Wanting to know which forecast error calculation to use is a common question we receive.
There is no one-size-fits-all answer, but there are a few things you can consider.
For more information on the calculations themselves, see the Forecast Error KPIs topic.
At its core, you want to select a metric that will be best for overall business performance. This discussion can (and has) fill a book, so here we will give you just some short summaries.
Mean Absolute Error
The default in StockIQ is to measure error via Mean Absolute Error Percent (MAE%).
MAE's strength lies in its consistency and relative insensitivity to outliers in your data, since each error is uniformly weighted.
Root Mean Squared Error
Commonly used in the machine learning world for evaluating the performance of ML algorithms, RMSE (and MSE) is another valid selection for your forecasting KPI, which StockIQ supports.
RMSE's strength is that it is not biased. RMSE is minimized when your forecast approaches the average (mean)
RMSE's weakness is in how much it penalizes outliers. If you have demand history with many large outliers, that would steer you more towards MAE.
Mean Absolute Percent Error (MAPE)
We do NOT recommend (nor does StockIQ support) using MAPE for measuring your forecast performance. Since it is calculated a sum of error percentages, it is skewed for large errors during periods with low demand. Because of this, MAPE rewards under-forecasting, and optimizing around this metric will result in an under-forecast bias. Do not use it.