# Demand forecasting

I am forecasting number of phone calls (y) i ll be getting based on the products sold(x) and I am doing the following: (forecasted y) = (y/x averaged for past three weeks) * (forecasted x). 1. Please let me know if this is apt to do there is any other method I can try. 2. Also how should I calculate the percentage change in y based on the percentage change in x

I feel this method is flawed because forecasted y are already forecasted using x historical and then by multiplying by forecasted x am I increasing the impact of x on y

Thank you so much everyone!

• Where does that formula come from? – user2974951 Oct 2 '19 at 11:30
• Historically it is assumed that there is a linear dependence of y on x. It is more like (Forecasted y = (historical y/historical x) * (forecasted x) – curious_raven Oct 2 '19 at 15:50
• But are you just assuming this is true, or do you have some data to back it up - your formula? My question is, why are you fixing your formula to be such? Why not use a model to determine this? – user2974951 Oct 3 '19 at 6:26
• Historically this formula worked well. However, now I am seeing a huge MAPE in the forecasted numbers. I also tried y as a rate of time using simple moving averages and ARIMA in R, the results were okay but I want to utilize the x dependent variable. What will be a starting point to redo this. Thank you so much! – curious_raven Oct 3 '19 at 18:51
• Why not let ARIMA decide the model? You can include exogenous variables in ARIMAX, so that is not an issue. – user2974951 Oct 4 '19 at 9:52