every month, the app automatically regenerates forecasts for every item at every location.
This process has three key stages:
Analyze the Sales History
The app reviews monthly sales history for each item-location combination, identifying trends, patterns, and gaps.
It cleans irregularities such as once-off bulk sales or missing months, so the forecast is based on meaningful patterns.
Test and Compare Forecasting Models
Each item’s sales pattern is tested against several statistical models (flat, trend, seasonal, sporadic).
The app then back-tests each model, essentially asking, “Which one would have predicted past sales most accurately?”
The model with the lowest forecast error becomes the best fit.
Assign the Demand Type and Generate Forecasts
The item is classified into a Demand Type that matches the winning model. This determines how its forecast is calculated in the future.
Forecasts are recalculated automatically every month and whenever new data is imported or manually refreshed.
Categorizing Sales: The Role of Demand Types
Not all products follow the same demand pattern. The app assigns demand types so each product can be forecasted in the most appropriate way:
A household staple like dishwashing soap sells steadily month after month, so it requires a flat, stable forecast.
Sunglasses sell in large volumes only during summer, so they need a seasonal forecast that accounts for recurring peaks.
Spare parts may not sell for months and then move in large bursts, so they require a sporadic forecast that allows for irregular demand.
If all of these products were forecasted in the same way, the results would be inaccurate. Classifying each item into a demand type ensures that every forecast reflects reality as closely as possible.
💡 Tip: You can view the Demand Type on the Item Inquiry page, just above the mini demand chart.
Understanding Each Demand Type
Below is a practical guide to each Demand Type, what it means, and how to manage items that fall under it.
No History
An item with no recorded sales history cannot generate a data-driven forecast, so the system assigns it a forecast of zero. This is common for brand-new or pre-launch products that have not yet recorded any sales activity. As soon as the first month of sales appears, the item automatically shifts from No History to Young, and regular forecasting begins.
Monitor the Top New Items section on your Dashboard. These are often new SKUs that could benefit from a manual starting forecast until actual sales accumulate.
💡 Tip: Use Supersessions when a new item replaces an older one. This transfers the historical sales from the old item to the new one, giving the system something solid to work with from day one.
Young
A Young item has only a few months of history. With so little data, Netstock uses a weighted moving average, giving greater importance to the most recent months.
Because the data set is small, forecasts for young items can fluctuate more than usual. You may want to review them frequently and add manual forecasts if you have better market insight, for example, if a product launch is accelerating faster than early data suggests.
Slow Mover
A Slow Mover sells in small, fairly consistent quantities with no seasonal pattern.
Netstock applies a simple moving average of past sales to produce a flat, steady forecast.
These items are sensitive to outliers. A single large order can distort the average. For this reason, some planners prefer to support slow movers with minimum stock levels rather than relying entirely on statistical forecasts.
Sporadic
Sporadic items don’t sell every month. They may have long periods of zero sales followed by a few high-volume transactions. Netstock smooths these peaks using a smoothed moving average, which prevents one large order from inflating future forecasts.
Because sporadic demand is unpredictable, these items deserve closer planner attention. Use manual forecasts, confirmed customer orders, or small safety-stock buffers to protect against surprise demand during long supplier lead times.
Seasonal
A Seasonal item shows a repeating monthly pattern year after year, for instance, umbrellas that peak each winter or sunscreen that rises each summer. The app automatically tests for seasonality using thresholds set in your configuration. If the pattern is strong enough, the item is classified as Seasonal and forecasted using a seasonal model, which may also include a growth or decline trend.
Once seasonal behavior is confirmed, you’ll see the same peaks projected forward into future months.
Linear Trend
A Linear Trend item displays a steady upward or downward pattern across many months. The forecast simply extends that slope into the future, increasing or decreasing at the same rate.
You’ll often see this with growing product lines, declining legacy items, or technology products approaching end-of-life.
No Trend / No Seasonality
Some products sell consistently with no visible trend or seasonality.
These are classified as No Trend / No Seasonality and forecasted using a weighted moving average that continues the recent average demand into future months.
Special Demand Types
Sum of Locations
Sum of Locations may apply to items in a “region” location created in the app. It indicates that the region’s forecasts have been aggregated (summed) from its source locations rather than generated afresh.
Seasonal Group
Seasonal Group is a special category that may apply if the Advanced Forecasting feature is enabled and a Seasonal Group has been selected. It indicates that the seasonal profile from a group of related items has been applied to the individual item.
For example, a new beverage flavor with limited history may follow the established summer peak of the entire drinks range. This ensures that even low-history SKUs benefit from a reliable seasonal curve.
⚠️ Watchouts
Limited or Missing Data: Items with no history or very young history are difficult to forecast accurately. Use business judgment when working with them.
Changing Demand Types: Demand types may change automatically over time as the system collects more history. Do not assume the current demand type is permanent.
Misreading Sporadic Demand: Misinterpreting sporadic demand as low demand can lead to stockouts when the next burst occurs.
💡 Tips
Review Regularly: Review demand types regularly, especially for new or seasonal products.
Use as a Diagnostic Tool: Treat demand types as an explanation of why the app produced a particular forecast. They are a valuable tool for understanding the system’s logic.
Forget about these 👇 😞 😐 😃 Have your say here!