Set up the project, bring in the team, work offline in the field, then get live QA and reports.
Bharat Survey is an end to end digital survey platform for teams running door-to-door and field surveys, and this is the order the work actually follows, from buying a plan to a panchayat-wise report. It comes in three parts: five setup steps in the client panel, ten field steps for volunteers, and the server’s checks through to the dashboard. At every stage you can see how far the work has got, and every record carries its evidence.
The surveyor introduces themselves, asks for consent and fills the form in 8–12 minutes. Everything before it (setup) and after it (checks) has one job: making sure this moment happens in the right place, with the right person, on the right form, and that the proof survives.
Project → areas → form → seats → assignments
The Owner or an Admin does this from a laptop or mobile browser. With area data already loaded, the longest task is settling the form, and if you already have an XLSForm, even that takes minutes.
A project is one assembly seat, district, group of wards or research round. Give it a name, a tenant type (political, company, NGO or government), a survey round, a date range and a consent-text version. The tenant type decides which modules appear: booth management and the war room show up only for political clients.
India’s administrative map is already loaded: state → district → block → gram panchayat → village, or urban body → ward, or assembly seat → booth. You tick boxes. Each area shows its boundary status (verified, approximate or not available), and a missing-mapping report arrives the same day. Hamlets missing from the official list get a local ID.
Start from a library template (voter mood, public services, household, retail audit, baseline) or upload an XLSForm from Kobo, ODK or SurveyCTO. You get skip logic, validation, Hindi and English labels and a “don’t know” option. Sensitive questions go live only after our approval; a published form is locked, and any change becomes a new version.
Add volunteers by name and mobile number; they activate through an invite link or OTP, and the device is registered. Create supervisors and hand them a team and an area. One seat is one person who fills surveys; Viewers and Owners never use a seat. If every seat is taken, one button sends a request for more, and the limit rises once we approve and payment clears.
An assignment ties together a volunteer, an area, a form, a date range and a target, say 300 households. It moves through Assigned → Accepted → In Progress → Paused/Completed → Closed. The volunteer gets a push notification and finds it under “Today’s work” after the next sync. Political projects can assign by booth. This is also where the dashboard’s target-versus-done figure comes from.
The ten screens of a volunteer’s day — hover a card or use the arrows
After setup, everything happens on these screens. Hover over the cards to fan them out; the arrows bring the next screen to the centre. Below, the same sequence, step by step.
Hello, I am Ram Kumar. This survey is run by Sample Organisation. Purpose: to learn about amenities in your area. It is voluntary, and you can say no at any time.
Offline, Hindi-first, with evidence at every step
These are the ten steps taught in the 45-minute training. The app shows the screens in the same order, so volunteers don’t have to memorise anything; they just press the next button.
Open the app and it pulls down today’s assignments, forms, instructions, household list and map tiles. The status is plain: “3 of 3 assignments ready”. From here the volunteer can work all day without a signal.
Tapping “Start shift” shows a clear notice about location sharing. If the volunteer agrees, they appear on the supervisor’s live map. A persistent notification stays on screen, so nothing runs in secret. The time left on the Offline Work Permit is shown here too.
One button on arrival records a foreground GPS fix with its time and accuracy. The app says either “probably inside your area” or “2.4 km outside the boundary — are you in the right place?”. That fix is the day’s first piece of evidence.
The surveyor says who is running the survey, why, and that taking part is voluntary. The time of consent and the text version are saved. “Declined” is a button too; it counts as data, and reports show the refusal rate separately. Consent under the DPDP Act 2023 starts here.
Hello, I am Ram Kumar. This survey is run by Sample Organisation. Purpose: to learn about amenities in your area. It is voluntary, and you can say no at any time.
One question per screen, large options, the next question chosen by skip logic, error messages in Hindi and a “don’t know” option. Time per question is logged quietly in the background, useful for both QA and training. Switching language takes one tap.
Complete, partial, declined, no contact, house locked, revisit: every household gets an outcome whether or not a survey happened. That is how the report gets an honest denominator: 300 households, 240 complete, 22 declined, 38 locked. Revisits drop into the next day’s list on their own.
Photos of a public problem, a shop or a facility, never of faces. The phone shrinks each one (5 MB → about 200 KB), keeps location and time metadata, and uploads it later through a signed URL. In retail audits and facility surveys, this is the evidence.
Each finished section is saved to an encrypted local database. If the app closes, the phone restarts or the battery dies, the volunteer picks up exactly where they stopped. Every submission gets a permanent UUID on the phone. The “safe on phone” count goes up; “on server” does not yet, and the app says so openly.
2G, 3G or WiFi, whatever turns up, sync runs in the background: answers first, then photos. Failed attempts retry at growing intervals. A record is marked “reached server” only after the server confirms it. Duplicates can’t happen, because the same UUID means the same record. On a normal connection the dashboard updates in about 5 seconds.
“End shift” turns location sharing off, shows the day’s summary (complete, declined, locked, revisit) and warns about anything still waiting to sync. Corrections sent back by the supervisor appear on the Pending/Corrections screen with reasons. Tomorrow’s work downloads overnight, so the morning starts in the field.
Hello, I am Ram Kumar. This survey is run by Sample Organisation. Purpose: to learn about amenities in your area. It is voluntary, and you can say no at any time.
Location state → QA queue → alerts → live dashboard → Sheets
What happens after a volunteer taps Submit isn’t done by hand. The server checks, the supervisor sees only the suspect records, and leaders and managers get accepted data live.
The server checks the evidence (lat/long, accuracy, age of the fix, mock-location signals) against the assigned boundary and sorts each record into one of five states: inside area · near boundary · outside area · weak signal · suspected fake GPS. Only “inside area” moves toward the accepted count; everything else goes to the queue or raises an alert.
Speed limits, time per question, duplicates, contradictions, impossible sequences and copy-paste patterns each add weight to a risk score. Suspect records line up in score order. The supervisor accepts, holds or rejects, sends a correction back with a reason, or assigns an independent back-check. Identical answers alone are never treated as fraud.
Outside area, fake GPS, interviews that are too fast, inactive for 45 minutes, sync pending, permit about to expire, booth not reporting: each lands in one list with a priority and an owner. Closing an alert requires a reason. Repeats of the same cause fold into one thread, so you see signal, not noise.
A live map (every volunteer on shift, approximate boundaries), an activity timeline, progress against target and four reports: work progress, data quality, local issues and help requests. Filter by area, date, volunteer or form version. Every report states its denominator, time window and form version. Export to CSV, Excel or PDF.
Five tabs (Overview, Area, Quality, Team, Issues) fill your Sheet automatically. It is a reporting copy: editing the Sheet never changes the source data, and if Sheets is down or over quota, the app and dashboard keep running. Two timestamps are shown: “last app server update” and “last Sheets sync”.
Demo → setup → training → pilot day → roll-out → first clean-up → weekly review
A planning estimate, not a commitment, but every project follows this order. Keep scrolling and the week moves forward.
We load your area and walk you through it: the panchayat or ward list, a sample form, the app and the live dashboard. This is where the tenant type, number of seats and form source (library, XLSForm or new) get decided.
Project, area selection, form (with approval), seats, supervisors, assignments, dashboard. Booth master import for political work and household list import happen here too.
Volunteers: app screens, consent script, the form, sync, corrections. Supervisors: 60 minutes on the QA queue, back-checks and alerts. Training videos live in the app.
Five to ten volunteers in one panchayat or ward. We sit with you on the live dashboard, and you run AT-04 (outside area) yourself. In the evening the form and thresholds get adjusted.
Every seat active, every assignment live. Daily: alert list, progress against target, data quality. Supervisors clear the QA queue every day.
Small fixes based on the first two or three days: a volunteer who hasn’t started a shift in 3 days hands the seat to someone new as a replacement; the first 5% back-checks go out; the review policy for “near boundary” records gets settled.
Panchayat- or ward-wise report (Excel, PDF, Sheets), lagging areas, QA rate, back-check match, seat needs. The form version is locked for the next round.
Super Admin · Owner/Leader · Supervisor · Volunteer · Viewer · Booth roles
Only the people who fill surveys in the field use a seat. Everyone else is free, so you never pay for leaders, managers or the people reading reports.
Clients, projects, plans and seats, payment receipts, approving or declining extra-seat requests, geography imports, form approval, QA policy, staff permissions, backup status. Every sensitive access is logged with a reason.
Team, seats, assignments, live map, alerts, reports, exports, seat requests; booth management and war room for political work. A Client Admin has the same access minus billing.
Their own team and area: the QA queue, sending corrections, exceptions with reasons, assigning back-checks, daily booth reports. A seat applies only if they fill surveys themselves.
Today’s work, shift, arrival check, consent, form, visit outcome, sync, corrections. One volunteer, one authorized device. Nobody else’s data is visible.
Read-only dashboards and reports for party office-bearers, research clients or department officials. The Owner decides whether exports are allowed. Add as many Viewers as you like; seats don’t change.
A daily booth report (2 minutes) and polling-day turnout entry. They don’t fill surveys, so no seat applies. Panna Pramukhs get roll page ranges assigned; voter list import is not part of v1.
45 minutes, in Hindi, phone in hand
Training happens with a phone in hand, not a slide deck. The bars show how long each part runs. The full form practice and the consent script get the most time, because those are what go wrong most in the field.
Supervisors get a separate 60-minute session: QA queue (20), sending corrections and exceptions (15), assigning back-checks (10), alerts and the daily routine (15).
Common mistakes in mobile field surveys, and the habit that prevents each
None of these is an app fault; each one is a habit. Name them out loud in training and in the first week’s briefings.
The volunteer then never appears on the live map and looks inactive to the supervisor. The app reminds them before the first survey; make it the first line of the briefing too.
Logging only completed surveys can’t give you “240 of 300”. House locked, declined, no contact: every door gets an outcome, even if it takes ten seconds.
“6 on phone” means six records haven’t reached the server. Open the app on home WiFi and let sync finish, or the permit may have expired by morning.
300 households in three days isn’t a target; it’s an invitation to fake surveys. Set assignment dates around 25–40 complete surveys a day.
Too strict and it flags honest fast volunteers; too loose and it flags nobody. Take the average length from the pilot day and set about 40% of it.
Write “Q7 photo unclear, retake it” or “Q4 and Q9 contradict, please confirm”. The reason is the training, and it’s what the audit log shows.
Back-checks only work when volunteers don’t know which 5% of households will be revisited. Give the assignment to someone independent.
Records sitting “in review” don’t count as accepted, so reports look behind. Supervisors should clear the queue top to bottom in 20 minutes a day.
Without a process vs with Bharat Survey
Data leaks at every link
Evidence at every step, a status for every record
Choosing the form source: template library vs XLSForm import vs building from scratch
This takes up most of the 1–2 setup days. All three routes end in the same form engine; the difference is time and preparation.
| Question | Library template | XLSForm import | New in the builder |
|---|---|---|---|
| Best when | Common needs: voter mood, public services, household, retail audit, baseline | You already have a working Kobo/ODK/SurveyCTO form | The questionnaire is new or still on paper |
| Time needed | ~1 hour (adding or trimming questions) | Minutes to import, plus a check of the import report | 1–3 hours |
| Skip logic / validation | Already in the template | Comes from the file (relevant, constraint) | Build it in the builder, test it on a mock household |
| Hindi / English labels | From label::hi / label::en columns | ||
| If there are sensitive questions | Published after approval | Published after approval | Published after approval |
| Risk | The template may not quite match your context | Unsupported items are listed line by line in the import report — read it | Logic mistakes the first time — that’s what the pilot day is for |
| Getting it back out | XLSForm export | XLSForm export | XLSForm export |
Think of a mobile field survey in three parts: before the field, in the field and after the field. Before the field, the most important job is testing the form: a “don’t know” option on every question, skip logic checked by filling the whole form for a mock household, and a Hindi consent script a volunteer can say in 20 seconds. Then look at the boundary status of your areas. Villages with approximate boundaries will produce more “near boundary” records, so warn supervisors in advance.
The volunteer’s ten field steps are above. Three things get missed most: (1) forgetting to tap Start shift, so nobody shows on the live map, which is why the app reminds volunteers before the first survey; (2) recording the visit outcome at every house, even without a survey, because that is what makes “240 of 300” possible; (3) checking the sync count in the evening. “6 on phone, 8 on server” means six haven’t arrived yet; open the app on home WiFi and let it finish.
After the field, the supervisor’s 20 minutes a day go on the QA queue from the top (sorted by risk score), closing high-priority alerts with reasons, and assigning the next day’s revisits. The leader or manager’s five minutes go on progress against target and the ratio of accepted records to those still in review.
One supervisor can handle the QA queue, alerts and revisits for 8–10 volunteers a day. Beyond that the queue piles up, and records sitting in review slow the report down.
With an 8–12 minute form, walking time and declined or locked houses, a volunteer completes 25–40 door-to-door surveys a day, depending on how dense the area is. A 300-household assignment is 8–12 days of work, not 3. An impossible target is an incentive to fake surveys.
Look at the pilot-day numbers: how long does an honest survey take on average? Set the speed limit at about 40% of that. Start the inactivity threshold at 45 minutes and adjust it for urban and rural areas.
For 5% of each volunteer’s households, a second person refills a short form. A match above 90% is normal; below 70% signals a need for retraining or investigation. Don’t tell volunteers the back-check list in advance.
Not “wrong”, but “Q7 photo unclear, retake it” or “Q4 and Q9 contradict, please confirm”. The reason is the training, and it’s what the audit log shows.
If a volunteer hasn’t started a shift in 3 days, move the seat to someone new as a replacement; there’s no charge. Add seats only when every existing seat is active every day.
“62% report a water problem.” 62% of what: complete surveys, accepted surveys or valid answers? Reports state all three; read them in the review.
This rule exists so that offline-first work and seat control can both be honest at the same time. Every successful sync gives the device a 24-hour permit, adjustable per project. Within the permit, a volunteer can work fully without network. When it expires, new surveys can’t start, but saved data stays safe and sync is still allowed, so not a single record is lost. The app warns four hours ahead and an alert appears in the panel.
When you close a volunteer’s seat or deactivate a device, an online phone stops immediately. A fully offline phone stops when its permit expires or when it next connects to the server, whichever comes first. We don’t promise that closing a seat shuts down an offline phone the same instant, because a phone that isn’t connected can’t receive any message. Every survey app has this limit; the difference is that we write it down and tie it to a permit.
The supervisor is the hinge of this whole workflow. Their day splits into four short blocks and adds up to less than an hour on the panel. Morning (10 minutes): a briefing in the village square covering yesterday’s returned records and reasons, today’s target, the form version, and whose phone didn’t sync last night. Around 9:30 (5 minutes): check the live map to see every shift is on, and call anyone who isn’t showing. Midday (10 minutes): the alert list, with “outside area” and “fake GPS” first, then “inactive” and “sync pending”. Evening (20 minutes): the QA queue in risk-score order, with one of three decisions on each record (accept, hold and send back, or reject) and a reason. Once a week, review back-check assignments and match rates.
Two habits set good supervisors apart. First, they grant few exceptions, and write each reason so a third person could understand it three months later (“the village’s eastern hamlet falls outside the radius; confirmed on site”). Second, they bring the “in review” count close to zero every day, because while a record sits in the queue, it doesn’t appear in the leader’s accepted count.
The pilot day isn’t a formality. It’s the first day you see real numbers for your form and your area, and it settles four things. Speed limit: work out the average length of an honest survey and set the limit at about 40% of it. Inactivity threshold: start at 45 minutes and raise it in scattered villages where walking between houses takes time. Daily target: the number of complete surveys a volunteer managed in the pilot is the realistic baseline, usually 25–40. Form flaws: check time per question; reword the one everyone gets stuck on, and ask whether the one nobody pauses on for more than two seconds is being read at all. After these changes a new form version is published, and the roll-out runs on it.
When this process is more than you need — or not what you need
Not every survey needs seats, assignments and a QA queue. In these situations we say so plainly during the demo.
The questions about setup, training, field work and supervision that come up in every demo.
Log in with mobile number and OTP → download today’s work → reach the area and tap “Start shift” → arrival check → introduction and consent at each door → the form, one question per screen → visit outcome → automatic sync when there’s network → “End shift” in the evening. The whole day can run offline, and the “safe on phone” and “reached server” counts are always visible.
Volunteers get a 45-minute session in Hindi covering the 5–6 app screens, consent, the form, sync and corrections, plus live support on day one. Supervisors get a separate 60-minute session on the review queue, sending corrections, back-checks and alerts. Training videos stay inside the app’s Help screen.
From the client panel: add volunteers within your seats, make supervisors and give them a team and area, assign work by area, watch the live map and activity timeline, act on alerts, and use the weekly report to spot areas that are falling behind. Every action is kept in the audit history.
Because area data is preloaded, one assembly seat, district or project usually goes live in 1–2 working days: project, areas, form (or XLSForm import), seats, assignments, dashboard. Larger projects with several districts, a household list import or a booth master take about a week. These are planning estimates, not commitments.
Every successful sync gives the device a 24-hour offline permit, adjustable per project. When it runs out, new surveys can’t start, but saved data stays safe and sync is still allowed. If you close a seat, an online phone stops at once; a fully offline phone stops when the permit ends or at its next connection. We don’t promise an instant shutdown we can’t deliver.
Deactivate the old device in the panel and authorize the new one. The seat stays the same; it’s a replacement, not an upgrade. An online phone stops at once, a fully offline one when its permit ends or it next connects. The phone’s local database is encrypted, so a lost phone can’t be read, and anything already synced is safe on the server.
Download today’s work in the morning on a home or market-square network, and the forms, areas and map tiles all land on the phone. After that, fill surveys all day with no signal; GPS fixes work without a network. Sync happens by itself wherever a signal turns up in the evening, even at home at night.
In the QA queue the supervisor picks “Send back” and writes a reason, such as “photo unclear” or “Q7 blank”. The volunteer gets a push notification and sees the record, with the reason, on the Pending/Corrections screen, fixes it and resubmits. History is kept version by version; original answers are never overwritten.
Three things, in five minutes: (1) the alert list, high priority first; (2) progress against target, to see which area is lagging; (3) data quality, the share of accepted records against those in review or outside the area. The app’s Leader view (Overview → Map → Activity → Alerts) shows the same thing, even from a car.
Yes. Projects, areas, forms, seats, assignments, the 10 field steps, QA and reports are shared by everyone. The difference is in modules: booth management and the war room are only for political clients; household rosters, repeat groups and beneficiary tracking appear in NGO and government projects; retail audits add a shop list and photo evidence.
Five to ten volunteers work a full day in one panchayat or ward while our team sits with you on the live dashboard. Morning: download and shift-start checks. Afternoon: sync counts. Evening: the first pass through the QA queue. You run test AT-04 yourself: have someone fill a survey from outside the area and watch it get flagged, stay out of the accepted count and raise an alert. By evening you agree changes to the form, speed limit and inactivity threshold.
With an 8–12 minute form, walking time between houses and the declined or locked doors, usually 25–40 complete surveys a day, depending on how dense the area is. So a 300-household assignment is 8–12 days of work. That is a planning estimate; the real number for your form and area comes from the pilot day, and targets should be set from it.
A workable ratio is one supervisor for 8–10 volunteers. At that size they can clear the QA queue daily, close high-priority alerts and assign the next day’s revisits. A supervisor who only reviews doesn’t use a seat; one who also fills surveys does.
No. The Owner or a supervisor adds the volunteer’s name and mobile number in the panel; the volunteer receives an invite link or OTP and the app activates on that phone. One volunteer means one authorized device. The first ten minutes of training go on this activation and the first download of today’s work.
The volunteer records “house locked”, “no contact” or “revisit” for that door, and that is a record too. Revisit households come back on the next day’s list automatically. Reports count each of these separately, which is how you get an honest line such as “240 of 300 complete, 22 declined, 38 locked”.
A published form stays locked. Changes become a new version, published after approval if sensitive questions are involved. Phones pick up the new version at their next sync, and any survey in progress finishes on the old one. Reports filter by form version, so old and new answers never mix. Check the version number on everyone’s phone at the morning briefing.
In a 20-minute demo we set up your assembly seat, ward or project in front of you, then sit with you through the pilot day.
Open reference pages built from ECI, Census 2011, LGD and MCA records, with the source credited on every page.
