0

I have already a SARIMA model working at my company (an ecommerce) to predict sales. Right now I am only trying to improve it.

The current model is only using endogenous variables (i.e the sales variable). Commercial teams claims that Stock info should be really useful to help predicting sales, as it is very correlated to our goal metric (they have an study on it)

That beeing said I am trying to include this exogenous variable to the model, but I aint getting no improvment (actually it is getting a little worse). I have already shown them the result but they arent getting convinced with it and I also dont know how to explain why this is happne.

Can someone help me undertand it ??

Miguel
  • 1
  • We likely won't be able to help you without actually seeing your data. That said, stocks can be argued to influence sales in opposite direction: if your stocks are low, people might get all "have to buy this before it is out", but conversely, high stocks are sometimes thought to drive "product pressure" (and there is at least one article that found this). When I see a predictor that could reasonably have opposite effects, I get very wary of it. Also, if stocks are highly correlated with sales, are you looking at lags? Then the correlation could simply be that stocks go down when you sell. – Stephan Kolassa Apr 11 '23 at 13:28
  • There is a big can of worms in here. You may find this interesting: https://stats.stackexchange.com/q/222179/1352 – Stephan Kolassa Apr 11 '23 at 13:29
  • Sales and stock could be related because when sales are increasing, your company produces more product, thereby increasing stocks. In this case the causation would be in opposite direction and thus no forecast improvement is to be expected. Or phrased differently: just because sales/stock are correlated, does not mean that you should expect one to be useful in predicting the other. – user2520938 Apr 13 '23 at 13:59

0 Answers0