Forecasting with the A/F Ratio Method

A/F Ratio

Actual demand/Forecasted demand

Building a Demand Forecast Distribution

an initial point forecast might be arrived at through quantitative or qualitative methods; we will build on the point forecast to develop a demand forecast distribution - an overview of al possible demands and their probabilities; the purpose of moving beyond a simple point forecast is to allow us to make better decisions

Step 3 of Generating a Normal Demand Forecast Distribution Using A/F Ratios

calculate the average of the A/F ratios, and the standard deviation of the A/F ratios

Step 1 of Generating a Normal Demand Forecast Distribution Using A/F Ratios

collect forecast and demand data for past products for which the forecasting task is similar to the current product of interest. the data should include an initial forecast of demand and the actual demand. we also need point forecast for the item for the upcoming season

A/F Ratio Method

demand forecast for this year's product with point forecast 'x' mean = X * (mean of A/F ratios) stdev = X * (steve of A/F ratios)

Step 4 of Generating a Normal Demand Forecast Distribution Using A/F Ratios

finally, take the point forecast of a current product (call it X). the demand forecast mean and standard deviation you should use are then: Expected Demand Mean = X * (Average of A/F ratios) Standard Deviation of Demand = X * (Standard Deviation of the A/F Ratios)

Step 2 of Generating a Normal Demand Forecast Distribution Using A/F Ratios

for each historical product, calculate the A/F ratio (Actual demand/forecast)

Key Assumptions of the A/F Ratio

forecast errors for the new product is similar to forecast errors for past products; forecast errors are bigger for larger quantities than smaller quantities

Other Assumptions of the A/F Ratio

not all items are "similar" in terms of how they are forecasted (be careful); differences are common from one product/category to another, and also differences will occur from one forecaster to another (retailer vs. manufacturer); grouping dissimilar items together can be misleading

A/F Ratio Method Allows Us

to combine past data from many forecasting tasks that were similar to the current forecasting task and adjust the current forecast according to: systematic bias in past forecasts forecasting accuracy