Funny Item Co-occurrences in 3.2 Million Instacart Orders
Funny item co-occurrences in 3.2M Instacart orders
I analyzed 3.2 million Instacart orders to find the strangest grocery combinations that make cashiers snicker. After realizing raw product data was too granular, I mapped items to GS1 Global Product Classification bricks using embeddings and an LLM. This approach revealed surprising pairings with the lowest statistical lift, uncovering the most unexpected shopping cart contents.
"The kind of shopping cart that makes the cashier snicker and later tell his friends, 'Dude, can you believe this guy came in and only bought condoms and apples?'"
HN discussion
- The perceived humor in co-occurring items like adult diapers and baby food stems from a 'terminally online' tendency to assume taboo sexual scenarios rather than recognizing mundane caregiving realities.
- Assuming seasonal availability for produce like strawberries or apples reflects a specific geographic privilege, as global supply chains make these items available year-round in many regions.
- The human brain naturally fabricates causal connections between unrelated items in a shopping basket, whereas the actual co-occurrence is often just a coincidence of different household members' needs or package sizes.
- Using LLM embeddings to classify products into larger categories represents a novel application of word vector techniques previously associated with word2vec.
- Bananas defy standard consumer prediction models by having the highest standard deviation in individual purchasing behavior, despite being the most frequently sold grocery item overall.