Calculator inputs
Enter aligned actual and predicted values. You can also include optional weights and custom labels.
Example data table
This sample shows how a forecast review table can be structured before calculation.
| Period | Actual | Predicted | Absolute Error | APE % |
|---|---|---|---|---|
| Week 1 | 120 | 118 | 2 | 1.67 |
| Week 2 | 135 | 140 | 5 | 3.70 |
| Week 3 | 150 | 147 | 3 | 2.00 |
| Week 4 | 145 | 150 | 5 | 3.45 |
| Week 5 | 160 | 158 | 2 | 1.25 |
Formula used
MAPE = (100 / n) × Σ |(Actual − Predicted) / Actual|
Weighted MAPE = Σ(Weight × APE) / ΣWeight
WAPE = [Σ |Actual − Predicted| / Σ |Actual|] × 100
sMAPE = (100 / n) × Σ [2 × |Actual − Predicted| / (|Actual| + |Predicted|)]
RMSE = √[(1 / n) × Σ (Actual − Predicted)2]
When an actual value is zero, this calculator replaces the denominator with epsilon to avoid undefined division and keep diagnostics comparable.
How to use this calculator
- Enter the model and dataset names to label your analysis.
- Paste actual values and predicted values in the same order.
- Add optional weights if some observations matter more.
- Add labels to make the result table easier to read.
- Choose an epsilon for zero actual values and a trim rate for outlier-resistant MAPE.
- Press Calculate MAPE to show the summary above the form.
- Review the Plotly chart, row-level errors, and diagnostic metrics.
- Download CSV or PDF if you need a report copy.
FAQs
1. What does MAPE measure?
MAPE measures the average absolute percentage gap between actual and predicted values. It helps compare forecast accuracy across scales because the error is normalized by the actual value.
2. Why can MAPE be misleading with zero values?
MAPE divides by the actual value, so zero actuals make the formula undefined. This page uses an epsilon safeguard, but you should still interpret those cases carefully.
3. When should I use weighted MAPE?
Use weighted MAPE when some rows carry more business importance, such as high-revenue products, critical customers, or larger transaction volumes. It shifts the score toward priority observations.
4. What is trimmed MAPE?
Trimmed MAPE removes a selected share of the highest percentage errors before averaging. It is useful when you want a more robust view that is less driven by extreme outliers.
5. How is WAPE different from MAPE?
WAPE divides the total absolute error by the total absolute actuals. Unlike standard MAPE, it weights naturally by scale, making it helpful for demand planning and portfolio-level reporting.
6. Is a lower MAPE always better?
Lower MAPE usually indicates better fit, but context matters. A low score can still hide bias, instability, or weak performance on critical segments, so review additional diagnostics too.
7. Why does this calculator also show RMSE and bias?
MAPE alone does not show direction or variance of errors. RMSE highlights larger misses, while bias shows systematic overprediction or underprediction across the evaluated sample.
8. Can I use this tool for classification models?
This calculator is designed for regression and forecasting outputs with numeric targets. Classification tasks usually need different evaluation metrics such as accuracy, log loss, F1, or AUC.