Personalization Drives Dark Store Assortment

This kind of personalized assortment is really a cost control system disguised as a better shopping experience. In a dark store model with only about 1,000 to 2,000 SKUs, every slot matters, so showing different customers different items helps Jokr steer demand toward the products most likely to convert in that neighborhood and time window, while keeping inventory turns high and spoilage low. That is how a 15 minute grocery app gets closer to profitability with a much smaller catalog than a supermarket.

• Jokr described its core loop as collecting data on customer type, neighborhood, and time of day, then using those patterns to decide what to buy, when to buy it, and what each customer sees. Personalization is not just merchandising, it is upstream inventory planning tied directly to procurement and category management.
  1 sacra
• This matters more in ultrafast delivery because dark stores are tiny. Jokr carried roughly 1,500 SKUs per store, versus around 30,000 in a supermarket. With that little shelf space, the app cannot be a neutral digital aisle. It has to act more like a smart bodega, surfacing the few items most likely to sell right now.
  2 sacra 3 sacra
• The economics explain why. In dark stores, labor and delivery are mostly fixed per order, so increasing basket size is the clearest path to contribution margin. Other operators like Farmstead and Duffl describe the same playbook, using demand prediction, local assortment, and rapid SKU testing to lift conversion, reduce waste, and make small footprints work.
  3 sacra 4 sacra 5 sacra

The next step is that ultrafast apps start to look less like miniature supermarkets and more like algorithmic convenience stores. The winners will use personalization to decide not only what to recommend, but what to stock, what to retire, and which local products belong in each micro market. That makes merchandising software, not just delivery speed, a core competitive advantage.