Building Supermarkets Before Demand

The core mistake was building a supermarket before earning supermarket level demand. Early online grocers filled large warehouses with tens of thousands of items, including short shelf life products like produce and meat, so unsold inventory quickly turned into waste. Later operators showed that customers do not need full supermarket breadth for the model to work, they need the right repeat purchase items in stock, at the right time, with enough density to keep inventory moving.

• Farmstead argued it could drive baskets above $80 with roughly 1,500 products, versus a traditional supermarket basket around $100 with about 30,000 products. That is the economic unlock, a much narrower assortment can still satisfy most demand if the selection is curated tightly around what turns fast.
  1 sacra
• Ultrafast operators like JOKR and Duffl built around this lesson. They use neighborhood or campus level demand data to decide what each micro warehouse carries, and they change assortment by location and time of day instead of forcing every site to mimic a full grocery store.
  2 sacra 3 sacra
• GoPuff pushed the logic furthest by positioning itself as a delivery app for convenience store items and groceries from dark stores, not as a full supermarket replacement. That shifts the job from weekly stock up shopping to urgent fill in purchases, where a smaller catalog and faster turns matter more than endless aisle breadth.
  4 sacra

The market has kept moving toward smaller, more data driven assortments. The winners are likely to look less like digital Kroger and more like software run convenience networks, adding fresh categories only when local demand is dense and predictable enough to support them without bringing back the old spoilage problem.