National Census

First Contact with the App

Opening the National Census app presents a login screen immediately. No tutorial, no walkthrough, no guest mode. The interface is sparse, with fields for a username and password. Below the login fields, a small link offers password recovery. The design is functional but not inviting. After logging in, the main dashboard appears. It shows a map with pins indicating survey locations. A menu bar at the bottom contains tabs for "Surveys", "Data", and "Profile". The "Surveys" tab lists assigned tasks with deadlines. Each task shows the number of households to visit and a status indicator. The "Data" tab displays submitted entries with timestamps and GPS coordinates. The "Profile" tab contains user information and logout option. The app requests location permission on first launch, which is necessary for tracking. It also asks for storage permission to save offline data. The interface language is English, but the app is clearly designed for use in developing regions. The onboarding process is minimal, which may be intentional for low-bandwidth environments. Users who are not technically proficient might find the lack of guidance frustrating. The app does not offer any in-app help or FAQ section. The only way to learn the workflow is through trial and error or external training.

The Two-Step Data Flow

The app combines satellite imagery analysis with real-time enumerator tracking to create a comprehensive data collection system. The first step happens before the enumerator even leaves the office. Machine learning algorithms process satellite images to identify potential survey targets. These targets are then assigned to enumerators as tasks on their dashboard. The second step occurs in the field. The enumerator travels to the assigned location, and the app tracks their movement in real-time. GPS coordinates are logged at regular intervals. When the enumerator arrives at the target location, they begin the survey. The app uses speech and image analysis to verify the quality of the data being collected. For example, if the enumerator takes a photo of a building, the app analyzes the image to confirm it matches the satellite imagery. If the speech recording contains certain keywords, the app flags it for review. This two-step process creates a feedback loop. The satellite data informs the field work, and the field data validates the satellite analysis. The system is designed to reduce errors and ensure data integrity. A practical limitation emerges when the satellite imagery is outdated or inaccurate. If the ML model misidentifies a target, the enumerator wastes time traveling to the wrong location. The app does not provide a way to manually correct the target location in the field. The enumerator must mark the task as "not found" and move on. This limitation can lead to incomplete data collection and frustration for field workers.

Performance in the Field

The app is designed for use in areas with unreliable internet connectivity. It supports offline data collection, with entries stored locally and synced when a connection is available. The sync process is automatic but can be slow with large datasets. The app does not show a progress bar during sync, leaving the user unsure if the process is working. The GPS tracking works reliably in open areas but struggles in dense urban environments or indoors. The app does not provide any feedback on GPS accuracy, which can lead to location data being recorded with low precision. The speech analysis feature requires a quiet environment to function properly. Background noise can interfere with the analysis, causing false positives or negatives. The image analysis feature works well with clear, well-lit photos but struggles with low-quality images. The app does not provide guidance on how to take acceptable photos. The battery consumption is high due to the constant GPS and camera usage. Field workers report needing to carry power banks to get through a full day of data collection. The app crashes occasionally, especially when switching between the camera and the survey form. These crashes can result in lost data if the user has not saved their progress. The app does not autosave, so users must remember to save frequently. The lack of auto-save is a significant design flaw that can lead to data loss and frustration.

Technical Details About the Offline Storage and Data Sync Mechanism

The app stores survey data locally in a SQLite database. Each survey entry is saved as a record with fields for the enumerator ID, location coordinates, timestamp, and the collected data. The data is encrypted at rest using AES-256. The encryption key is derived from the user's login credentials. When the user logs out, the key is discarded, rendering the data inaccessible. The sync process uses a REST API over HTTPS. The data is compressed using GZIP before transmission to reduce bandwidth usage. The sync endpoint is hardcoded in the app, which means it cannot be changed without an update. The app checks for new tasks on every sync, pulling down any assignments that have been added since the last sync. The sync frequency is configurable in the settings, with options for every 15 minutes, every hour, or manually. The default is every hour, which balances data freshness with battery life. The offline storage capacity is limited by the device's available storage. The app does not provide a way to clear old data or manage storage usage. Users must manually uninstall and reinstall the app to free up space if the database grows too large. The data retention policy is not documented, leaving users unsure how long their data is stored locally or on the server.