Lord Ganesha and Machine Learning
I live in the city of Frisco, and I frequently shop at a local Kroger for some of my groceries. On the 9th of Sep, I went to Kroger looking for a few specific things, flowers being one of them.
Although I was expecting it, I found the flower section nearly empty with a lot of shoppers like me scrapping the leftover flowers. Although I had planned to buy a bunch of roses, I ended up buying Orchids.
I again visited the same store yesterday, and the flower section was full, but hardly anyone was buying.
While am pretty sure the Analytics team at Kroger is savvy and would have noticed this peak of flower demand. Their forecasting model would probably factor in peak demand in Sep.
However, am curious about how they would deal with this next year because there is a cultural nuance behind the peak. Frisco has a very large South Indian (East India) diaspora and Sep 9th, and 10th was a special festival called “Ganesh Chaturthi”. This day most Indian South Indian Hindu households would pray to Lord Ganesha, and this is where the flowers are used. It's common that bouquets are broken into individual flowers minus the stem to be used in the prayers.
领英推荐
This cultural nuance is certainly not yet known to Kroger, otherwise, they would stock flowers differently. In a larger context, this may not be relevant for Kroger because lost flower sales at a few stores will just be a blip in their annual sales, but this is certainly an opportunity.
In 2017 soon after I had moved to Frisco, I had created a few slides that I wanted to send to Kroger. I estimated the grocery expenses by Indian at $35 M a year and remember sharing with a few Kroger folks before I forgot all about it.
Frisco’s Indian population has grown dramatically since but while the real estate industry has understood the Indian buying behavior, the large retailers have not.
The many Indian festivals and the fact that those follow a Lunar calendar would give the machine learning models a tough time.
Next year Ganesh Chaturthi would fall around Aug 31st. I really hope Kroger doesn’t stock up flowers for Sep 9th!!!! I would definitely be watching.
Technology Leader | Data Engineering, AI & Analytics | Data Platforms (GCP, AWS, Azure) | Data Products | Business Intelligence | Product | Strategy | Global Teams | Supply Chain Transformation | Fulfillment Networks
3 年Good writeup. Grocery demand planning models generally are not planned at a SKU level. Although it is not an easy problem to address, one way of approaching is to consider latest demographics and SKU demand data and forecast at Store-SKU level.
Senior Manager - Marketing Lead at Bizmetric (Head of Marketing)
3 年Even the flower sales may have been a fraction of their overall sales, but hopefully they would have explored as to the reasons behind the sudden burst in flower sales during this period and maybe they would be better prepared to meet the demand for flowers next Ganesh Chaturthi. A great example of an opportunity area for not only Machine Learning helping address the changing demand for flowers but also an opportunity to appreciate how the culture of the local population impacts an organization's sales - even though this also may be less of a factor affecting their annual balance sheets.
Very interesting Maruti S Peri