Browsing the Aisles or Browsing the App? How Online Grocery Shopping is Changing What We Buy

This paper studies the extent to which the variety and composition of online grocery shopping baskets systematically differ from offline grocery baskets using data from around two million offline (brick & mortar) and online (Instacart) trips. We use unsupervised machine learning algorithms agnostic to the channel type to infer what constitutes a regular shopping trip for each household in order to evaluate comparable offline and online trips. We find that shopping basket variety, as measured by the number of categories purchased, is significantly lower for online shopping trips. Within a given household, the Instacart baskets are significantly more similar to each other than offline baskets, suggesting that filter bubbles and past-order shortcuts may be accelerating consumer inertia. Finally, we find that Instacart shopping baskets typically have 13% fewer fresh vegetable items and 5-7% fewer items from impulse purchase categories that include candy, bakery desserts, and savory snacks. Importantly, these fresh vegetables and impulse purchases are not picked up via alternative or additional shopping trips. We show that these variety and composition differences are unlikely to be driven by price or assortment differences across the two channels or propensity to shop online due to Covid-19 risks.