Build guide
Building a field survey app that works with no signal
In short
A field data app has to work with no signal and sync cleanly later. The core is an offline queue, a form engine driven by configuration, roles for collectors and checkers, and GPS check-ins. The hard part is conflict-free sync. A first release takes roughly 12-18 weeks.
Who this is for
Research firms, enterprises and NGOs that send teams door to door and need verified data back, even from areas with weak coverage.
What the product has to do
- Offline queue and syncStores records on the device and uploads them when a connection returns, without duplicates or lost data.
- Dynamic form engineRenders multi-step surveys from configuration, so a new questionnaire does not need a new app release.
- Roles and workflowsCollectors, back checkers and supervisors each see their own tasks and screens.
- GPS attendanceCheck-in with location so supervisors know where and when teams worked.
- Media capturePhoto, audio and video attached to records and compressed on the device before upload.
- Back-check and reviewA second person re-verifies a sample of records and flags problems.
- Multi-language and right-to-left interfaceThe whole app in the languages the field team speaks, with correct mirrored layouts for right-to-left scripts.
- Supervisor dashboardProgress, quality flags and team activity.
The stack, and why
| Layer | Choice | Why |
|---|---|---|
| App framework | Flutter and Dart | One codebase for Android and iOS, with strong right-to-left and localisation support. |
| State management | BLoC and Cubit | Clear, testable state for long multi-step forms. |
| Dependency injection | get_it | Keeps feature modules loosely coupled. |
| Offline storage | Hive as an offline queue | Fast local storage for records waiting to sync. Today I would also consider an SQL-based store for complex querying and conflict handling. |
| Structure | Feature-first folders | Each role and workflow lives in its own area, so teams can work in parallel. |
What ships first, and what waits
Version one
- Offline-capable survey form for one role
- Local queue and manual or automatic sync
- GPS check-in
- One interface language plus right-to-left readiness
- Basic supervisor view of submissions
Later
- Back-check role and re-verification workflow
- Audio and video capture
- Additional languages
- Dynamic form builder for non-developers
- Quality analytics
Build order and timeline
- 2-3 weeksField research and data modelRoles, forms, sync rules and device list agreed with the people who will use the app.
- 4-6 weeksCore app and offline queueForms, local storage, sync and a test server.
- 3-5 weeksRoles, GPS and mediaRole workflows, check-in and capture.
- 2-4 weeksLanguages and field pilotTranslations, RTL review and a pilot on real devices in the field.
Want a number for your own scope? The app cost calculator turns platforms and features into an estimate.
Rules you cannot skip
- Location permissions
Request the narrowest location access that works. Google Play allows background location only for core functionality and requires a prominent in-app disclosure plus a Play Console declaration; attendance taken at check-in does not need background access.
Official source - Consent and purpose for data
Apple expects consent before collecting user data and clear purpose strings, and says location use must be relevant to the app's features.
Official source - Data of the people surveyed
If respondents are in the EU, GDPR principles apply, including purpose limitation, data minimisation and storage limitation. Collect only the fields the study needs.
Official source
Where these projects go wrong
- Sync conflicts and duplicate records
The same record edited on two devices, or uploaded twice after a drop in coverage, ruins data quality. Design record IDs and conflict rules before writing screens.
- Test only in the office
Offline behaviour, slow networks and old phones only show up in the field. Pilot on the real devices.
- Right-to-left added at the end
Layouts, icons and forms that were never mirrored are expensive to fix late. Build with it from the first screen.
- Forms that cannot change
If every questionnaire change needs a new app release, the project stalls. Drive forms from configuration.
- Fake or careless data
Without back-checks and GPS evidence, quality is a guess. Build the verification role early.
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
emit was called after an event handler completed normally.Flutter Bloc throws that emit was called after the event handler completedBlocProvider.of() called with a context that does not contain a CounterBloc.flutter_bloc cannot find my Bloc: BlocProvider.of() called with a context that does not contain itCannot emit new states after calling closeBloc or Cubit throws StateError after it was closedThe method 'play' isn't defined for the type 'AudioCache'.AudioCache has no play method after updating audioplayersThe URI 'package:flutter_lints/flutter.yaml' included in '/path/analysis_options.yaml' can't be found when analyzing '/path'.analysis_options.yaml: the URI it includes can't be foundThe body might complete normally, causing 'null' to be returned, but the return type, 'int', is a potentially non-nullable type.Dart error: the body might complete normally and return null for a non-nullable type