Hacker News Reader

Comment by m12k | Hacker News Reader

From the article:

"There are 24-hour surveillance cameras too, which alert the store's manager Domenica Gerlach if there's a break-in or a stock spillage.

She looks after four stores in the region, usually visiting once a week to clean, stack the shelves and put together click-and-collect orders made online. Lifvs uses artificial intelligence to work out what stock to order for each store, based on the data it collects about locals' shopping habits"

View on HN

> Lifvs uses artificial intelligence to work out what stock to order for each store

The "AI" is an if test

  if stock < stock_threshold:
    buy_more(stock)

You can use statistics to determine what can of stock you should favor (e.g. winter clothes vs summer clothes) and the quantity (more Christmas food in December less after). If you have enough data from previous years or a good heuristics, you can start this one easily. Even a plain old markov chains could help.

I'm stunned that week after week, I'll come across the same shelf, with one variant of a thing sold out by midweek, and the unpopular variant next to it virtually untouched, as always. The amount of statistics or programming required to prevent this, and balance the stock to what the market shows that it wants, is fairly trivial, and yet my local supermarket has failed to do so for several decades now.

I had a good chuckle in March of last year when I realized the shelves had plenty of "all natural, non-toxic cleaner" sort of products during the run on Clorox wipes and bleach.

The on hand inventory is most likely wrong in the system, so it probably thinks there are phantom units available. This happens a lot with products that could have been lost, stolen or expired and the system was not updated.

The article also speaks of "fresh fruit and vegetables" though. I don't think visiting once a week would be enough for many fruits and vegetables.

Reply on news.ycombinator.com