1 of 1 people found this helpful
I think that the answers are as many as there are companies, meaning no two companies appear to forecast in the same manner. I have seen quite sophisticated ARIMA, or modified ARIMA models used in a few cases. Another way I have see this done is to forecast quarterly quantities and then to spread this into monthly or weekly buckets using a standard profile. I have also seen Holt-Winters used with adequate results, provided there isn't an overarching seasonal component too, such as Christmas.
Are you also doing some demand segmentation to determine which SKUs need this more sophisticated forecasting approach? Or do you find that all your demand follows the quarter end spike?
Thanks, Trevor. Appreciate your thoughtful and detailed response.
I remain eager to see if there are others from the user community who confirm that a variety of statistical models are in use for demand forecasting. None of the 7 companies I am intimately familiar with used statistical forecasting other than for ROP. Lumpiness of demand at the end of quarter was only one of several reasons. Lack of familiarity with statistical forecasting seemed to be the biggest reason; lack of appropriate statistical forecasting model to address the short life cycles was a close second.
Pleased to hear that you have found companies using ARIMA and Holt-Winters.
Inviting all you end-users, demand planners, master schedulers to weigh in.