Perimattic
Laptop screen showing analytics charts and performance graphs

Free Tool

Demand Forecast Accuracy Calculator

Enter your forecast and actual demand figures to compute MAPE, bias, and weighted MAPE. See the formulas and benchmark your accuracy.

Book a Demo

How do you calculate demand forecast accuracy?

Demand forecast accuracy is measured by MAPE (Mean Absolute Percentage Error): for each period, divide the absolute difference between forecast and actual demand by the actual demand, then average across all periods and multiply by 100. A MAPE of 20% means your forecast is off by 20% on average. Bias — the signed average error — tells you whether the forecast consistently over-predicts or under-predicts. Demand OS computes both metrics automatically at the SKU level and retrains models when accuracy drifts.

Calculator

Enter forecast vs actual demand

ForecastActual

Formulas used

MAPE= (1/n) × Σ |Forecast − Actual| / Actual × 100
Bias= (1/n) × Σ (Forecast − Actual) / Actual × 100
Weighted MAPE= Σ |Forecast − Actual| / Σ Actual × 100

Results

Enter your forecast and actual values, then click “Calculate Results” to see your accuracy metrics.

How to use it

Forecast accuracy, worked through step by step

The calculator takes one forecast and one actual value per period and returns four numbers: MAPE, forecast accuracy (100% minus MAPE), bias and weighted MAPE. Periods with an actual of zero are skipped, because MAPE divides by the actual value.

A worked example with three periods

Say you forecast 100, 150 and 90 units, and actual demand came in at 120, 140 and 100. The absolute percentage errors are 20 ÷ 120 = 16.7%, 10 ÷ 140 = 7.1% and 10 ÷ 100 = 10.0%. Their average is a MAPE of 11.3%, so forecast accuracy is 88.7%.

Bias keeps the sign: (−16.7% + 7.1% − 10.0%) ÷ 3 = −6.5%, which means the forecast ran low. Weighted MAPE adds the absolute errors and divides by total demand: 40 ÷ 360 = 11.1%.

How to read your result

The calculator labels MAPE of 15% or less as excellent, up to 25% as good, up to 35% as fair, and anything higher as needing improvement. Bias within ±3% is treated as minimal. Positive bias means you over-forecast and build excess stock; negative bias, as in the example, means you under-forecast and risk stockouts. Look at both together: a low MAPE with a steady negative bias still needs fixing.

Common mistakes

Measuring too few periods, so one unusual month dominates. Comparing MAPE across SKUs with very different volumes instead of using weighted MAPE. Mixing the lag: accuracy should compare the forecast you made before the period with what happened, not a forecast revised after the fact. And dropping bias, which is the number that explains why inventory keeps drifting too high or too low.

What to do next

Forecast error feeds directly into how much buffer stock you need. Take the variability you see here into the safety stock calculator to size the buffer for your target service level.

Planning team reviewing sticky notes on a wall during a forecast meeting

Frequently Asked Questions

Forecast accuracy, answered

What is a good MAPE for demand forecasting?+
A MAPE below 20% is generally considered good for consumer goods; below 30% is acceptable for most industries. Fast-moving SKUs with stable demand often achieve 10–15% MAPE. Intermittent or new-product SKUs may exceed 40%. What matters more than the absolute number is the trend — accuracy should improve over time as models learn from more data.
What is the difference between MAPE and forecast bias?+
MAPE measures the magnitude of forecast error (always positive — it tells you how far off you are on average). Bias measures the direction of error (positive bias means you consistently over-forecast; negative means under-forecast). A forecast can have low MAPE but high bias if errors are consistently in one direction. Both should be tracked: MAPE for overall accuracy, bias for systematic skew that inflates or depletes inventory.
Why is weighted MAPE better than simple MAPE?+
Simple MAPE gives equal weight to every SKU or period regardless of volume. A 50% error on a SKU selling 10 units counts the same as a 50% error on a SKU selling 10,000 units. Weighted MAPE (wMAPE) weights each period by its actual demand, so high-volume items drive the metric more than low-volume ones. This better reflects the revenue and inventory impact of forecast errors.
How does Demand OS improve forecast accuracy?+
Demand OS runs an ensemble of statistical and ML models per SKU, automatically selects the best-performing model based on cross-validated accuracy, and retrains when accuracy drifts. Demand sensing adds daily signal detection to catch near-term shifts before the next planning cycle. Explainable AI shows ranked factor attribution for every forecast so planners know why the number changed — not just that it changed.
Can MAPE be calculated when actual demand is zero?+
No — MAPE divides by actual demand, so zero actuals produce a division-by-zero error. For periods with zero demand, alternatives include MAE (Mean Absolute Error, not a percentage), SMAPE (Symmetric MAPE, which uses the average of forecast and actual in the denominator), or simply excluding zero-demand periods from the MAPE calculation and tracking them separately.
Perimattic Demand OS

Ready to automate forecast accuracy tracking?

Demand OS computes MAPE, bias, and accuracy drift at the SKU level across your entire catalogue — continuously, not once a quarter.
  • Free consultation
  • No commitment
  • Tailored to your team