CPG / Retail bof at rstudio::global

Here's a recap of some of the activity related to CFG/Retail at rstudio::global(2021)


R in manufactuRing & consumeR products
Tuesday, March 2, 2021, 12:00 PM to 1:00 PM EST
I also wanted to share with this group that my colleague and I are organizing an industry Meetup for folks in the manufacturing and consumer products fields. We will have invited speakers from industry sharing how they're using R, what their hard lessons learned have been, best practices, and more. We hope to make it a place for networking and industry connection. Event details will be shared here: https://www.meetup.com/RStudio-Enterprise-Community-Meetup/events/275770080/


Daniel to CPG / Retail - January 21st, 1:06 PM
Do you have preferred packages for creating forecasts?

Elaine
January 21st, 2:56 PM
I have used prophet in the past but the forecasting effort involved didn't get far enough along to really validate how good a choice it was.

Elaine - January 21st, 6:03 PM
Any have experience with fable?

Daniel - January 22nd, 8:32 AM
I've started using modeltime and the packages put together by Matt Dancho at Business-Science. I really like them because it, relatively easily, allows for using multiple models (prophet, arima, xgboost, etc.) all within the same time-series-ish framework.


Daniel to CPG / Retail - January 21st, 1:06 PM
What tips do you have for creating forecasts at scale?

Vaclav - January 21st, 4:08 PM
Definitely Matt Dancho's course: https://university.business-science.io/p/ds4b-203-r-high-performance-time-series-forecasting
It is using Matt's packages timetk and modeltime as well as Python's GluonTS


Elaine to CPG / Retail - January 21st, 11:28 AM
Is anyone aware of a reliable up-to-date data source for grocery UPCs? (In my case, beer and wine are the categories of interest). I'd love to be able to verify UPCs and get properties of the products (wine varietal, or brand, for example). The only possibilities I've found don't seem to be very updated with the ever changing landscape of UPCs.

Aarron - January 21st, 12:02 PM
Two things here I'd suggest (for better or worse)... 1) Depending on what you mean by "verify UPCs", you can absolutely script some business rules around the values you have to validate the structure of the upc (upc 13/14, etc. and check digit calculations). 2) Your best bet for getting current UPCs and associated attribution might be to create some webscrapes for a site(s) that have the majority of the info you're looking for. I'd probably start with the sites of some of the bigger distributors (RNDC, etc.) rather than the odd-ball sites. Best of luck on it!