React Native + native modulesBluetooth unlockstaged rollouts per city
Overview
Overview
Spin is Ford Mobility's shared micro-mobility platform — electric scooters and bikes operating across 100+ cities in North America and Europe, serving more than 1M riders.
I spent three years on the production iOS and Android apps shipping React Native features across the full ride lifecycle: discovery, rental, in-ride state, payments, and trip history. Work spanned cross-platform React Native, native modules in Swift and Kotlin for location and Bluetooth, and the CI pipeline that got it all to the stores.
The challenge
Shared mobility apps live and die by latency.
A rider staring at a blank map walks to a competitor's scooter. The legacy app had GPS accuracy problems, a battery-draining location loop, a crash rate eating into retention, and a feature pipeline that couldn't ship fast across two platforms and a dozen city operations.
Instant mapor riders walk away
Battery awarelocation tuned per ride state
Two platformsone feature pipeline
The solution
Native where it counts, React Native everywhere else.
Rebuilt the hot paths as native modules (location, BLE, maps) exposed to React Native, so the UI stayed cross-platform without giving up native performance. Tuned location polling per rider state, added predictive map tile and scooter caching so first-paint felt instant, and gated every rider-facing release behind feature flags with staged rollouts across cities.
Native modulesCached mapFlagged rollout
Architecture
Native where it counts.
A cross-platform React Native app with native modules for the latency-sensitive paths, backed by a typed state layer and an offline-tolerant data flow. A flaky cell signal should never interrupt a ride.
Client
React Native UI
Redux Toolkit
React Navigation
Optimistic updates
Native
Swift / Kotlin modules
Core Location / FusedLocation
BLE for scooter unlock
Turbo Modules
Data
Typed API layer
Predictive map tile cache
Offline queue
Deep link dispatch
Platform
Google Maps SDK
Geofencing
Apple Pay / Google Pay
Bitrise CI + staged rollouts
One request, start to finish
Rider opens the app and location is acquired
Map streams with live scooter positions from cache + network
BLE unlock triggers a scooter in milliseconds
Trip state syncs live; offline queue catches signal drops
Payment and receipt settle at end-of-ride
Decisions
Five calls and why I made them.
Call 01
React Native + native modules, not fully native
Shared ~85% of code across iOS and Android while keeping native code for the paths that actually need it (location, BLE, maps). The alternative — two separate native codebases — meant every rider feature shipped twice.
Call 02
Adaptive location polling tied to rider state
Battery drain was killing retention. A fixed-cadence GPS loop is wasteful; tuning polling to state (idle vs. riding vs. searching) dropped polling cost dramatically without hurting accuracy where it mattered.
Call 03
Predictive map tile and scooter caching
First-paint of the map was the single biggest drop-off point. Caching the last view + pre-fetching adjacent tiles made the map feel instant even on cold starts, which directly moved conversion.
Call 04
Feature flags end-to-end, staged rollouts per city
Multi-city operations meant a bad release could strand riders in a city I'd never visited. Every rider-facing change shipped behind a flag with per-city staged rollouts and instant kill switches.
Call 05
Crash-first observability
Crashes are a hidden tax on retention you don't notice in dashboards. Crashlytics triage built into the sprint cadence, with SLOs on crash-free sessions, drove the rate down and lifted retention alongside it.
Reflections
What I took away.
Perceived performance lives in the first 300ms
Map first-paint was the single biggest drop-off point in the funnel. No rider waits for a blank map — they walk to a competitor's scooter. We threw caching, tile pre-fetch, and adaptive location polling at it until cold start felt instant, and that alone moved conversion more than any visual refresh ever did. I've used the same playbook on every app since: optimize the first 300ms like nothing else matters.
Battery drain is a silent retention killer
A fixed-cadence GPS loop looks fine in a one-day test and disastrous over a month. I didn't take this seriously until I watched weekly-active riders drop off in a cohort that had nothing to do with product quality — they were just turning location off. Tuning polling to rider state (idle vs. riding vs. searching) clawed back retention we didn't know we were losing.
Feature flags are how small teams ship safely across many cities
Shipping to 100+ city operations without flags would have meant that every release was a coordination call. Per-city staged rollouts plus kill switches meant I could ship aggressively and roll back a bad release for a single market in minutes. It changed my relationship with 'risky' changes permanently.
Outcomes
What shipped.
Riders1M+Active users across 100+ cities
App Rating4.8App Store and Google Play
Load Time-25%Faster cold start for 1M+ riders
Stack
Technology
Frontend
React Native
TypeScript
Redux Toolkit
React Navigation
Maps & Location
Google Maps SDK
MapKit
Core Location
Geofencing
Native
Swift
Kotlin
Native Modules
Turbo Modules
BLE
CI & Testing
Bitrise
GitHub Actions
Jest
Detox
XCUITest
Crashlytics
Features
What it does
Cross-platform React Native shared across iOS and Android, native modules for latency-sensitive paths
60fps live map with clustered scooter markers and predictive tile caching
BLE scooter unlock with millisecond feedback and offline-safe retry
Smart geofencing for parking and no-ride zones with client-side fallback
Deep links across auth, trip lifecycle, and transit partner apps
Apple Pay and Google Pay integration with tokenized card-on-file flow
Adaptive location polling — battery-aware background behavior tuned per rider state
Automated tests (Jest + Detox + XCUITest) on Bitrise with feature-flagged staged rollouts per city