# Forecast Accuracy (MAPE and Bias)

> Source: https://netray.co/glossary/forecast-accuracy
> Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems.
> Section: Glossary

## Short answer

Forecast accuracy measures how close a demand forecast is to actual demand, commonly expressed through MAPE (mean absolute percentage error), WMAPE (weighted MAPE) and bias. Lower error and bias mean planning can rely on the forecast.

## Aliases

- MAPE
- WMAPE
- forecast error
- demand forecast accuracy
- forecast bias

## Category

manufacturing

## Full Definition

MAPE is the average of absolute error divided by actual demand across items and periods. It is easy to explain but breaks down when actual demand is zero or very small, and it overweights low-volume items. WMAPE fixes this by dividing total absolute error by total actual demand, so high-volume items count more. Accuracy is often reported as 100 percent minus the error.

Bias is the signed measure: total forecast minus total actual, divided by total actual. A forecast can have low absolute error and still be consistently high or low. Persistent positive bias builds excess inventory, while negative bias causes shortages. Planners track bias because it is correctable, unlike random noise.

Measure accuracy at the level and lag the decision is made. A monthly family forecast may look good while the weekly item-level forecast that drives MRP is poor. Measuring at the lag matching your supply lead time (for example the forecast frozen 8 weeks out) avoids flattering numbers from forecasts revised close to the actual.

ERPs hold the forecast and actual shipments, but accuracy reporting is often done in a planning module or outside the ERP. SyteLine, NetSuite demand planning and Infor LN have forecasting data sets, and many teams export to a BI tool to calculate WMAPE and bias by product family, customer and lag.

## Why It Matters

- Forecast error is the main driver of safety stock, so measuring it sizes buffers correctly.
- Bias tracking reveals systematic sales over-optimism or under-calling that plain error hides.
- A baseline lets you test whether new statistical or AI models actually beat the current method.
- Measuring at the right lag stops forecast revisions from disguising real planning performance.

## In Practice

For three items the actual demand is 100, 20 and 5, and forecasts are 90, 30 and 10. Absolute errors are 10, 10 and 5, totalling 25 against actual of 125. WMAPE = 25 / 125 = 20 percent. Simple MAPE is the mean of 10, 50 and 100 percent = 53 percent, dominated by the tiny item. Bias = (130 - 125) / 125 = +4 percent, a mild over-forecast.

## Faq

### What is the difference between MAPE and WMAPE?

**Answer:** MAPE averages the percentage error of each item, so low-volume items can dominate. WMAPE divides total absolute error by total actual demand, weighting results by volume.

### What is forecast bias?

**Answer:** Bias is the signed error: forecast minus actual over actual. A positive value means you consistently over-forecast and a negative value means you under-forecast.

### What is a good forecast accuracy?

**Answer:** It depends on demand volatility and aggregation level. Stable family-level forecasts may reach 85 to 95 percent, while item-level, intermittent demand can be far lower. Track improvement against your own baseline.

## Search terms

forecast accuracy, MAPE formula, WMAPE, forecast bias, how to measure forecast accuracy, demand forecast error, MAPE vs WMAPE, forecast accuracy manufacturing

---
Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json