Apoorva Mehta at Startup School NY 2014

After his learning plateaued at Amazon, Mehta quit in 2010 and moved from Seattle to San Francisco to start a company. He and his co-founder tried around 20 products, including an analytics platform and a social network for lawyers. The lawyer product failed because they were solving a problem they did not understand or care about.

Mehta had long disliked grocery shopping, so he built Instacart to deliver groceries to customers’ doors within an hour. He tested the first version by ordering groceries for himself on June 2, 2012, then noticed that friends used it without being pushed. When the YC application deadline had passed, he sent a beer order to partner Garry Tan and was invited to meet the partners.

Mehta says founders need resilience and should handle work manually while they learn what customers need. He joined a shopper to deliver 202-liter bottles of soda for a YC dinner, and personally fulfilled orders when no shopper was available. Early customer calls also exposed product problems, such as a banana listing that led a shopper to bring 10 bushels instead of 10 bananas.

Investors hesitated because earlier grocery delivery companies such as Webvan and Kozmo had failed, but Instacart closed a seed round. With no Trader Joe’s product catalog available, the team bought one of every item, photographed them, and added them to the service. Mehta says the company then expanded from the Bay Area to Chicago, Boston, and DC; by the talk, it served 10 U.S. cities and reported 10 percent weekly revenue growth for 20 weeks.

As Instacart grew, it had to coordinate orders of different sizes and schedules, shoppers with different speeds, and stores with varying selections and inventory. Mehta describes this as a difficult computer science and operations research problem: providing an Amazon-like experience without warehouses, trucks, or inventory. He says the company is only one percent of the way toward its goal and emphasizes persistence through repeated failures.