Intelligent Buyer Routing
ByteDance / Tokopedia · Senior Product Manager · Jul 2022 – Feb 2024
Problem
Buyers were matched with sellers based on price and availability alone, without geographic proximity. This led to long fulfillment times, high shipping costs, and inflated promotional subsidies.
Approach
I designed a routing algorithm that factored in seller-buyer distance alongside price and stock availability. The system dynamically selected the optimal fulfillment location for each order, balancing speed, cost, and seller capacity.
The judgment call
We could have built this as a visible feature — “see your nearest seller” or “fastest delivery” badge. Instead, I chose to make it invisible. No new UI, no new button. The algorithm ran silently in checkout. This was a hard sell internally because invisible work does not get celebrated. But the buyer does not need to know why the order arrived faster. They just need it to arrive faster.
Outcome
Average fulfillment time dropped by several hours. Seller-buyer distance decreased measurably. Platform promotional costs fell as the routing reduced the need for subsidized shipping on long-distance orders.
What I think about differently now
The buyer never saw this feature. Some of the highest-impact product work is invisible — like the aerobic base that lets a runner hold pace without thinking about it. The best infrastructure disappears into the experience. I now evaluate every feature request with two questions: “Who notices when this works?” and “Who notices when it fails?” If the answer to the first is “nobody,” I am usually working on the right thing.