How will you select respondents?
Listen for a sampling frame, a random selection rule inside each area and a replacement rule. Quota sampling can be fine if the agency says so openly.
Guide · Buying research
For companies, departments, NGOs and campaign teams commissioning fieldwork
If you are working out how to choose a survey agency for a project in India, judge bidders on four things you can check: how they draw the sample, how they prove each interview happened, who owns the raw data, and whether they can interview in your respondents' language. This guide gives a 7-step selection method, questions to ask, and an RFP checklist.
- 7-step selection method
- Questions for the first call
- RFP and contract checklist
- Red flags that predict bad data
Sampling plan in writingGPS and time on every recordBack-check share namedRaw data handed overPilot before full fieldworkLocal-language interviewersClear consent textQuotation, not guesswork
Quick answer
Look for a written sampling plan, verifiable fieldwork with GPS, time and duration on each record, a stated back-check share, raw data and questionnaires as deliverables, interviewers who speak the local language, and a pilot before full launch. Price matters after these, because cheap data that cannot be audited is expensive later.
Why this decision goes wrong
Most buyers compare proposals by page count, sample size and price. None of these tells you whether interviews will be real, whether the sample represents the people you care about, or whether you can re-check the work. Agencies that cut corners usually do it in the field, where the buyer cannot see, so selection has to focus on what the agency will let you see.
Research agency or software, or both
Some teams need a full-service agency that designs, collects and analyses. Others have researchers in-house and only need a field team, or only need survey software for their own staff. Decide which gap you are filling before you issue a request, because the evaluation criteria differ for each.
Have ready before you ask
- Objective in two sentences
- Who must be interviewed
- Geography and languages
- Decision date
- Budget range, if fixed
The field
Where survey quality is won or lost
The proposal is written in an office. The data is made at a doorstep in a district two hundred kilometres away, often without network, by an interviewer paid per completed form. Choose the agency whose process still holds up there.
The method
Follow these in order. Steps 1 and 2 cost nothing and remove most unsuitable bidders before you read a single proposal.
01
1. Write a one-page brief
State the decision the data supports, the respondents, geography, languages and deadline. Vague briefs get vague quotes.
02
2. Shortlist on fit
Keep agencies that have worked with your respondent type and can field in your states and languages, and ask them to show how.
03
3. Ask for a method note
Request a short note on sampling, questionnaire length, pilot, supervision ratio and verification before asking for price.
04
4. Inspect sample data
Ask for an anonymised extract or demo dashboard showing what each record contains: location, time, duration, supervisor status.
05
5. Compare like with like
Normalise quotes to the same sample, length, languages and deliverables. A lower price for fewer back-checks is not cheaper.
06
6. Pilot before committing
Pay for a small pilot in two or three locations. It tests the questionnaire and the agency at the same time.
07
7. Contract the checks
Write back-check share, data handover, rejection rules and data protection into the contract, not only the proposal.
First call
Specific questions get specific answers. Evasive answers to these are the most useful signal you will receive.
How do I know an interview happened?
Good answers mention GPS, timestamps, duration, audio or photo where consented, and a named share of supervisor back-checks.
Who are the interviewers?
Ask about recruitment, training days, language ability, and how many interviewers each supervisor oversees.
What data do I receive?
Raw record-level data, codebook, final questionnaires in every language and the cleaning log should be standard deliverables.
How is personal data protected?
Ask about consent wording, who can access identifiable data, where it is stored and when it is deleted.
What happens to bad interviews?
The agency should explain rejection rules, replacement at its own cost, and how rejected records are reported to you.
Our position on choosing a partner
Signals
Before
Red flags
Each one predicts problems in the field
- Large sample quoted with no sampling method
- Delivery in days for a multi-district study
- Only summary tables offered, no raw data
- No pilot in the timeline
- Verification described as 'strict quality control' with no detail
- Promises of guaranteed accuracy or outcomes
With Bharat Survey
Good signs
Things a careful agency volunteers
- Margin of error stated with its assumptions
- Named back-check share and who performs it
- Record-level data with location and time fields
- Questionnaire length tested in a pilot
- Clear consent and data retention terms
- Limitations written into the proposal
Template
Copy these into your request for proposal or terms of reference. They make proposals comparable and protect data quality after award.
Put in the RFP
- Objective and decisions the data will inform
- Target respondents and geography
- Languages of interview
- Expected questionnaire length
- Deliverables list and formats
- Evaluation criteria and weight on method vs price
Ask bidders to state
- Sampling design and sample size calculation
- Pilot plan and sign-off point
- Supervisor to interviewer ratio
- Back-check share and method
- Verification fields on each record
- Team roles and time commitment
Write into the contract
- Raw data, codebook and questionnaires as deliverables
- Rejection and replacement rules
- Consent text and data retention period
- Data ownership with the client
- Interim progress reporting schedule
- Payment linked to verified completes
Decision aid
A full-service agency suits teams without in-house researchers who need design, fieldwork and analysis. A field-only team suits organisations with researchers who need interviews collected well. Software alone suits teams with their own staff in the field. Many projects mix these, and the table shows where each fits.
| Your situation | Full-service agency | Field team only | Survey software only |
|---|---|---|---|
| No in-house researcher | |||
| Researchers in-house, no field staff | |||
| Own field staff, need verification | |||
| Repeated monitoring by your own team | |||
| One-off multi-state study |
Bharat Survey offers field services and an offline Android survey app with a web panel, so projects can use either or both.
Costs
There is no fair standard rate. Cost is driven by sample size, interview length, how dispersed the sample is, the number of languages, travel to remote areas, verification depth and the analysis required. Ask for a quotation broken down by these drivers so you can see what changes the price and compare bids properly.
What moves the price most
Geographic spread usually matters more than sample size. Five hundred interviews in one city cost far less than five hundred spread across forty villages in hilly blocks. Each extra language adds translation, back-translation and interviewer matching. Deeper verification adds supervisor time but reduces the cost of discarded data later.
- Number of locations and travel time
- Interview length
- Languages
- Back-check share
- Analysis and reporting depth
Choosing a survey agency India-wide or local
A local firm may know one district well but struggle to run the same protocol across states. A larger survey agency India-wide can keep one method in many places but may subcontract fieldwork. Ask directly who will do the interviews in each state and how the protocol is enforced there.
Not the right fit if
- You need a CATI phone centre or IVR calling today
- You need an iOS-only field app
- You need a vendor with a specific empanelment we do not hold
- The study needs clinical or laboratory measurement
A good fit if
- Face-to-face fieldwork in Hindi and regional languages
- Projects that need GPS, time and duration on every record
- Teams that want software for their own staff
- Low-network districts where interviews must work offline
Bharat Survey home
Survey software and field services in one place.
Learn moreभारत सर्वे (Hindi)
The Bharat Survey homepage in Hindi.
Learn morePolitical surveys
Opinion research and booth-level fieldwork with neutral methods.
Learn moreCorporate research
Market research, retail audits and customer surveys.
Learn moreFeatures
Offline capture, GPS verification, web panel and reports.
Learn moreFAQ
Straight answers to the questions people ask most about this topic.
Should I choose the cheapest survey agency?
Only after the method and verification are equal. Price differences often come from fewer back-checks, shorter training or clustered interviews that cut travel. Normalise every quote to the same sample design, interview length, languages and deliverables, then compare. A slightly higher price for auditable data usually costs less than repeating a failed study.
How do I verify that interviews were really conducted?
Ask for record-level data with GPS coordinates, start and end times and interview duration, then check a sample yourself. Look for clusters of very short interviews or many records from one spot. Request the back-check log showing which respondents were re-contacted and what was found.
Who should own the survey data?
The client should, and the contract should say so. The agency delivers raw data, codebook and final questionnaires, and deletes or returns identifiable data after an agreed retention period. Agencies may keep anonymised copies only if the contract allows it.
Is a pilot worth the extra time?
Yes. A pilot in a few locations shows whether questions are understood, how long interviews really take and whether the agency's field process works. Problems found in a pilot cost little to fix; the same problems found after full fieldwork can invalidate the dataset.
How long should selecting an agency take?
For a straightforward project, a brief, a shortlist call and a method note can be done in a week or two. Government procurement follows its own timelines. Rushing selection to save a week often costs more time later when data has to be cleaned or re-collected.
India Nationwide, Hindi-first
Send the brief you are sending others. We will reply with a method note, a sample record and a quotation you can compare line by line.