Forecast accuracy formula: how to calculate MAPE, WAPE and bias

There is no single forecast accuracy formula. Most teams use one of three measures, and each can tell a different story from the same numbers. Here is how to calculate them and which to report.

The formulas

MAPE = average of (|forecast − actual| ÷ actual)

WAPE = sum of |forecast − actual| ÷ sum of actual

Forecast accuracy = 1 − WAPE

Bias = (sum of forecast − sum of actual) ÷ sum of actual

A worked example

Item	Forecast	Actual	Error	Absolute error	% error
A	100	120	−20	20	16.7%
B	500	450	+50	50	11.1%
C	40	20	+20	20	100.0%
D	250	260	−10	10	3.8%
E	110	150	−40	40	26.7%
Total	1,000	1,000	0	140

Measure	Calculation	Result
MAPE	Average of the five % errors	31.7%
WAPE	140 ÷ 1,000	14.0%
Forecast accuracy	1 − 14.0%	86.0%
Bias	(1,000 − 1,000) ÷ 1,000	0.0%

MAPE is more than double WAPE because item C, at just 20 units, counts as much as item B at 450. Bias is zero because the highs and lows cancel out, even though most individual items were wrong. That is why you need both an error measure and a bias measure.

To skip the math, paste your numbers into the free forecast accuracy calculator.

In Excel

With forecasts in B2:B6 and actuals in C2:C6:

WAPE: =SUMPRODUCT(ABS(B2:B6-C2:C6))/SUM(C2:C6)

MAPE: =SUMPRODUCT(ABS(B2:B6-C2:C6)/C2:C6)/COUNT(C2:C6)

Bias: =(SUM(B2:B6)-SUM(C2:C6))/SUM(C2:C6)

MAPE returns an error if any actual is zero, which is common for slow items.

Which measure to report

• WAPE for the overall number. It weights items by volume and copes with zero sales.
• Bias alongside it. A forecast that runs high every month builds excess stock, and one that runs low causes shortages.
• MAPE only for steady items with no zero-sales periods.

How to measure it properly

1. Use the right lag. If a supplier takes three months, judge the forecast you made three months ago.
2. Measure at the level you buy, such as item by location. Totals hide errors.
3. Compare it with a simple method, such as last year’s sales. That shows whether your process adds anything.
4. Split by item class. Erratic items will always score worse.

What to do with the result

Error tells you how much safety stock an item needs. Bias tells you to look for a cause, such as optimistic sales numbers. For the wider process, read our guide to demand planning software. The Statadventure Optimization Lab shows WAPE for each item’s forecast on the Planner and Growth plans.

Frequently asked questions

What is a good forecast accuracy?

It depends on the item. Steady, high-volume items can be forecast closely, and erratic, low-volume ones cannot. Compare your result with a simple method and track it by item class.

Why use WAPE instead of MAPE?

WAPE weights items by volume, so tiny items do not distort the result, and it still works when some items have zero sales.

What is forecast bias?

The tendency to forecast consistently too high or too low. A positive bias means you are over-forecasting.