Primary vs Secondary Research: When Published Data Runs Out
Primary research collects fresh information from people; secondary research reuses what others collected. Secondary research stops answering when your decision needs a city, brand or reason the published data skips.
By Jitendra Kumar Nirala, Founder · Updated 13 Nov 2026

Key takeaways
- Primary research is information you collect for your own question. Secondary research is information someone else collected for another purpose. Decide fact by fact, not project by project.
- Secondary research is enough for a fact only if the source passes five checks: level, category, sample, definition and date.
- Official Indian tables are strong on how many and how much. They lump categories, report at state or national level and use definitions that differ from yours, so reasons and city-level behaviour need fieldwork.
A review of 23 saturation tests (Hennink and Kaiser, 2022) found most narrow studies reached saturation, the point where new interviews stop giving new themes, within 9 to 17 interviews or 4 to 8 groups.
In this article
- 1What is the difference between primary research and secondary research?
- 2When does secondary research stop being enough for a business decision?
- 3Does the source publish your level: city, segment or brand?
- 4Is your category its own line, or lumped in with others?
- 5Is the sample behind your cell large enough?
- 6Does the source's definition match the words in your question?
- 7Was the fieldwork done before the change you care about?
- 8What does the five-check test show on an illustrative beverage decision?
- 9Which primary research format fills which gap?
- 10How many interviews or groups does one gap need?
- 11How do we turn a failed check into a study?
- 12Frequently asked questions
The Census counts 52.83 crore Hindi speakers (1 crore is 10 million), but only 61% of them gave Hindi itself as their mother tongue (our calculation, shown below). A team planning Hindi interviews from that headline would recruit the wrong people. Published Indian data also tells you how many households own a smartphone and how much urban families spend on processed food. It rarely tells you why a shopper in a tier-2 town (a mid-sized city below the metros) switched brands.
That gap, more than the textbook line about who collected the data, is what separates primary research from secondary research for a consultant or a brand team. The textbook line is correct, but it does not help you decide what to commission on Monday. The useful question is whether a particular number can carry your decision.
This page gives you a five-check test to run on any published source, worked through on three official Indian datasets. Where a source fails a check on a fact your decision rests on, that fact needs fieldwork with recruited people. Where it passes, you can spend the budget where it is needed.
What is the difference between primary research and secondary research?
Primary research is information you collect yourself, from people or places, for your own question: interviews, group discussions, observation, diary tasks, product tests and surveys. Secondary research, also called desk research, is information someone else collected for another purpose, such as government tables, company reports and published studies. The difference is who collected the data, and why.
What matters in practice is fit. Secondary data answers someone else's question, in someone else's categories, on someone else's date. Primary research is built around yours. So most projects need both. Read what already exists first, then spend fieldwork money only on the facts it cannot give. Our market research glossary entry covers the wider picture.
When does secondary research stop being enough for a business decision?
Secondary research stops being enough when a fact your decision depends on fails any of five checks: level, category, sample, definition or date. Run the checks on one fact at a time. A pass means you can use the number. A fail means that fact needs fieldwork with recruited people. Ignore facts that are only background; they do not need a study.
Does the source publish your level: city, segment or brand?
Pass if your exact cell is a published row. Fail if you would have to infer it from a larger area.
The Household Consumption Expenditure Survey 2023-24, run by the Ministry of Statistics and Programme Implementation (MoSPI), collected data from 2,61,953 households (1,54,357 rural and 1,07,596 urban) between August 2023 and July 2024. Its factsheet gives the split of spending by item group for rural and urban India as a whole. It gives average monthly spending per person (monthly per capita consumption expenditure, MPCE) of ₹4,122 in rural India and ₹6,996 in urban India, with separate MPCE estimates for each state and union territory. A city is not a row in the factsheet.
Is your category its own line, or lumped in with others?
Pass if your product has its own line. Fail if it sits inside a bigger group.
In the MoSPI Household Consumption Expenditure Survey 2023-24 factsheet, "beverages, refreshments, processed food" is a single item group that also includes purchased cooked meals. It comes to ₹406 per person a month in rural India (9.84% of spending) and ₹776 in urban India (11.09%). Our calculation: ₹776 × 4 people = ₹3,104 a month for an illustrative household of four. The group includes cooked meals and many other products, so this is not the size of any one product's market. The household size of 4 is our assumption.
Is the sample behind your cell large enough?
Pass if the publisher gives estimates at your level without a warning. Fail if it warns you off.
The Comprehensive Modular Survey: Telecom (CMS-T) covered 34,950 households between January and March 2025. It found that 85.5% of households had at least one smartphone and 86.3% had internet access at home. MoSPI's press note of 29 May 2025 says the main aim was national-level estimates. It says state-level figures must be read with the sample size and the relative standard error (RSE, a measure of how far an estimate could move because only some households were asked) in mind.
Our calculation: 34,950 households ÷ 36 states and union territories is about 970 each on a simple average. The sample was spread in proportion to population, so small states have far fewer. Any age or gender cut inside a state has fewer still.
Does the source's definition match the words in your question?
Pass if the definition matches your wording. Fail if the same word means something else.
In CMS-T, "owns a mobile phone" means having a device with an active SIM card for personal use, including employer-provided phones. "Used" means at least once in the last three months for calls or the internet.
Census 2011 Table C-16 defines mother tongue as the language the person's mother spoke to them in childhood. That is not the language someone is comfortable being interviewed in. The Census also groups many mother tongues under one heading. "Hindi" is 52,83,47,193 people (43.63%, or 52.83 crore; 1 crore is 10 million). Of these, 32,22,30,097 gave Hindi itself as their mother tongue. The other 20.61 crore gave mother tongues grouped under it, such as Bhojpuri (5.06 crore) and Rajasthani (2.58 crore).
Our calculation: 32,22,30,097 ÷ 52,83,47,193 = 61%. So "Hindi 43.63%" is the wrong number to plan Hindi sessions on. Ask about language comfort in the screener (the short set of questions that checks whether a person fits the study) instead, and read our note on multilingual research.
Was the fieldwork done before the change you care about?
Pass if fieldwork finished after the last big change in your category, such as a price move, a launch or a policy. Fail if it came before.
The Census language tables date from 2011, 15 years before 2026. The consumption survey ran from August 2023 to July 2024 and the telecom survey in January to March 2025. Language geography moves slowly, so an old language table can still help. A spending or phone table from before your category's last big change cannot.
A current example: the 2023-24 spending data was collected before the GST rate changes of 22 September 2025, which cut tax on most packaged foods and non-carbonated juice drinks to 5% and raised it on carbonated drinks to 40% (Press Information Bureau). A beverage team should treat price-sensitive spending figures from that survey as background.
Add up the fails on the facts your decision rests on. Each one is a line in your fieldwork brief.
What does the five-check test show on an illustrative beverage decision?
On a typical decision the test passes the background and fails the facts that matter. The scenario below is illustrative, not a real client.
A beverage team is deciding whether to test a ready-to-drink tea with urban households in tier-2 cities.
- Level: fails. There is no city row.
- Category: fails. Tea sits inside a lumped group that also includes purchased cooked meals.
- Sample: fails. There is no city-level estimate to test in the factsheet.
- Definition: fails by default. "Regular ready-to-drink tea buyer" is the team's own phrase, so the team must define it in the screener.
- Date: fails for price-sensitive facts. The 2023-24 fieldwork ended in July 2024, before the GST rate changes of 22 September 2025 cut tax on most packaged foods and non-carbonated juice drinks to 5% and raised it on carbonated drinks to 40%. Use the ₹776 figure only as background.
Usable as background: urban households spent ₹776 per person a month on beverages, refreshments and processed food, including purchased cooked meals, in 2023-24.
The "existing knowledge" paragraph for the brief then reads: "Urban households spent ₹776 per person per month on beverages, refreshments and processed food, including purchased cooked meals, in 2023-24. We hold nothing that splits this by city or by product. Unknown: whether 2023-24 spending still holds after the GST rate changes of 22 September 2025, why households in the target cities choose or avoid ready-to-drink tea, where they buy it and how it fits into the day at home."

Which primary research format fills which gap?
Match the type of gap to the format. Reasons need conversation, behaviour needs observation and product use needs a trial at home.
| The gap | Format we recruit for | What it involves |
|---|---|---|
| How big is the pool, who owns which device | Secondary research | Use the published tables if they pass the five checks |
| Why people choose, switch or refuse | In-Depth Interview (IDI) | One-on-one, 30 to 120 minutes, online or in person |
| How people compare options and react to each other | Focus Group Discussion (FGD) | Moderated, 1 to 3 hours, online or in person |
| What actually happens at home | Ethnography (see contextual inquiry) | Home visits or virtual sessions |
| What happens over several days | Digital diary study | Multi-day photo and video tasks on a mobile app or the web |
| What happens at the shelf | Shop-along | In-store or online |
| Whether a product works in daily use | Product test (in-home use test, IHUT) | Participants try the product at home |
| How an app or site is used | Usability and app testing | Remote or lab-based |
How many interviews or groups does one gap need?
Hennink and Kaiser's 2022 review of 23 tests of saturation, the point where new interviews stop producing new themes, found most narrow studies reached saturation within 9 to 17 interviews or 4 to 8 groups. Our recommendation is to plan that many per segment, and more when the aim is broad or participants differ.
Our recommendation is also to size each city or user type separately and to recruit above your target to cover no-shows. If you need counts alongside explanations, our guide to mixed methods research shows how to link surveys and interviews.

How do we turn a failed check into a study?
You send us a brief. We check that it is feasible and send a tailored proposal and quote. Once scope, commercials and timelines are agreed, we commission the project and begin recruitment.
Put your failed checks in the "existing knowledge" field of our brief template. Tell us the research objective, the method, the audience, the participant profile, the geography, the sample size, the usership (who uses the product or category, and how often), the timelines and the recruitment criteria.
We recruit by city and target by household and demographic profile. We check participant identity before invitation. We run sessions in languages other than English, and we can supply moderators. Studies may be conducted by us, by research partners or by other research organisations, depending on the project. Deliverables depend on scope. They can include recruitment, fieldwork materials, recordings, transcripts, structured data exports and research reports. All of it is agreed before the study begins.
We do not work from a fixed rate card. Pricing depends on the method, audience, participant profile, sample, usership, geography, recruitment complexity and overall scope. We offer a free pilot, sample or trial.
To start, email your brief to support@qualfacto.io. We aim to reply within one business day, Monday to Friday, 10:00 to 18:00 IST (Indian Standard Time).
If you would like to take part in studies as a participant, see how to join and get invited. Invitations depend on each project's requirements and your profile, and rewards depend on the specific study's terms.
Turn your failed checks into a study
Email your brief to support@qualfacto.io and we aim to reply within one business day, Monday to Friday, 10:00 to 18:00 IST.
Frequently asked questions
Sources
- 1.Fact Sheet: Survey on Household Consumption Expenditure 2023-24, Ministry of Statistics and Programme Implementation, Government of India
- 2.Press Note: Results of Comprehensive Modular Survey: Telecom, 2025, Ministry of Statistics and Programme Implementation, Government of India, 29 May 2025
- 3.Census of India 2011, Paper 1 of 2018, Language, Table C-16, Office of the Registrar General & Census Commissioner, India
- 4.Next-Gen GST Overhaul: Staple Foods Tax-Free, Processed Foods at 5%, Press Information Bureau, Government of India
- 5.Sample sizes for saturation in qualitative research: a systematic review of empirical tests, Hennink, M. and Kaiser, B. N. Social Science & Medicine, 292, 114523, 2022



