0

I'm an intern in a company that sells about 900 products.

We are trying to forecast our sales for the next year. I already have the monthly data for the 3 previous years, cleaned it, and analysed the trends plus various kinds of seasonality.

I tried both Holt-Winters and SARIMA. I ran quite a lot of tests through Excel plugins PEERforecaster and Real-Statistics, plus a manual Excel model developped by Régis Bourbonnais, PhD to compute the data with Holt-Winters, and the automatic forecasting method included in Excel 2016.

Of everything, the automatic excel forecasting method included in Excel 2016 seems to be the most accurate, but works only for very few products.

Did I miss anything ? Is there anything that I should try instead ?

If you could give me any guidance, I would be grateful. Thank you.

Seb
  • 1
  • 1
  • Are there any drivers for sales, like promotions or other incentives? If so, have you modeled them? Are you accounting for new product introductions and/or delistings? Also, there is always some remaining unforecastability. – Stephan Kolassa Feb 27 '20 at 12:04
  • Why don't you post another question AND post an example of your data that is the MOST troublesome and I will try and help further by spotlighting a reasonable model which you can then examine.. – IrishStat Feb 27 '20 at 12:16
  • @StephanKolassa thank you for the link, it was very interesting. There is no particular driver, except for some products that I already modeled. For now I just stick with regular products, so no new products/delistings at the moment. – Seb Feb 27 '20 at 13:34
  • @IrishStat thank you for your help, I will provide you everything as soon as I can. – Seb Feb 27 '20 at 13:35

0 Answers0