San Francisco-born startup Instacart has built its success on offering to do its users’ grocery shopping for them through Instacart assistants – shoppers, check-out assistants. Obviously, the profitability of the business depends on how fast the delivery assistants can turn the shopping run around. In order to optimise delivery speed, the startup has started to use Artificial Intelligence, and more specifically deep learning. This technique broadly follows the way the neural networks in our brain work. It enables computer software tools to learn and improve all by themselves by training up on vast volumes of existing data. This approach provides a method for predicting a default order in which the quickest Instacart shoppers would collect grocery items from the shelves, avoiding unnecessary to-ing and fro-ing.
The algorithm used to draw up this sequence is trained up on a complex set of data, including the actual products, the specific store, and the shopper’s individual profile. The deep learning algorithm helps to capitalise on all the existing shopping and time data so as to improve assistants’ overall efficiency. Another service provided by Instacart is feeding supermarkets information to which they would not otherwise have access. Jeremy Stanley, VP for Data Science at Instacart, points out: “When a customer goes to a supermarket with the intention of buying 30 different items and only finds 2, the store has no way of knowing that. By comparing the items requested by the customer and those which our shopper actually found at the store, we are able to provide them with this information.”