Survey software India-ready, down to the last village

Yes. Once the form, area and map are downloaded, surveys can be filled all day without a network. Every section is saved to an encrypted local database and survives the app closing or the phone restarting. Sync happens whenever there is signal, even at night at home. Acceptance test: an 8-hour field session, 100 surveys, and no confirmed save lost across retries.

Geofencing only asks whether a point is inside a circle, and usually blocks the survey if it isn't. Verification stores evidence with every submission (lat/long, accuracy, fix age, mock flag), sorts it into five states — in area, near boundary, outside, weak signal, fake GPS — and adds only ‘in area’ records to the approved count. Because the evidence stays on the record, you can audit it later.

As accurate as the boundary data behind it. Districts, assembly and Lok Sabha seats use open polygons; villages and wards use an estimated centroid and radius. That is why the badge says ‘approximately in area’, why only clearly distant records (say 14 km) get flagged ‘outside area’, and why edge cases go to review as ‘near boundary’. Android accuracy is an estimate, not proof, and we show it that way.

XLSForm is an Excel-based standard: survey, choices and settings sheets hold questions, options, skip logic (relevant), validation (constraint) and multilingual labels. Kobo, ODK and SurveyCTO all run on it. Bharat Survey imports XLSForms directly and lists anything unsupported line by line in the import report. You can also export your form back to XLSForm.

Yes. Questions, options, error messages and instructions can be in Hindi and English, with a one-tap language toggle in the app. XLSForm label::hi and label::en columns are supported, and columns for other Indian languages can be added.

Tracking runs only during a shift, after the volunteer taps ‘Start shift’, with a persistent notification — roughly every 30–60 seconds while moving, less often when still. It stops at ‘End shift’. No background tracking (Play Store policy). Leaders see locations only for shift time, and voters or respondents are never tracked. The battery figure is a test target, measured in the pilot.

No. A seat is needed only by a named person who fills surveys in the field. Owners, Admins, Viewers and supervisors who only review are seat-free. A supervisor who also fills surveys takes a seat. Swapping an old volunteer for a new one is a replacement (no charge); going from 10 to 11 is an upgrade.

Location state (5 categories), interview length against the speed limit, time per question, duplicates (house label + location + answer pattern), contradictory answers, impossible time or distance sequences, repeated patterns and the mock-location flag. The signals add up to a 0–100 risk score; suspicious records go to review and the rest are approved automatically. Supervisors can send records back or assign independent back-checks.

No. Sheets is a reporting copy with a one-way sync (Overview, Area, Quality, Team, Issues). The core records live in our database, and the app and dashboard keep running even if Sheets is down or over quota. Full data exports to CSV, Excel and PDF.

Photo questions capture a public problem (a drain, a road, a hand pump) or a shop or facility from the camera; faces are not the default and never without consent. Photos are compressed on the phone (5 MB → ~200 KB), uploaded with location and time metadata, and shown in reports through signed URLs.

Android 8.0 or newer with 2 GB RAM and about 60 MB free space is enough; the app is native Kotlin/Compose, with no heavy webview. Forms are tested up to 300 questions and 10 photos per survey. For fieldwork: location on ‘High accuracy’, date and time on ‘Automatic’, the app excluded from battery saver, and mock location off. There is no iOS app in v1.

No. Each section autosaves to an encrypted local database as soon as it is complete. Switch the phone back on, open the app, and the same survey continues where it stopped. Records already submitted stay in the ‘safe on phone’ count and reach the server when there is signal — with the same UUID, so no duplicates are created.

Yes. A repeat group creates a roster such as ‘these 6 questions for each household member’, with a maximum. Relevance makes the next question depend on earlier answers, cascading selects narrow district → block → panchayat, and piping carries an earlier answer (like a member’s name) into later question text. Exports put each repeat group on its own sheet.

No. 1 seat = 1 named person = 1 authorised device. To change phones, deactivate the old device in the panel and authorise the new one — the seat stays the same, and it counts as a replacement, not an upgrade. The rule is enforced on the server, so changing app settings cannot get around it.

In v1, data leaves as CSV, Excel, PDF and a one-way Google Sheets copy, and forms move in and out as XLSForm (migration from Kobo, ODK, SurveyCTO). A public API is planned for a later version — we don’t list it as available yet. Every export is recorded in the audit log: who, when and what.

No. Sheets is only a copy of the reports; the core data stays in our database. If Sheets fails, the app, sync, QA queue and dashboard carry on as before, and reports honestly show an older ‘last Sheets sync’ time. When Sheets comes back, the copy refreshes on its own.