Build guide
Building a real-time logistics dispatch platform
In short
A logistics dispatch platform turns orders, driver state and locations into a live event stream that drives the map, assignment and customer updates. The hard parts are the event pipeline and fast releases without downtime. A focused team can ship version one in roughly 8-14 weeks.
Who this is for
Logistics operators moving from spreadsheets and phone calls to GPS-driven dispatch who need a working first version in weeks, not quarters.
What the product has to do
- Order intakeCreates delivery orders with pickup, drop-off and time rules.
- Event pipelineCarries location, order and driver-state events so every screen reads the same live truth.
- Driver appShows assignments, navigation steps and proof of delivery, and reports location.
- Customer appLets customers place orders and follow deliveries.
- Dispatch logicMatches orders to available drivers from rules you define.
- Live operations mapShows drivers, orders and exceptions for the operator.
- NotificationsUpdates to customers and alerts to drivers and staff.
- Delivery and release pipelineContainer builds and rolling releases so updates ship every few days without downtime.
The stack, and why
| Layer | Choice | Why |
|---|---|---|
| Mobile apps | Flutter, with Kotlin where native code was needed | One codebase for the driver and customer apps, with native code for device-specific parts. |
| Services | Spring Boot and WebFlux | Non-blocking services handle many concurrent location updates. |
| Event streaming | Apache Kafka | Location and order events are written once and consumed by dispatch, the map and notifications. Today I would also add a schema registry so event formats stay compatible. |
| Cloud and delivery | AWS with Docker and Kubernetes | Repeatable deployments and room to scale. |
What ships first, and what waits
Version one
- Order creation and a live order list
- Driver app with assignment and status updates
- Location events flowing through the pipeline to an operator map
- Manual dispatch with a clear audit trail
- Automated build and deploy
Later
- Rule-based auto-assignment
- Customer app with live tracking
- Proof of delivery with photo or signature
- Route optimisation
- Reports and billing
Build order and timeline
- 1-2 weeksOperations mappingHow orders, drivers and exceptions work today, and the first-release cut.
- 3-4 weeksEvent pipeline and servicesKafka topics, core services and an operator view on test data.
- 3-5 weeksDriver appAssignments, status updates and location reporting on real devices.
- 2-4 weeksPilot and customer appA live pilot with the operator and a first customer-facing release.
Want a number for your own scope? The app cost calculator turns platforms and features into an estimate.
Rules you cannot skip
- Driver location on Google Play
Background location is allowed only for core functionality with a prominent in-app disclosure and a Play Console declaration. Foreground access is preferred wherever it is enough.
Official source - Location foreground services on Android
A location foreground service needs the matching permission and cannot start from the background without background location access.
Official source - Location and background use on iOS
Apple requires location use to be relevant to the app's features with user consent, and allows background services only for their intended purposes, location among them.
Official source
Where these projects go wrong
- Event pipeline designed as an afterthought
If location and order events are bolted on later, screens disagree and the map lags. Design topics and event formats first.
- Event format changes breaking consumers
Dispatch, map and notifications all read the same events. Agree a versioning approach before the second consumer appears.
- Driver phones that stop reporting
Battery savers and permissions can silence location updates. Show stale drivers clearly on the operator map.
- Shipping fast without safety
Releases every few days only work with automated builds, health checks and rollbacks. Build the pipeline in the first weeks.
- Replacing spreadsheets in one step
Operators need to trust the system. Run it alongside the old process during the pilot.
Sources
How was this page made?
I drafted this page with AI assistance from the official documents listed under Sources. Found a mistake? Tell me.
Problems you will meet on this stack
Cannot perform an interactive login from a non TTY devicedocker login to Amazon ECR fails with: Cannot perform an interactive login from a non TTY deviceError: Running AWS SAM projects locally requires Docker. Have you got it installed?sam local fails: Running AWS SAM projects locally requires Dockersoftware.amazon.awssdk.core.exception.SdkClientException: Unable to load credentials from any of the providers in the chain AwsCredentialsProviderChainThe Java SDK finds no AWS credentials on an EC2 instance or in a container on itError: spawnSync docker ENOENTcdk deploy fails in a GitLab job with spawnSync docker ENOENTKubernetes status in Docker Desktop stays at starting and never reaches runningKubernetes in Docker Desktop will not start or stays on startingno basic auth credentialsdocker push to Amazon ECR stops with: no basic auth credentials