The toothpaste analogy really points out quite insidious thought schemes.
I catch myself thinking about this when I see a pineapple or some meat at the supermarket about to expire.
Surely it would be a good deed to buy that item - otherwise it would go to waste entirely, even if it is plain overconsumption on my part.
But then the supermarket's metrics will show they sold all their stock and will order the same amount again next time - so in the end the overconsumption feeds itself.
I'm really glad that a lot of supermarkets around here (in Switzerland) put products on the edge of (legally speaking) hitting their expiry date on sale and then on apps like Too Good To Go (where you can pick up an assortment of food that would otherwise get thrown out for a low price).
I do hope that internally, the analytics are accurate enough to capture the actual margin on each product to reduce the amount of stock ordered ahead of time.
The problem is that demand has variance, both temporally, and quantitatively. They don't sell exactly 12 pineapples each week. One week, they sell 9 and the next week they sell 15. And someone is having a fun Pacific Island-themed party and buys 12 in one day. So they generally used weighted averages coupled with seasonality to order, but even then, demand is not predictable. It's within a certain range, and they generally aim for the middle of that range when ordered. That's assuming infinite supply, which obviously isn't the case.
And no, AI isn't any better at it than humans, because AI is just trying to predict the behavior of humans, which is fundamentally chaotic and prone to variance.
Comments
The toothpaste analogy really points out quite insidious thought schemes.
I catch myself thinking about this when I see a pineapple or some meat at the supermarket about to expire.
Surely it would be a good deed to buy that item - otherwise it would go to waste entirely, even if it is plain overconsumption on my part.
But then the supermarket's metrics will show they sold all their stock and will order the same amount again next time - so in the end the overconsumption feeds itself.
I'm really glad that a lot of supermarkets around here (in Switzerland) put products on the edge of (legally speaking) hitting their expiry date on sale and then on apps like Too Good To Go (where you can pick up an assortment of food that would otherwise get thrown out for a low price).
I do hope that internally, the analytics are accurate enough to capture the actual margin on each product to reduce the amount of stock ordered ahead of time.
The problem is that demand has variance, both temporally, and quantitatively. They don't sell exactly 12 pineapples each week. One week, they sell 9 and the next week they sell 15. And someone is having a fun Pacific Island-themed party and buys 12 in one day. So they generally used weighted averages coupled with seasonality to order, but even then, demand is not predictable. It's within a certain range, and they generally aim for the middle of that range when ordered. That's assuming infinite supply, which obviously isn't the case.
And no, AI isn't any better at it than humans, because AI is just trying to predict the behavior of humans, which is fundamentally chaotic and prone to variance.