Uber · 2022 - 2023
UberX Share: fewer confused cancellations, server-driven UI, and a 40% faster MapStore
On UberX Share I built rider flows that explain shared rides before people cancel over them, brought server-driven UI to our microservice and pitched it to the wider shared rides org, and cut MapStore's average query time 40%.
- role
- software engineer
- team
- UberX Share, shared rides and matching
- location
- New York
- worked with
- legal, policy, operations
UberX Share, from Uber's official channel.
Context
UberX Share brought shared rides back after Uber paused them in 2020. The trade is simple on paper: you might pick up a co-rider heading the same way, and in exchange you get an upfront discount and up to 20% off if you’re matched, while arriving no more than 8 minutes later than a regular UberX (Uber Newsroom).
On paper. In the app, a rider who didn’t understand that trade would see an unexpected stop or a stranger getting in, and cancel.
Flows that explain the ride before it surprises you
Some cancellations weren’t about price or wait time at all. They came from riders who didn’t understand what they had booked. I built flows to cut those cancellations, including the first-time-user experience for UberX Share in the Rider app, so the trade is explained before the ride instead of discovered during it.
Alongside that, I ran A/B experiments on the opt-in funnels and turned the research and data analysis into feature changes that shipped, lifting shared-ride adoption 4.5%. I also led the rollout across 4 markets, coordinating legal, policy and operations so the technical implementation matched each market’s regulatory requirements.
Server-driven UI
Rider flows change constantly, and every change that waits on an app release is slow to test. I onboarded server-driven UI to our microservice, so the backend describes what a screen shows and we could change and experiment with flows without shipping a new app version. Once it worked for us, I evangelized it across the shared rides and ride matching org.
MapStore, 40% faster
MapStore is an internal service other engineering teams use for map data. I redesigned the service interface and the query architecture: paginated responses that used to return entire result sets, rewrote the slowest SQL queries, and trimmed payloads so the internal tools on top of it loaded faster. Average query time dropped 40% for the teams that depend on it.
What I learned
The cheapest fix for a confusing product is often an explanation at the right moment, not a new feature. And internal infrastructure is a product too. MapStore’s users were engineers on other teams, and they noticed a 40% faster query as much as riders notice a faster pickup.