Demand forecasting estimates future customer demand so planners can make better decisions about purchasing, production, inventory, capacity and logistics. A useful forecast is not simply a number: it should be measurable, repeatable and monitored for both accuracy and bias.
Why Demand Forecasting Matters
- Sets the demand signal used in supply and inventory planning.
- Supports safety-stock and replenishment decisions.
- Helps capacity, labour and supplier planning.
- Provides the starting point for S&OP demand review.
- Highlights uncertainty so planners can prepare scenarios rather than rely on one number.
Four Practical Forecasting Methods
1. Naive Forecast
The next period is forecast to equal the latest actual demand. It is simple, transparent and useful as a baseline. More complex methods should normally outperform it before they are considered worthwhile.
2. Moving Average
A moving average uses the average of the most recent periods. A three-period moving average is:
Forecast = (Demand t-1 + Demand t-2 + Demand t-3) / 3
This smooths short-term variation but can lag when demand is moving consistently upward or downward.
3. Weighted Moving Average
A weighted moving average gives more importance to recent demand. For example, with weights of 50%, 30% and 20%:
Forecast = 0.50 × most recent demand + 0.30 × previous demand + 0.20 × demand three periods ago
4. Exponential Smoothing
Exponential smoothing updates the previous forecast using the latest forecast error:
New Forecast = Previous Forecast + α × (Actual − Previous Forecast)
The smoothing factor α controls responsiveness. A higher alpha reacts more quickly to recent changes; a lower alpha produces a smoother forecast.
Forecast Accuracy Metrics
| Metric | What it tells you |
|---|---|
| MAE / MAD | Average absolute forecast error in the same units as demand. |
| MAPE | Average absolute percentage error; easy to communicate but problematic when actual demand is very low or zero. |
| WAPE | Total absolute error divided by total actual demand; useful for aggregate performance. |
| Bias | Shows whether forecasts systematically over- or under-estimate demand. |
| Tracking Signal | Compares cumulative error with average absolute error to highlight persistent bias. |
Accuracy Is Not Enough: Check Bias
A forecast can have an acceptable average error and still be consistently too high or too low. Persistent over-forecasting can create excess stock and working-capital pressure, while persistent under-forecasting can create stockouts, expedites and lost service.
How to Select a Forecasting Method
- Start with a simple baseline such as naive forecasting.
- Test moving average, weighted moving average and exponential smoothing on the same historical periods.
- Compare error metrics over a common evaluation window.
- Check bias, not just average error.
- Segment items where needed; one method rarely fits every SKU.
- Apply justified business overrides for promotions, launches, known disruptions or structural changes.
- Measure the override afterwards to determine whether it improved the statistical forecast.
Common Forecasting Mistakes
- Choosing the most sophisticated model rather than the best-performing model.
- Using MAPE blindly when demand contains zeros or very low volumes.
- Mixing forecast accuracy periods across different models.
- Changing forecasts manually without recording the reason.
- Ignoring lost sales, stockouts or promotions that distorted historical demand.
- Measuring accuracy at a level different from the level at which decisions are made.
- Treating the forecast as a commitment rather than an estimate with uncertainty.
How Forecasting Fits into S&OP
The demand forecast is the analytical starting point for the demand review. Commercial intelligence, promotions, customer changes and market events are then considered before the consensus demand plan is balanced against supply, inventory, capacity and financial objectives.
Download the Practical Workbook
Use the SCMANA Demand Forecasting Calculator to compare Naive, 3-Period Moving Average, Weighted Moving Average and Exponential Smoothing forecasts using a common evaluation window. The workbook calculates MAE, MAPE, WAPE, bias and tracking signal and identifies the lowest-WAPE method automatically.
Download the Demand Forecasting Calculator (Excel)
