HN Explorer

Ask HN: How to datamine grocery stores?

9 points by evanrmurphy ↗ HN
Ideally it would be real-time inventory data, but just a list of all items a store typically carries would be a good start. Some possibilities:

  * Self-checkout screenshots
  * Crowd-source customer receipts (thanks Kliment)
  * Browse aisles with camera
  * Ask store for listing
  * Mine store websites
[Edit: Updated possibilities.]

4 comments

[ 2.8 ms ] story [ 21.5 ms ] thread
Kliment
Grocery store websites tend to show very, very little of inventory. The typical grocery store considers the data a trade secret, and are unlikely to tell you if you ask. You could walk around the store with a camera mounted to your hat (seriously) and then work from the images. You are more likely to get data from smaller stores than from big chains, but they tend to have less accurate data (weekly inventory rather than real-time). But I don't know if you will get far without the big stores. If they notice you gathering data they may get suspicious.

One trick would be to ask people to upload receipts from stores. Then you work from the product listing on the receipt. This will tell you where people are buying things also, in the sense of "Store X is where most people in this area buy product Y." Think about it, this might be the easiest way.

evanrmurphy
Crowd-sourcing receipts is an interesting approach, thanks for suggesting. The rest your comment makes a lot of sense.
Predictor
I wonder how many receipts could be acquired from the ground or garbage cans outside the store? It would be a biased sample, possibly, but it is hard data and free of cost.
ig1
Tesco offers an API (still in beta until later this year), although it's poorly documented and all the examples are in Microsoft languages...