What is MAPE and What's a Good MAPE Score?

Quick Answer: MAPE (Mean Absolute Percentage Error) measures how far off a demand forecast is from actual sales, expressed as a percentage. A MAPE of 20% means forecasts were off by an average of 20%. For consumer goods, a MAPE below 20% is considered good. Below 10% is excellent. Above 40% indicates a forecast that needs significant improvement.

MAPE (Mean Absolute Percentage Error) is the most widely used metric for measuring demand forecast accuracy. It expresses forecast error as a percentage of actual demand, making it easy to compare accuracy across products with very different sales volumes — a 100-unit error on a 200-unit SKU (50% MAPE) is far worse than a 100-unit error on a 10,000-unit SKU (1% MAPE).

MAPE is the standard accuracy metric in demand planning because it is intuitive, scale-independent, and easy to communicate to non-technical stakeholders.

How to Calculate MAPE

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

In plain language: for each period, calculate the absolute percentage difference between what you forecast and what actually sold. Average those percentages across all periods.

Example:

Month	Forecast	Actual	Absolute % Error
January	500	450	11.1%
February	480	510	5.9%
March	600	520	15.4%
Average			10.8% MAPE

What is a Good MAPE Score?

MAPE benchmarks vary significantly by product type, industry, and planning horizon. Use these as general reference points:

MAPE Range	Interpretation
< 10%	Excellent — high-accuracy forecasting
10–20%	Good — meets planning needs for most use cases
20–30%	Acceptable — review root causes and refine inputs
30–50%	Poor — significant improvement needed
> 50%	Very poor — forecasting model or data has fundamental issues

Context matters. A 30% MAPE for a new product with 3 months of history is understandable. A 30% MAPE for a stable, fast-moving SKU with 2 years of history is a problem worth investigating.

MAPE Limitations to Know

MAPE has three well-known limitations:

1. Undefined at zero — MAPE cannot be calculated when actual sales are zero (dividing by zero). This is a problem for slow-moving or intermittent-demand SKUs.

2. Asymmetric penalty — MAPE penalizes over-forecasting and under-forecasting differently. Under-forecasting a product with actual sales of 10 units (forecast: 8) shows as 20% error. Over-forecasting the same product (forecast: 12) shows as 20% error too — but the business impact is different.

3. Sensitive to low-volume items — A single-unit miss on a product that sold 5 units produces a 20% error. This can inflate MAPE for slow-moving SKUs in a portfolio context.

Alternatives to consider alongside MAPE:

Bias : Measures whether the forecast systematically over-predicts or under-predicts. A low MAPE with high bias indicates a structural problem.

Weighted MAPE (WMAPE) : Weights each SKU's error by its revenue or volume contribution — giving more influence to your highest-impact items.

MAE (Mean Absolute Error) : Raw unit error without percentages; useful when items have similar volumes but problematic for cross-SKU comparison.

How to Improve MAPE

• Clean your historical data — Remove stockout periods, correct data entry errors, and separate promotional demand from baseline

• Increase data history — More history improves seasonal pattern detection

• Apply manual overrides — For high-value SKUs with known upcoming changes, use human judgment to refine the statistical baseline

• Track error by SKU tier — Focus improvement efforts on your A-tier SKUs first, where accuracy has the most financial impact

• Review bias regularly — If you're consistently over- or under-forecasting, the model inputs need adjustment

How Moselle Tracks MAPE

Moselle calculates forecast accuracy — including MAPE — at the SKU level and surfaces results in the Forecast Performance Report. You can identify which products have the highest error, how accuracy has trended over time, and where to focus forecast refinement effort before placing purchase orders.

https://github.com/Moselle-io/moselle-docs/blob/master/faq/analytics/reporting/moselle-reports/performance-reports/demand-forecast-performance.mdHow Accurate is AI Demand Forecasting?

Frequently Asked Questions

Should I measure MAPE at the SKU level or the category level?

Answer: Both. SKU-level MAPE identifies which specific products have problems. Category-level MAPE gives a useful operational benchmark and is less sensitive to noise from individual SKU outliers.

How do stockouts affect MAPE?

Answer: Stockouts artificially suppress actual sales, making the forecast look like an over-prediction. MAPE measured against stockout-affected actuals will be inflated. Best practice is to exclude stockout periods or use unconstrained demand estimates when calculating accuracy.

What's a realistic MAPE target for my first 90 days using a new planning tool?

Answer: Expect MAPE to improve as your data quality improves and as the system accumulates more history. Setting a target of 20–30% MAPE in the first quarter and tracking improvement monthly is a reasonable starting framework.