Forecast Accuracy & Forecast Value Added: MAPE, WAPE, Bias & FVA

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Calculate the three building blocks of every accuracy metric for one SKU of Fernhill Home Care, then the MAE, and see why the plain average error is not an accuracy measure.
• Calculate MAPE, WAPE and sMAPE for one month of ten SKUs and see how two tiny SKUs push the MAPE to an absurd level while WAPE tells the real story.
• Measure the bias of the laundry family forecast as a percentage, build a running tracking signal and let a formula rule flag the months where it leaves the ±4 band.
• Calculate MAE and RMSE for three dish-care SKUs with AVERAGEIFS, add the SKUs up to the family level and see how much smaller the family error is.
• Build the naive forecast for the surface spray, compare its WAPE with the statistical forecast and calculate the forecast value added of the statistical model.
• Build an FVA stairstep report for the laundry pods: the WAPE of every process step, its value added against the naive forecast and against the step before, and you name the step that destroys accuracy.
• Classify Fernhill's SKUs by demand variability (XYZ), look up a fair accuracy target per class, check which SKUs meet it and calculate the volume-weighted WAPE of the A items.
• Build Fernhill's monthly accuracy dashboard: accuracy per month for the forecast and the naive benchmark, the six-month FVA and bias, and a line chart of both accuracies.