Jessica Batbayar

What does a new product cost the people it wasn’t primarily built for?

As an aspiring product manager, I got interested in what happens to the parts of a marketplace that a new product wasn’t primarily built for. I built two projects to find out what Cart Assistant does to the people who fulfill its orders and whether Instacart’s fulfillment layer is ready for it.

The shared argument

Cart Assistant knows things no one else in the fulfillment chain can see: the recipe it built the cart around, the allergy the consumer mentioned, the meal plan the whole order is in service of. When the order leaves Cart Assistant, that context disappears. Project A is about what happens when a shopper encounters an unsafe substitute without it. Project B is about what happens to a shopper’s earnings when they navigate a complex order without it. Both projects trace back to the same missing piece in the system. The fix is one pipeline serving two surfaces. That is the whole argument.

2,457 blocked

Project A

Catalog-Powered Smart Substitution via Store View

A popularity-based substitution system introduced allergen mismatches in 8.44% of cases. This engine brought that number to zero without reducing coverage by a single percentage point.

0.00% allergen mismatch rate 84.8% of OOS items matched 2,457 unsafe substitutes blocked
Hourly Earnings 0% 25% Cart Assistant Adoption −12.6% at 25% adoption

Project B

Shopper Earnings at Risk

Cart Assistant was built to make grocery shopping smarter for consumers. This project asks what that costs the person who walks into the store and fills the cart.

48.5% lower hourly earnings on complex orders −12.6% shopper earnings at 25% Cart Assistant adoption 3.2M orders analyzed