Mean absolute percentage error calculator

Mean absolute percentage error (MAPE)
—

Enter actual and forecast values to get started.

Rating
—
Accuracy
—
Rows used
—
WAPE
—
sMAPE
—
MAE
—

Forecast accuracy gauge

The needle points at your MAPE. The coloured zones follow the common rating scale: under 10% highly accurate, 10–20% good, 20–50% reasonable, over 50% inaccurate.

Percentage error per row

Bars to the right mean the forecast was too low. Bars to the left mean it was too high. The dashed lines mark ± MAPE.

Step-by-step working

How the MAPE is calculated with your numbers.

    #ActualForecastError|Error| ÷ |Actual|% error

    How mean absolute percentage error calculator works

    The mean absolute percentage error calculator works in 4 steps: it subtracts each forecast from its actual value, divides the error by the actual value, drops the sign and multiplies by 100, then averages the percentage errors. The result is the mean absolute percentage error (MAPE), a unit-free percentage that compares forecasts for items of any size.

    1. ErrorAᵢ − Fᵢ for every row.
    2. ScaleDivide each error by its actual value Aᵢ.
    3. PercentDrop the sign and multiply by 100.
    4. AverageAdd the percentages and divide by n.

    Rows with an actual value of 0 are skipped, since division by 0 is undefined. The calculator reports 7 results: MAPE, a rating, forecast accuracy, rows used, weighted absolute percentage error (WAPE), symmetric MAPE (sMAPE) and mean absolute error (MAE). The example gives MAPE = 11.25%.

    0%10%20%10%200 vs 18020%50 vs 605%120 vs 12610%80 vs 72MAPE 11.25%
    Each bar is one row's absolute percentage error. The line marks the average, 11.25%. The 50-unit row misses by only 10 units but carries the largest error, 20%.

    Where 11.25% sits on the Lewis (1982) scale

    Under 10%Highly accurate
    10–20%Good
    20–50%Reasonable
    Over 50%Inaccurate

    Mean absolute percentage error calculator formula

    The mean absolute percentage error calculator formula is MAPE = Σ|(Aᵢ − Fᵢ) ÷ Aᵢ| ÷ n × 100%: the average of the absolute errors as fractions of the actual values, written as a percentage.

    • MAPE: mean absolute percentage error, the typical miss in percent.
    • Σ (sigma): add the percentage error of every row.
    • | |: absolute value. A 10% over-forecast and a 10% under-forecast both count as 10%.
    • Aᵢ: the actual value of row i, used twice: in the error and as the divisor.
    • Fᵢ: the forecast of row i.
    • n: the number of rows with a non-zero actual value.

    The formula with the example data

    MAPE = (|20 ÷ 200| + |-10 ÷ 50| + |-6 ÷ 120| + |8 ÷ 80|) ÷ 4 × 100% = (10% + 20% + 5% + 10%) ÷ 4 = 11.25%

    4 percentage and absolute error formulas

    MeasureFormulaExample value
    Mean absolute percentage error (MAPE)Σ|(Aᵢ − Fᵢ) ÷ Aᵢ| ÷ n × 100%11.25%
    Weighted absolute percentage error (WAPE)Σ|Aᵢ − Fᵢ| ÷ ΣAᵢ × 100%9.78%
    Symmetric MAPE (sMAPE)Σ|Aᵢ − Fᵢ| ÷ ((Aᵢ + Fᵢ) ÷ 2) ÷ n × 100%11.03%
    Mean absolute error (MAE)Σ|Aᵢ − Fᵢ| ÷ n11 units

    MAPE weights every row equally. WAPE weights rows by size, so the 200-unit row counts 4 times as much as the 50-unit row. sMAPE stays between 0% and 200%.

    Mean absolute percentage error calculator technique

    The clearest mean absolute percentage error calculator technique is the 6-column percentage error table: actual, forecast, error, absolute error, absolute error ÷ actual and percentage error, with the last column averaged. The fraction column shows which rows inflate MAPE.

    AᵢFᵢAᵢ − Fᵢ|Aᵢ − Fᵢ||e| ÷ AᵢAPE
    20018020200.110%
    5060-10100.220%
    120126-660.055%
    8072880.110%
    Σ 450Σ 44÷ 4 = 11.25%
    0%5%10%15%11.25%MAPE9.78%WAPE11.03%sMAPE
    WAPE (9.78%) sits below MAPE (11.25%) for the example. The small 50-unit row carries the largest percentage error, and WAPE gives that row less weight.

    The small-number effect

    ActualForecastMissPercentage error
    4 units6 units2 units50%
    400 units402 units2 units0.5%

    The same 2-unit miss costs 100 times more MAPE on the small row. Use WAPE for data with many small or near-zero actual values, such as daily sales of slow-moving items. Excel and Google Sheets compute MAPE with =AVERAGE(ABS((A2:A5-B2:B5)/A2:A5))*100, and Python (scikit-learn) returns MAPE as a fraction with mean_absolute_percentage_error(y, y_pred).

    Frequently asked questions

    How do you calculate MAPE?

    Calculate MAPE by dividing each absolute error by its actual value, averaging those fractions and multiplying by 100: MAPE = Σ|(Aᵢ − Fᵢ) ÷ Aᵢ| ÷ n × 100%. For actual values 200, 50, 120, 80 and forecasts 180, 60, 126, 72, the percentage errors are 10%, 20%, 5% and 10%, and MAPE = 45% ÷ 4 = 11.25%.

    Can MAPE be over 100%?

    Yes, MAPE can be over 100%. MAPE has no upper limit when forecasts run high: a forecast of 30 for an actual value of 10 gives a 200% error. MAPE for forecasts that run low stays at or below 100%, so MAPE penalizes over-forecasts more than under-forecasts.

    How do you convert MAPE to forecast accuracy?

    Convert MAPE to forecast accuracy with accuracy = 100% − MAPE. A MAPE of 11.25% gives a forecast accuracy of 88.75%. The calculator shows 0% accuracy when MAPE reaches 100% or more.

    Is MAPE the same as mean absolute error?

    No, MAPE is not the same as mean absolute error (MAE): MAPE reports the typical miss as a percentage of the actual value, and MAE reports the typical miss in data units. The example data gives MAE = 11 units and MAPE = 11.25%. MAPE compares accuracy across products of different sizes; MAE does not.

    What is a good MAPE?

    A good MAPE is under 20%: Lewis (1982) rates under 10% as highly accurate, 10–20% as good, 20–50% as reasonable and over 50% as inaccurate. The right target depends on the data. Daily sales of 1 product are harder to forecast than yearly totals.

    What happens when an actual value is 0?

    An actual value of 0 makes MAPE undefined for that row, so the calculator skips the row and reports how many rows it skipped. WAPE and sMAPE keep those rows, so check WAPE and sMAPE for data with zeros.

    Why is my MAPE so high when the errors look small?

    MAPE is high when the errors fall on small actual values: a miss of 2 on an actual value of 4 is a 50% error, and the same miss on 400 is a 0.5% error. WAPE gives a fairer picture when small values dominate MAPE.

    What is the difference between MAPE and WAPE?

    MAPE averages the percentage error of each row, and WAPE divides the total absolute error by the total actual value. Every row counts equally in MAPE. Large rows count for more in WAPE, which makes WAPE common in retail and supply chain reporting.