How To Calculate Mean Percentage Error In Excel

Calculate the accuracy of your forecasts in Excel with this interactive tool

Complete Guide: How to Calculate Mean Percentage Error (MPE) in Excel

The Mean Percentage Error (MPE) is a critical metric for evaluating forecast accuracy, particularly in business forecasting, economics, and data science. Unlike other error metrics, MPE provides insight into the direction of forecast errors (over-forecasting vs. under-forecasting) while maintaining a percentage scale that’s easy to interpret.

What is Mean Percentage Error?

Mean Percentage Error measures the average percentage difference between actual values and forecasted values. The formula is:

MPE Formula:

MPE = (1/n) × Σ[(Actual – Forecast)/|Actual|] × 100%

Where:
n = number of observations
Σ = summation symbol
|Actual| = absolute value of actual observation

Why Use MPE Instead of Other Error Metrics?

While metrics like Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) are popular, MPE offers unique advantages:

• Directional Insight: Positive MPE indicates systematic under-forecasting; negative MPE indicates over-forecasting
• Scale Independence: Percentage errors are unit-less, allowing comparison across different scales
• Business Relevance: Percentages are intuitively understandable for stakeholders
• Sensitivity Analysis: Helps identify whether errors are consistent across different magnitude values

Step-by-Step: Calculating MPE in Excel

1. Organize Your Data:

   Create two columns in Excel: one for Actual Values and one for Forecast Values. Ensure they’re aligned row-by-row.

   A	B
   Actual	Forecast
   100	105
   120	118

2. Calculate Percentage Errors:

   In column C, create a formula to calculate the percentage error for each pair:

   = (A2-B2)/ABS(A2)

   Drag this formula down for all rows.

3. Compute the Average:

   Use Excel’s AVERAGE function to calculate the mean of your percentage errors:

   =AVERAGE(C2:C100)*100

   Multiply by 100 to convert to percentage format.

4. Format the Result:

   Right-click the result cell → Format Cells → Percentage with 2 decimal places.

Advanced MPE Applications in Excel

Conditional Formatting

Apply color scales to visualize error distribution:

1. Select your percentage error column
2. Home → Conditional Formatting → Color Scales
3. Choose a red-yellow-green scale

This helps quickly identify periods with extreme errors.

Error Analysis Dashboard

Create a dynamic dashboard with:

• Sparkline charts for error trends
• Slicers to filter by time periods
• Pivot tables summarizing errors by category

MPE vs. Other Forecast Accuracy Metrics

Metric	Formula	Pros	Cons	Best For
Mean Percentage Error (MPE)	(1/n) × Σ[(A-F)/|A|] × 100%	Shows bias direction Unit-less comparison Business-friendly	Undefined for zero actuals Asymmetric treatment of errors Can be dominated by small values	Bias analysis, business reporting
Mean Absolute Error (MAE)	(1/n) × Σ|A-F|	Easy to understand Same units as data Less sensitive to outliers	No directional information Scale-dependent	General accuracy assessment
Root Mean Squared Error (RMSE)	√[(1/n) × Σ(A-F)²]	Penalizes large errors Mathematically elegant	Sensitive to outliers Harder to interpret Scale-dependent	Model comparison, statistical analysis
Mean Absolute Percentage Error (MAPE)	(1/n) × Σ|(A-F)/A| × 100%	Percentage interpretation Scale-independent	Undefined for zero actuals Biased toward small values	Percentage accuracy reporting

Common Pitfalls and How to Avoid Them

1. Division by Zero:

   MPE becomes undefined when any actual value is zero. Solutions:

   • Use =IF(A2=0,0,(A2-B2)/ABS(A2)) to handle zeros
   • Consider Mean Absolute Scaled Error (MASE) as alternative
2. Asymmetric Error Treatment:

   MPE treats over- and under-forecasts differently due to the absolute value in the denominator. For symmetric treatment, consider:

   =AVERAGE((Actual-Forecast)/Actual)*100
3. Small Value Dominance:

   When actual values vary greatly in magnitude, small values can dominate the MPE. Mitigation strategies:

   • Segment analysis by value ranges
   • Use weighted MPE where weights reflect value importance

Real-World Applications of MPE

Supply Chain Forecasting

A major retailer used MPE to evaluate their demand forecasting system across 500 SKUs. By analyzing MPE by product category, they identified that:

• Perishable goods had MPE of +8.3% (consistent under-forecasting)
• Electronics had MPE of -3.1% (slight over-forecasting)

This led to category-specific safety stock adjustments, reducing stockouts by 22% while maintaining 98% service levels.

Financial Projections

A Fortune 500 company compared three forecasting methods for quarterly revenue:

Method	MPE	MAE ($M)
Time Series (ARIMA)	-2.4%	$12.3
Machine Learning (XGBoost)	+1.8%	$9.7
Judgmental Forecasts	+5.2%	$15.1

The MPE revealed that judgmental forecasts consistently under-estimated revenue, while ARIMA slightly over-estimated. The company implemented a hybrid approach combining XGBoost with ARIMA residuals.

Academic Research on Forecast Error Metrics

The selection and interpretation of forecast error metrics has been extensively studied in academic literature. Key findings include:

• Hyndman & Koehler (2006) demonstrated that percentage errors can be misleading when actual values vary greatly in scale. They recommend scale-dependent metrics like MASE for such cases.
  Another look at measures of forecast accuracy (Monash University)
• The M4 Competition (2018) analyzed 100,000 time series and found that error metric choice significantly impacts method ranking. MPE was particularly useful for identifying biased forecasts in economic data.
  M4 Competition results (International Journal of Forecasting)
• The U.S. Census Bureau uses MPE variants to evaluate their economic indicators. Their research shows that MPE is particularly valuable for:
  • Seasonally adjusted series
  • High-frequency economic data
  • Comparing forecast performance across different economic sectors
  Census Bureau Forecasting Methodology

Excel Automation with VBA

For frequent MPE calculations, create a custom VBA function:

VBA Function for MPE:

Function MPE(actualRange As Range, forecastRange As Range) As Double
    Dim i As Long
    Dim sumPE As Double
    Dim count As Long
    Dim pe As Double

    count = 0
    sumPE = 0

    For i = 1 To actualRange.Rows.Count
        If actualRange.Cells(i, 1).Value <> 0 Then
            pe = (actualRange.Cells(i, 1).Value - forecastRange.Cells(i, 1).Value) / Abs(actualRange.Cells(i, 1).Value)
            sumPE = sumPE + pe
            count = count + 1
        End If
    Next i

    If count > 0 Then
        MPE = (sumPE / count) * 100
    Else
        MPE = CVErr(xlErrDiv0)
    End If
End Function

Usage: =MPE(A2:A100, B2:B100)
Note: This function automatically handles zero values in the actual series.

Alternative Implementations

Python Implementation

import numpy as np

def mean_percentage_error(actual, forecast):
    """
    Calculate Mean Percentage Error

    Parameters:
    actual (array-like): Array of actual values
    forecast (array-like): Array of forecast values

    Returns:
    float: MPE value in percentage
    """
    actual = np.asarray(actual)
    forecast = np.asarray(forecast)

    # Handle division by zero
    non_zero = actual != 0
    if not np.any(non_zero):
        return np.nan

    pe = (actual[non_zero] - forecast[non_zero]) / np.abs(actual[non_zero])
    return np.mean(pe) * 100

R Implementation

mean_percentage_error <- function(actual, forecast) {
  # Calculate Mean Percentage Error
  # actual: vector of actual values
  # forecast: vector of forecast values

  # Remove cases where actual is zero
  valid <- actual != 0
  if (sum(valid) == 0) {
    return(NA)
  }

  pe <- (actual[valid] - forecast[valid]) / abs(actual[valid])
  return(mean(pe) * 100)
}

Frequently Asked Questions

Q: Can MPE exceed 100%?

A: Yes, MPE can theoretically exceed 100%. This occurs when the average percentage error across all observations is greater than 100%. For example, if you consistently forecast 50 when the actual value is 100, your percentage error for each observation would be (100-50)/100 = 50%, but if you have some observations where you forecast 0 when actual is 100, those would contribute -100% errors, potentially driving the average beyond ±100%.

Q: How does MPE differ from MAPE?

A: The key differences are:

Aspect	MPE	MAPE
Error Direction	Preserved (shows bias)	Always positive
Interpretation	Average percentage error with direction	Average absolute percentage error
Use Case	Identifying forecast bias	Measuring absolute accuracy

Q: What's a good MPE value?

A: "Good" MPE values are domain-specific, but general guidelines:

• |MPE| < 5%: Excellent forecast with minimal bias
• 5% ≤ |MPE| < 10%: Good forecast with some bias
• 10% ≤ |MPE| < 20%: Acceptable but may need improvement
• |MPE| ≥ 20%: Significant bias - investigate forecast method

Note that industry standards vary. For example:

• Retail demand forecasting often targets |MPE| < 10%
• Financial projections may accept |MPE| up to 15% due to volatility
• Weather forecasting aims for |MPE| < 5% for temperature predictions

Conclusion and Best Practices

Mean Percentage Error is a powerful tool for forecast evaluation when used appropriately. Remember these best practices:

1. Combine with Other Metrics: Always use MPE alongside scale-dependent metrics like MAE or RMSE for complete analysis
2. Segment Your Analysis: Calculate MPE separately for different product categories, time periods, or value ranges
3. Visualize Errors: Create control charts of percentage errors over time to identify patterns
4. Document Assumptions: Clearly state how you handled zero values, outliers, and data transformations
5. Benchmark Continuously: Track MPE over time to monitor forecast performance improvements

For advanced forecasting applications, consider implementing forecast value added (FVA) analysis which decomposes MPE by forecasting process stage, or probabilistic forecasting which evaluates entire prediction distributions rather than point forecasts.

Ready to calculate your MPE?

Use our interactive calculator at the top of this page, or implement the Excel formulas in your own spreadsheets.

For questions about advanced forecasting techniques, consult the Forecasting Principles resource from leading forecasting researchers.