Mean Absolute Percentage Error (MAPE) Calculator
Measure forecast error as a scale-independent percentage.
Mean Absolute Percentage Error (MAPE) Template — free, works offline, formulas included.
What Is MAPE? (And Why Should You Care?)
Mean Absolute Percentage Error (MAPE) is MAD's percentage-based cousin — instead of "off by 3.5 units," it tells you "off by 3.1%." That small shift matters more than it sounds: percentages are comparable across products no matter how big or small their volumes are, which raw-unit error isn't.
This is the metric most planners actually report up the chain, precisely because it lets you say "our forecast accuracy across the whole category is 92%" as a single, meaningful number, rather than a pile of MAD values in different units for every SKU. It's also the benchmark most industry accuracy targets are expressed in.
How Does It Work?
Each period's absolute error gets divided by that period's actual demand before averaging — turning a raw unit miss into a percentage miss. Average those percentages across all periods and multiply by 100. Because it divides by actual demand, MAPE is undefined when actual demand is zero, which is the one hard constraint on using it.
| MAPE | Interpretation |
|---|---|
| < 5% | Excellent forecast accuracy |
| 5-10% | Good forecast accuracy |
| 10-25% | Reasonable, but room to improve |
| > 25% | Poor accuracy, reconsider the method |
Real-World Example
Actual demand: 100, 120, 110, 130
Forecast: 105, 115, 108, 128
A MAPE around 3% falls solidly in the "excellent" range.
Now compare the same rougher forecast used in the MAD example (80, 140, 90, 150) against the same actual demand:
A MAPE of roughly 18% lands in "reasonable, but room to improve" — not disastrous, but noticeably worse than the 3% forecast. This is the same relative gap MAD showed (3.5 versus 20), just expressed on a scale that's directly comparable to a different product's forecast accuracy, which raw MAD numbers wouldn't be.
Key Assumptions & Limitations: When Does This Work?
MAPE breaks down completely when actual demand is zero or very close to it — dividing by a tiny number inflates the percentage error enormously even for a small absolute miss. It's also asymmetric: an under-forecast can produce a percentage error larger than 100%, while an over-forecast is mathematically capped, which can make MAPE look worse for under-forecasting than for over-forecasting even when the raw miss is the same size.
For intermittent or lumpy demand (lots of zero-demand periods), MAPE often isn't the right tool at all — other accuracy metrics designed for intermittent demand tend to work better.
5 Ways People Get MAPE Wrong
Applying it to demand that includes zeros. A single zero-demand period makes MAPE undefined or wildly misleading. Filter those periods out or use a different metric for intermittent demand.
Expecting sub-2% accuracy without advanced methods.That level of precision is rare outside of very stable, high-volume products with sophisticated forecasting behind them. A realistic target for most businesses is well above that.
Not accounting for its asymmetry. MAPE can make under-forecasting look disproportionately worse than over-forecasting of the same absolute size — worth knowing before drawing conclusions from a single number.
Averaging MAPE across wildly different products without weighting. A blended MAPE across the whole catalog can hide the fact that your highest-volume, highest-value items are forecast poorly while low-value items pull the average down.
Treating one good MAPE reading as a solved problem.Forecast accuracy drifts as demand patterns change — recheck periodically instead of assuming a good score stays good.
Industry Benchmarks & Context
The table above is a widely-used rule of thumb: under 5% is excellent, under 10% is good, 10-25% is workable but improvable, and above 25% suggests the forecasting method itself needs rethinking. High-volume, stable products (grocery staples, for instance) often achieve MAPE well under 10%; new products, fashion items, and anything demand-volatile routinely run 20-40% even with a well-run forecasting process — the benchmark should account for how inherently predictable the product actually is.
Next Steps & Related Tools
Once you know your forecast accuracy:
- Compare across forecasting methods — run the same actual demand through Moving Average, Weighted Moving Average, and Exponential Smoothing, and see which produces the lowest MAPE.
- Prioritize accuracy work by value — a high MAPE on an A-item matters more than the same MAPE on a C-item.
Learn More
Books:
- Forecasting: Principles and Practice by Rob J Hyndman and George Athanasopoulos (free online)
Standards & curricula:
- APICS (ASCM) CSCP certification curriculum (demand planning module)
General references for further study, not endorsements — verify course availability and content directly with the provider.
Interview Preparation
Questions like these come up in supply chain and operations interviews — here's how to answer them.
What is MAPE, and is it a forecasting method or an accuracy metric?
MAPE (Mean Absolute Percentage Error) is an accuracy metric, not a forecasting method — it doesn't generate a forecast, it measures how far off an existing forecast has been running, expressed as a percentage rather than raw units.
How would you explain MAPE to someone with no statistics background?
For each period, figure out what percentage the forecast missed the actual number by, then average those percentages. A MAPE of 3% means your forecast is typically off by about 3%, regardless of how big the product's sales volume actually is.
Why divide by actual demand instead of just using MAD?
Dividing by actual demand turns a raw-unit miss into a percentage miss, which makes it comparable across products with very different volumes — a MAD of 5 means something completely different for a 1,000-unit product than a 10-unit product, but a MAPE of 5% means roughly the same thing for both.
When would MAPE not be the right metric to use?
When actual demand includes zero or near-zero values — dividing by a tiny number inflates the percentage error enormously, sometimes to the point of being undefined. It's also a poor fit for intermittent or lumpy demand, where other accuracy metrics designed for that pattern work better.
What's a subtle limitation of MAPE that's easy to overlook?
It's asymmetric — an under-forecast can produce a percentage error greater than 100%, while an over-forecast is mathematically capped, so MAPE can make under-forecasting look disproportionately worse than an over-forecast of the same absolute size.
Frequently Asked Questions
- What counts as a good MAPE score?
- As a rough rule of thumb: under 5% is excellent, 5-10% is good, 10-25% is workable but improvable, and above 25% suggests the forecasting method needs rethinking. What's realistic depends heavily on how inherently predictable the product is.
- Why is my MAPE calculation showing an error or undefined result?
- MAPE divides each error by that period's actual demand, so any period with zero actual demand makes the calculation undefined. Filter out zero-demand periods or use a different accuracy metric designed for intermittent demand.
- Should I use MAPE or MAD?
- Use MAPE when comparing forecast accuracy across products with different demand volumes, since it's scale-independent. Use MAD when you want the error expressed in the same units as demand, which some people find more intuitive for a single product.
- Can I average MAPE across my whole product catalog to get one number?
- You can, but be careful — a blended MAPE can hide the fact that high-volume, high-value items are forecast poorly while low-value items pull the average down. Consider weighting by value or tracking A-items separately.