A buyer-side guide to scoping primary market research in India, comparing quotes, controlling fieldwork quality and defining decision-ready outputs.
A market research proposal can look precise and still buy the wrong evidence. In India, two agencies may quote for the same number of interviews while assuming different cities, recruitment routes, respondent languages, quality controls and deliverables. The totals are not comparable because the scopes are not comparable.
This guide is for the person commissioning the work. It shows how to turn a business decision into a quote-ready brief, compare proposals on the same basis, and define evidence, fieldwork and data standards before the project begins.
Fix the decision, population, geography, languages, sample logic, deliverables and acceptance rules first. Then compare methodology, risk and price.
Context: the ₹29,008 crore figure is for the wider Indian research and insights industry, not primary fieldwork alone. MRSI reports that analytics accounts for nearly 60% of the total. Source checked 2 November 2026: Market Research Society of India.
1. Write the decision before the methodology
“Understand the Indian market” is not a research objective. It does not tell a provider which evidence matters, who can supply it or what management will do with the answer. A useful decision statement names the choice, alternatives and deadline.
Use that statement as the control point. Every interview, survey question and output should help resolve the decision. If an item will not change a choice, reduce uncertainty or explain a material risk, remove it.
2. Build a quote-ready brief in ten lines
Give every bidder the same one-page brief. Do not force a methodology before hearing the agency's reasoning, but remove ambiguity from the commercial scope.
- Decision: what management will approve, reject or change.
- Known facts: what desk research or internal data has already established.
- Unknowns: the evidence that is still missing.
- Population: who must be represented and why.
- Geography: states, cities, city tiers and sales environments.
- Languages: respondent language, working language and report language.
- Evidence standard: exploratory insight, directional validation or population estimate.
- Timing: management decision date and dependencies.
- Required outputs: raw data, fieldwork records, analysis and executive readout.
- Commercial format: itemized price, taxes, assumptions, exclusions and change rates.
Ask each bidder to list assumptions and exclusions in a separate section. This exposes whether translation, respondent incentives, travel, tax, data cleaning, transcripts or presentation work are missing from the headline total.
3. Choose the method around the evidence
| Business question | Useful method | What it cannot prove by itself |
|---|---|---|
| How common is awareness or preference? | Structured survey | Why respondents think or behave that way |
| Why do buyers reject the offer? | In-depth interviews | How prevalent each reason is in the population |
| What is actually stocked and priced? | Retail audit or observation | Demand outside the selected outlets |
| How is a service delivered? | Mystery shopping | Customer motivations not visible in the visit |
| Will a concept survive real purchase friction? | Pilot or test sale | Long-term repeat demand without follow-up |
| Who influences a B2B purchase? | Stakeholder interviews and account mapping | A statistically representative market share |
Online research is efficient for digitally reachable audiences. It should not automatically stand in for rural, low-literacy, older or offline populations. Face-to-face work can improve reach, but it adds interviewer, travel, supervision and safety requirements. Use mixed methods only when each method closes a named evidence gap.
4. Define whom the study represents before discussing sample size
A request for “500 Indian consumers” is incomplete. The number alone does not define what the results represent. Fix the following before accepting a sample recommendation:
Buyer, user, influencer or decision-maker, with verifiable screening conditions.
States, metros, tier-2 or tier-3 cities, rural areas and channel environments.
Probability, panel, quota, purposive, intercept or referral recruitment.
The smallest subgroup management expects to compare reliably.
Quota balance is not the same as probability sampling. A neatly balanced online panel may be useful for directional decisions, but it should not be described as representative of India unless the sampling frame, selection probabilities, coverage and weighting support that claim.
For a large national design, request the sampling frame, selection stages, replacement rules, expected design effect, weighting plan and limitations. For exploratory B2B research, ask how the provider will verify job role, purchasing involvement and company eligibility.
5. Treat language as part of research design
Translation is not a final production task. It can change who participates, how a concept is understood and whether response scales work as intended. Decide the language pathway before the questionnaire is approved.
- Write a clear master questionnaire and mark technical or culturally sensitive terms.
- Translate for meaning, not word-for-word equivalence.
- Use an independent review or reconciliation step for each field language.
- Test screening questions, scales, product concepts and consent text in-language.
- Record the approved version number used in each interview.
- For qualitative work, agree how quotations will be translated and checked.
A respondent may speak English at work but explain household behaviour, price sensitivity or product experience more accurately in another language. The bidder should justify the language plan by audience and location, not by internal convenience.
6. Normalize quotes before comparing price
There is no defensible single “average price” for primary research in India. Public rate pages mix fieldwork-only fees, full-service studies, different audiences and different deliverables. Use public prices for early budgeting, not as proof that two proposals are equivalent.
| Cost line | What the bidder should state | Common hidden difference |
|---|---|---|
| Design | Workshop, questionnaire or guide, sampling plan | Client is expected to supply a final instrument |
| Recruitment | Source, screening, verification and replacement | Convenience recruits used for hard-to-reach profiles |
| Respondent payments | Amount, mode, eligibility and tax treatment | Incentives billed later at actual cost |
| Languages | Translation, review, scripting and transcript language | Only the questionnaire, not open ends or transcripts, is translated |
| Fieldwork | Mode, locations, interview length and supervision | Travel or venue costs excluded |
| Quality control | Checks, audit trail, rejection and replacement rules | Basic automated cleaning presented as full QC |
| Analysis | Tables, coding, weighting, interpretation and workshop | Only raw data and standard charts included |
| Commercials | GST, currency, payment schedule and change rates | Taxes, bank charges or scope changes omitted |
If the client pays from outside India, also confirm invoice currency, beneficiary account, withholding documentation, bank charges and who bears exchange-rate movement. Commercial friction discovered after fieldwork can delay data release and final reporting.
7. Write fieldwork acceptance criteria before launch
“Standard quality checks included” is not an acceptance criterion. The contract or statement of work should name the evidence the buyer will receive and what happens when a record fails.
Do not impose one fixed back-check percentage on every project. The appropriate control depends on mode, audience, risk and available contact data. Require the bidder to state its rule and justify it. The 2025 ICC/ESOMAR Code also says clients should receive enough technical information to assess validity and may request independent checks on collection and preparation, subject to confidentiality.
8. Set consent, data and AI rules in the contract
Primary research can involve contact details, recordings, demographic attributes, opinions and behavioural data. Map the data flow before collection: client, research agency, panel or recruiter, fieldwork partner, transcription provider, technology platform and analyst.
| Contract question | Required answer |
|---|---|
| Why is each personal-data field collected? | Named purpose and minimum necessary data |
| What does the participant see? | Clear notice, consent language and withdrawal route |
| Who receives identifiable data? | Named roles, subcontractors and transfer conditions |
| How long is data retained? | Retention schedule plus deletion or anonymization evidence |
| Will AI or synthetic data be used? | Use case, data environment, human oversight and disclosure |
| What happens after a breach? | Escalation owner, notification process and response timeline |
India's Digital Personal Data Protection Act 2023 and Rules 2025 should be assessed against the project's actual roles and effective dates. MRSI's 2026 industry guidance specifically notes that governance considerations apply to clients, agencies, panels, fieldwork specialists, technology providers and cross-border engagements. Obtain legal advice for your facts; this article is operational guidance, not a legal opinion.
Primary references: Ministry of Electronics and Information Technology, MRSI professional standards, and ICC/ESOMAR Code 2025.
9. Specify decision-ready deliverables
A slide deck is not a complete definition of delivery. Specify the evidence package and the management output separately.
- Final questionnaire or discussion guide by language
- Sampling and recruitment record
- Fieldwork and quality-control log
- Clean raw data, codebook and weighting variables
- Transcripts, recordings or observation records as agreed
- List of exclusions, replacements and deviations
- Answer to the original decision statement
- Findings separated from interpretation
- Segment or city differences with base sizes
- Evidence that challenges the preferred hypothesis
- Limitations and what cannot be concluded
- Recommended next action, owner and trigger
Define file formats, ownership, permitted reuse and the number of review rounds. If the buyer needs editable charts, respondent-level data or Japanese-language management material, state that before pricing.
10. Score vendors on evidence, not presentation
Use the same scorecard for every proposal. Weighting should reflect project risk, but the categories below prevent a persuasive pitch from hiding a weak execution plan.
| Criterion | Suggested weight | Evidence to request |
|---|---|---|
| Understanding of the decision | 15% | Restated decision, hypotheses and exclusions |
| Method and sample logic | 20% | Coverage, recruitment, limitations and analysis plan |
| India execution capability | 15% | Relevant locations, languages, audience and senior oversight |
| Quality-control plan | 20% | Pilot, checks, rejection rules and audit trail |
| Data, consent and security | 10% | Data map, roles, retention and incident process |
| Deliverables and decision support | 10% | Exact files, analysis, workshop and limitations |
| Commercial clarity | 10% | Itemized price, assumptions, taxes and change rates |
The weights total 100%. Change them before bids are opened, not after seeing which vendor is cheapest. Record the reason for each score and any condition that must be resolved before award.
11. Use a gated project timeline
A schedule should show decisions and dependencies, not just start and end dates.
Do not compress the pilot to recover time lost during procurement. A short pilot can expose unclear screening, broken translations, unrealistic interview length and recruitment fraud before those problems spread across the full sample.
12. Frequently asked questions
How many respondents do we need for India?
There is no India-wide default. Sample size follows the claim you need to make, the selection method, expected variability, design effect, subgroup analysis and non-response. Define the population and smallest reporting cell before asking for a number.
Is online research enough?
It can be appropriate for digitally reachable audiences and directional decisions. It is not automatically representative of offline, rural, low-literacy or hard-to-reach groups. The proposal should explain coverage gaps.
Should we select the lowest quote?
Only after normalizing scope. A low bid may exclude translation, incentives, travel, quality checks, transcripts, analysis or tax. Compare the evidence and risk purchased, not only the total.
What should the client own?
State ownership and permitted reuse in the contract. Common requirements include the final instruments, clean data, codebook, agreed recordings or transcripts, QC documentation and final report, subject to participant consent and privacy restrictions.
Can AI be used in analysis?
Yes, if the use is disclosed, confidential data remains controlled, human oversight is defined and outputs are validated. The ICC/ESOMAR 2025 Code specifically requires clients to be informed about AI, synthetic data and synthetic personas used in datasets, analysis, reporting or interpretation.
Planning primary research in India?
Jindoxx can help turn a management question into a comparable research scope, coordinate multilingual fieldwork and build a decision-ready evidence package.
Discuss your research brief