Nov 19, 2026
No image
Flutter Rewrite of a Hybrid Tide Prediction App
Completed

Flutter Rewrite of a Hybrid Tide Prediction App

$50,000+
7-12 months
United States
6-9
view project
Service categories
Service Lines
IT Services
Mobile Development
Domain focus
Consumer Products & Services
Other
Frameworks
Cordova
Flutter
Subcategories
IT Services
App Management & Support
Mobile Development
Cross-platform

Challenge

A tide prediction application serving coastal users had been built on a web container framework and could no longer be maintained economically. Two separate problems had to be solved at once.

The first was platform drift. Minimum and target platform versions, privacy manifests, data safety declarations, permission justifications and signing requirements all change on a schedule the container plugin ecosystem does not follow. The advertising SDK integration had fallen behind policy and was failing review, and the web view rendering path was producing layout defects on newer device generations including blank regions, mis-sized panes and inconsistent safe area handling.

The second was correctness. Tide heights are computed from harmonic constituents published for each station rather than fetched as a forecast, so the calculation is deterministic and runs on the device. A deterministic calculation does not fail, it simply returns a wrong number when the frame of reference is wrong. Results are only meaningful against a stated vertical datum and the station time zone, and assuming device local time equals station time produces confidently wrong high water times with no error in the logs.

Solution

The application was rebuilt in Flutter with Dart, which renders its own widget tree and removes the class of layout defect caused by web view differences, with consistent behaviour across current device generations from a single codebase.

A model view controller separation with a state management library such as GetX keeps the prediction logic, station data access and presentation apart, so the calculation layer is testable without a widget involved.

The structure covers a station data layer holding bundled or synced constituent and metadata sets with a versioning scheme so station corrections ship as data rather than requiring a full release, a prediction engine in pure Dart that is deterministic and unit tested against known reference values including astronomical calculations, a selection flow built as region, area, zone and site with saved favourites, offline behaviour treated as a requirement because the shoreline is where connectivity fails, and a correctness surface with explicit vertical datum labelling and station time zone handling including daylight saving transitions.

The prediction and data layers were ported first and validated against the previous implementation across a broad sample of stations and dates before the interface was rebuilt screen by screen against existing behaviour.

Results

The application renders consistently on current device generations, so the layout defects caused by web view differences stop recurring with each platform update.

Predicted heights and times remain faithful to the previous implementation, which is what made the rebuild safe to ship to an existing user base that had been reading those numbers for years.

Station corrections now reach users through data updates instead of store releases. The build satisfies current store, privacy and advertising policy requirements, and the deterministic calculation layer is testable independently of the interface. Source history was retained through the migration so prior revisions remain traceable.