Field survey app for market research: CAPI interviews and store audits with GPS and photo proof.
Bharat Survey is a market research company in India that runs consumer surveys, retail shelf audits, distributor visits and mystery-shopping-style outlet checks, and licenses the same offline-first Android app to agencies with their own field teams. Every visit is GPS-checked and photographed. Import your XLSForm questionnaire, watch quotas live, clear a risk-scored QA queue and export to CSV or Excel.
Most market research field data is unverifiable. Yours will not be.
The brand manager reads the report, the agency sends the tables, and in between are 40 interviewers nobody watched at work. Most survey apps only geo-tag a record. They can’t tell you whether the interview really happened at that shop or in that home. Here are six things that go wrong on almost every project, and how Bharat Survey catches each one.
Keep scrolling and the audit moves forward. The example is a sample (outlet OUT-0412, Lucknow), but the sequence is the same on any FMCG, telecom or pharma project.
A CSV with name, address, channel (kirana, modern trade, chemist) and coordinates goes in. Each outlet gets a permanent ID, such as OUT-0412, that stays the same in every wave. Rows without coordinates are listed separately so you can fix them.
Each auditor receives today's outlets and a daily target. Lists and forms download on the office network, so nobody needs signal in a narrow bazaar lane or a mall basement.
One tap on arrival. The app records location, time, accuracy and fix age, then matches them against the outlet's GPS pin. If the shop is shut, that outcome is recorded instead and does not count toward quota.
Yes/no for every focus SKU in a repeat group, then facing count, stock-outs, MRP and promo price. Each question can be answered in ten seconds by looking at the shelf, with limits on numbers and error messages the auditor understands.
Photos come only from the in-app camera and attach to the same record with time and location. They shrink on the phone from 5 MB to roughly 200 KB. Anything picked from the gallery, or taken long before submission, goes to review.
Completed, shop closed, owner refused or outlet not found: one of four. The record sits in the phone's encrypted database and reaches the server with a unique ID once signal returns, so duplicates don't appear.
The server checks duration, per-question timing, location state, duplicate outlets and photo time. Clean records are accepted automatically; the rest line up by risk score. The back-check form stays linked to its parent record, and corrections go back with a reason.
The dashboard shows city × channel progress and quota gaps. CSV/Excel carries question codes, labels, form version and verification-state columns; a PDF summary goes to the client. Clients can also watch live with a Viewer login.
India geography, verified visits and seat pricing in INR — built in.
Location, time, accuracy, fix age and mock-GPS signals are matched to the outlet or area and resolved into five states. An ‘out of area’ interview never reaches the accepted table, and client reports can show verification state beside every figure. When a boundary is approximate, the badge says ‘approximately inside area’. It never shows a false ‘Verified’.
Districts, blocks, gram panchayats, villages, urban local bodies and wards for all 36 states and UTs, with LGD codes. Assign by city, report by ward, upload an outlet list and see it on the map, with no areas drawn by hand. A rural haat audit and a metro ward consumer survey can run in the same project.
As many seats as active interviewers, for as many months as the project runs. Viewers and project managers don’t use seats. No monthly dollar bill, and no high floor sold as ‘unlimited users’. When an interviewer leaves, the seat moves to a new named person, and the server checks the limit at every activation.
Consumer interviews, retail audits, quotas, QA and the agency panel. Switch tabs to see what a project manager sees.
From outlet list to analyst-ready tables — with proof at every step.
Retail audit app with GPS and photo: shelf share, stock-out, POSM, planogram
A retail audit is only as credible as two facts: the auditor really went to the shop, and the shelf photo was taken then. Bharat Survey makes both part of the record. Upload the outlet list as CSV (name, address, channel such as kirana, modern trade or chemist, and coordinates). Each outlet gets a permanent ID and appears on the auditor’s ‘today’ map.
Form engine with skip logic, validation, piping and XLSForm import
Single and multiple choice, number, text, date, rating, photo, signature, barcode, repeat groups (a household roster or brand list), cascading selects, calculations and piping. Validation comes with Hindi error messages so the interviewer knows what to fix. Published forms are locked; a change creates a new version, and old answers keep the meaning of the old question, so a wave-on-wave brand track stays intact.
Field force survey app: visit plan vs actual, with duration and location state
Turn a sales or merchandising visit into a short form: arrival check, a 6–8 question checklist, one photo and a visit outcome. Reports list each person’s visits, duration, location state and missed outlets. This is a survey platform, not order booking or invoicing, but it is where the proof that a visit happened comes from.
Offline-first with honest sync status on 2–3 GB RAM Android phones
Forms, outlet lists and instructions download on the office network in the morning. In the field, each section is saved to the phone’s encrypted local database, and it survives the app closing or the phone restarting. Records sync on their own once signal returns, with unique IDs so nothing duplicates. The interviewer can always see how many records are on the phone and how many reached the server.
Consumer survey, shelf audit, distributor visit, outlet check, brand track, CSAT
FMCG, telecom, BFSI, pharma, durables and research agencies: one platform, set up per project. Forms, assignments, QA and exports work the same in every case; only the questionnaire and sampling frame change. Hover over or tap a panel.
Multi-client workspaces for research agencies: one panel, isolated tenants, seat-free client viewers
The four cards below stack as you scroll, the same way access layers build from the agency panel down to a client’s report. Figures on screen are sample data.
The agency admin sees every client workspace, its seats and status in a single list. The form library sits at agency level, so last month's retail audit checklist can be copied and adapted for the next client without touching the previous client's data.
Interviewers, outlet lists, forms, photos and reports for FMCG client A and telecom client B live in separate tenant spaces. The separation isn't only on screen: it is checked at the API, database, file storage, export and live-update layers.
Give the brand manager or research head a read-only Viewer login. They see live progress, quota gaps and verification states; contact details and editing stay closed. A login rather than a link, so every access is on record. Viewers don't use a seat.
When the project ends, hand the client a full export: CSV/Excel/PDF plus photos through signed URLs. Data then stays read-only for later export; it is deleted when the client instructs or on a valid DPDP Act 2023 request, and that is written to the audit log.
Shop, shelf, phone and dashboard: each tile below is a moment on the same route. Screen tiles show sample data; photos are illustrations, not customers.
SurveyCTO/Kobo-style tools are good form engines. India needs more than a form engine.
SurveyCTO, KoboToolbox and ODK are excellent XLSForm engines, which is exactly why we import XLSForm. The difference is in everything an Indian field project otherwise has to bolt on by hand, every time.
Such as SurveyCTO / KoboToolbox / ODK
The same XLSForm, with the India layer included
CAPI vs CATI vs paper: what each method can and cannot prove
| Criterion | CAPI (phone/tablet, face to face) | CATI (phone call) | Paper |
|---|---|---|---|
| Location proof of the visit | Yes: five states, matched to outlet/area | No: respondent’s location unknown | No |
| Photos / shelf observation | Yes: in-app camera, time-stamped | No | Separate camera, linked to records by hand |
| Skip logic and validation followed | Enforced by the form | Depends on the software | Depends on the interviewer’s memory |
| Data-entry cost and errors | None: answers go straight to the database | None | Double entry, typos, delays |
| Practical interview length | Long questionnaires, show cards possible | Calls drop after 8–10 minutes | Long is possible, but skips get missed when tired |
| Who you can reach | Any home or shop the team can visit | Only people whose number you hold | Any home or shop the team can visit |
| Retail audit / outlet check | Built for it | Not possible | Possible, but without proof |
| In Bharat Survey v1 | Help migrating off paper |
Retail audit checklist, CAPI vs CATI vs paper, GPS+photo proof, XLSForm import
These are the four questions that come up in every first demo. The answers are product-neutral: they hold whichever tool you pick.
A good retail audit checklist is short, and each question can be answered in ten seconds by looking at the shelf. Start with outlet identity: ID, channel (kirana, modern trade, chemist, paan-plus) and the GPS pin from the list. Then availability: yes/no for each focus SKU, inside a repeat group, so adding an SKU doesn’t mean rebuilding the form. After that come facing count and share of shelf, stock-outs, MRP and promotions, POSM presence and one mandatory photo.
Three mistakes show up in almost every first checklist: (1) too many open-text questions, because an auditor won’t type at a counter; (2) optional photos, after which 30% of records arrive without one; (3) audits without outlet IDs, which makes the second wave impossible to compare. In Bharat Survey, mandatory photos, mandatory outlet IDs and arrival checks are all form settings, set per project.
For monthly audits, keep a fixed outlet panel so trends read cleanly; rotate 10–15% of the panel each quarter to drop shops that have closed. Set a quota for each city × channel and count it on accepted visits only. ‘Shop closed’ and ‘out of area’ visits should never add to quota.
CAPI (Computer-Assisted Personal Interview) has the interviewer fill a phone or tablet in front of the respondent. For long questionnaires, show cards, photos, location proof and observation of a home or shop, nothing replaces it. It costs the most, but data entry is zero and verification is possible.
CATI (Computer-Assisted Telephone Interview) is fast and cheap, but it reaches only people whose numbers you hold, interviews longer than 8–10 minutes tend to drop, and you get no location, photos or observation. A retail audit can’t be done by CATI at all.
Paper looks cheapest at the start. Then come data entry, keying errors, skip logic nobody followed and no record of when or where the survey happened. A simple rule: if you need proof of the visit, or the questionnaire has more than five skips, use CAPI. Bharat Survey is built for CAPI; CATI is outside v1.
A geo-tag sticks a location point onto a record. It doesn’t tell you whether that point was near the outlet, fresh, or produced by a fake-GPS app. Verification matches location, time, accuracy, fix age and mock-location signals against the outlet or area and gives one state (inside area, near boundary, out of area, weak signal, suspected mock GPS), with a record of every exception.
Accuracy is an estimate, not absolute proof, so an honest system won’t show ‘Verified’ when a boundary is approximate; it says ‘approximately inside area’. Photos follow the same rule: in-app camera only, with time and location, and gallery photos or ones taken long before submission go to review. Analysts get the data that came from the spot; everything else is counted separately, with a reason.
XLSForm is an open standard (survey, choices and settings sheets) used by KoboToolbox, ODK and SurveyCTO alike, so your questionnaire isn’t locked into any one tool. On import, Bharat Survey reads question types, relevance (skip logic), constraints and their Hindi messages, calculations, choice lists, cascading selects, repeat groups and label::hi/label::en language columns.
After import you get a line-by-line report: which questions came through unchanged, which were adjusted, and which aren’t supported (such as some external data-source functions). Nothing changes silently. Importing historical data from CSV and attaching outlet or household IDs is part of the setup service, so comparisons with earlier waves don’t break.
Sample CAPI questionnaire structure, sample client report outline, a five-day rollout and the mistakes to avoid
The structures below are samples and will change with your questionnaire and your client’s format. The point is to show what has to be decided between the first demo and the first field day.
A consumer questionnaire usually splits into seven blocks. (1) Screener: age, gender, SEC, category use; if the quota cell is already full, the interview stops here, saving the respondent’s time and your budget. (2) Introduction and consent: who is running it, the purpose, that it’s voluntary; time and version logged automatically. (3) Category use: how often, where from, who buys. (4) Brand funnel: unaided then aided awareness, consideration, last purchase; the brand list sits in a repeat group so every brand gets the same questions. (5) Ad recall and messaging. (6) Price and pack size: constraints on numbers, Hindi error messages. (7) Demographics and close: visit outcome, interviewer’s note.
Set a minimum form length (sample: 9 minutes for a 12-minute questionnaire). An interview that finishes faster isn’t rejected automatically; it goes to review with a risk score. Keep open-ended questions to two at most. An interviewer standing at a door won’t type much, and the one who does often repeats the same sentence everywhere.
A good field report accounts for the fieldwork before it shows results. Page one: period, form version, cities, total submissions, accepted, in review, rejected, and the ‘out of area’ count. Page two: target vs accepted for each quota cell, with the reason any cell fell short. Page three: quality, meaning average interview length, how many back-checks were done and matched, and the split of visit outcomes. Only then the results tables, with n beside every percentage and ‘don’t know’ counted separately.
Bharat Survey delivers that accounting with the export: verification-state and form-version columns in CSV/Excel, and the header in the PDF summary. The app itself is not a cross-tab analysis tool, and SPSS/Stata labelled exports aren’t in v1. Analysis and presentation stay with your team, or can be scoped with us as a separate service.
1. No coordinates in the outlet list. Then ‘near the outlet’ can’t be matched; capture the pin on the first visit and start matching from wave two. 2. Counting quota on total submissions. Once pending and rejected records drop out, cells come up short on the last day; always count accepted. 3. Changing questions mid-field. If you must, create a new version and report response counts for both versions separately. 4. Asking supervisors to read every record. They’ll give up by day three; give them the risk-score queue only. 5. Skipping sync status in training. Most first-day calls are ‘did my data go through?’, so teach people to read both counts on screen.
Not every research project needs a field app. In the five situations below we’ll tell you to choose another method, and we say the same in demos.
Assignments, quotas, exports and client access — one panel.
By role, the work splits like this: the project manager handles setup and quotas, supervisors handle the QA queue and back-checks, interviewers see only ‘today’s work’, and clients get a read-only dashboard.
Each client’s data separated at database level; encrypted in transit and at rest; consent aligned with the DPDP Act 2023.
In-app camera, compressed on the phone, uploaded with the submission, accessed through signed URLs.
Speed, duplicates, location, contradictions: suspicious records first; back-checks linked to the parent record.
Question codes, labels, form version and verification-state columns in CSV/Excel; a PDF summary for the client.
From submission to dashboard on a normal network (SSE realtime), measured during the pilot.
In Hindi, with live support on day one and separate review and back-check training for supervisors.
Behind every table: how many interviews were accepted, how many are in review, how many were out of area. That number can go in the report. An agency that can show the quality of its fieldwork is the one that wins the next project.
Book an agency demoCAPI, quotas, XLSForm, pricing and privacy: the questions every agency demo brings up.
Both. We run survey and field services, including market research fieldwork and data collection teams, and we license the same offline, GPS-verified Android app and web panel to brands and agencies that run their own field. Either way, every interview carries the same location, timing and photo evidence.
Yes, same category: offline CAPI interviews, XLSForm support, skip logic, validation, multilingual labels. The difference is that Bharat Survey ships with Indian geography preloaded (district, block and ward with LGD codes), verifies GPS against five location states instead of only tagging it, prices seats in INR, has a Hindi-first app interface and hosts data in India.
Each outlet has a GPS pin and a permanent ID. The auditor’s submission location, time, accuracy, fix age and mock-location signals are matched against the outlet. Submissions far from the outlet are flagged ‘out of area’ and kept out of the accepted count. Shelf photo time is compared with submission time, so old or gallery photos go to review.
Yes. XLSForms from Kobo, ODK or SurveyCTO import with question types, relevance (skip logic), constraints, calculations, choice lists, repeat groups and label::hi / label::en language columns. Anything unsupported shows up line by line in the import report. Nothing changes silently.
CAPI: the interviewer sits face to face and fills a phone or tablet, which allows long questionnaires, photos and location proof. CATI: phone-call interviews, fast and cheap, but no location or photos, and you only reach people whose numbers you hold. Paper looks cheapest until data entry, errors and verification add up. Retail audits and door-to-door work need CAPI. Bharat Survey is built for CAPI; there is no CATI module in v1.
At setup you enter targets by area × segment, for example city × SEC × age band. Each accepted interview counts in its cell; records in review or rejected do not. When a cell fills, the app won’t start new interviews in it. The quota gap report is on the dashboard and in CSV.
Yes. B2B studies with dealers, distributors, chemists or service centres use the same CAPI forms, outlet IDs and GPS checks, so each interview is tied to the business premises it claims. Screeners confirm the respondent’s role before the main questionnaire starts. Contact details are restricted by role and never shown to client Viewers.
No. Our field services, geography data and data hosting are built for India only: all 36 states and union territories, down to village and ward. If your study spans several countries, we can cover the India leg and export in CSV/Excel so your coordinating agency can merge it with the other markets.
No. We don’t publish syndicated reports or off-the-shelf industry data. Every study is commissioned: your questionnaire, your sample, your data. If you need a category overview, a custom study with a defined sample is the honest route, and you own the results.
CSV, Excel and PDF reports, filtered by area, date, interviewer or form version. A one-way report copy to Google Sheets. Photos export separately with signed URLs. SPSS/Stata labelled exports and an API are planned for a later version, not v1.
Active interviewer seats per project plus plan length. One seat is one named person who can fill surveys in the field. Supervisors need a seat only if they fill surveys themselves; Viewers are free. Short projects can use 1–3 month plans. Agencies running several projects get a volume quotation in writing after a demo.
Location is collected only during a shift, starting when the interviewer taps ‘Start shift’, with a clear notice on the phone, and stops at ‘End shift’. There is no background tracking, in line with Google Play policy. Respondents and shopkeepers are never tracked; only the submission point is recorded.
Because geography is preloaded, a single-city or single-district project is usually live in 1–2 working days, including forms (or XLSForm import), the outlet or household list, seats, assignments, quotas and dashboard. We run a 45-minute interviewer training in Hindi and provide live support on the first day.
Forms, outlet lists and assignments download in the morning while there is signal. After that, the whole interview or audit (arrival check, checklist, photos, outcome) works offline and is stored in the phone’s encrypted database. The Offline Work Permit lasts 24 hours; records sync automatically with unique IDs when signal returns. Weak indoor GPS is labelled ‘weak signal’, never passed off as ‘inside area’.
Our September 2026 check found SurveyCTO plans at about $225–630 a month, KoboToolbox paid tiers at $129–166 a month and ODK Cloud at $199–499 a month, all billed monthly in dollars. Prices change, so confirm on their sites. Bharat Survey works differently: active interviewer seats × plan length, in INR, with 1–3 month project plans and free Viewers. Quotes come in writing after a demo.
Bharat Survey checks six signals on every submission: location state against the outlet or area, mock-location signals, total and per-question duration below the minimum, duplicate outlet or household IDs, contradictory answers, and impossible time-distance sequences. No single signal proves fraud. Together they form a risk score that sends the record to a supervisor, a person decides with a reason, and the original evidence is never deleted.
Send your XLSForm or questionnaire. We’ll load it and walk you through the app, quota tracker, QA queue and dashboard in 20 minutes.
Open reference pages built from ECI, Census 2011, LGD and MCA records, with the source credited on every page.
