Quota Sampling Calculator for India: Cell Sizes and Minimum Sample

• Cell target = total sample × the cell's population share, and the smallest cell sets the minimum total: minimum per cell ÷ smallest share. For India's 10 adult age and gender cells at 30 per cell, that is 388 interviews.
• Crossing gender with age roughly doubles that minimum: age bands alone need 188 for the same 30-person floor.
• Equal cells cost little precision for age and gender in India (design effect 1.026, meaning how much unequal weights inflate variance). They reach 30 per cell with 300 interviews instead of 388.
• A quota sample does not normally support a margin of error. Report the quotas, the counts per cell and how people were selected, and leave out a ± figure.

A quota sampling calculator works out three numbers: how many people to recruit in each cell, the smallest total sample your grid can carry, and how much precision unequal quotas cost.

Quota sampling means recruiting from defined groups until each group's target count is filled, with the people inside each group chosen by the recruiter rather than at random. It is the non-probabilistic version of stratified sampling. A cell is one combination of groups, such as women aged 25 to 34. On this page, an interlocking quota means a target for each combination, and an independent quota means a target for each variable on its own.

Inputs are the total sample (n, default 400), the minimum per cell (m, default 30, our rule of thumb, not a standard), the population share of each cell (pᵢ, % of adults aged 18+, in the table in the next section) and the planned sample share of each cell (sᵢ, % of n, default equal at 10% for each of 10 cells).

The design effect is the factor by which unequal weights inflate variance. The effective sample size is the size of a random sample that would give the same precision. Here, weights are cell population share ÷ cell sample share. When sᵢ = pᵢ, deff is exactly 1. The results and their formulas are below; for the design effect and effective sample size, see PracTools.

Among India's adults aged 18 and over, the ten age and gender cells run from 7.73% (men aged 60+) to 12.60% (men aged 25 to 34), based on the government's projected population for 1 March 2026.

How we worked this out. We took the projected population by age and sex for 1 March 2026 from Table 18 of the Technical Group on Population Projections report (July 2020). For the 18–24 band we added single ages 18 and 19 from Table 20 of the same report to the 20–24 band. Dividing each cell by the adult total of 1,018.3 million (1,018,315 thousand) gives the shares. The age bands are our choice. Targets at n = 400 are rounded to whole people by largest remainder (round down, then give the leftover people to the cells with the biggest fractions), so they sum to exactly 400.

This is a projection from the 2011 Census, not a count. For the urban and rural split, the Ministry of Statistics and Programme Implementation (MoSPI) projects 513.45 million of 1,425.90 million people as urban in 2026, which is 36.0%, leaving 64.0% rural. That share covers all ages. The age tables are not split by urban and rural, so treat location as a separate quota. Do not cross it with age and gender.

For socio-economic class quotas, see our guide to NCCS classification.

A quota sample needs at least the minimum per cell divided by the smallest cell's share. With 30 per cell and men aged 60+ at 7.7326% (0.077326), that is 30 ÷ 0.077326 = 387.97, so 388 interviews (rounded up). For 50 per cell it is 50 ÷ 0.077326 = 646.6, so 647.

The table shows cells below a 30-person minimum, with targets rounded to whole people. At 300, the four short cells are men and women aged 18–24 and men and women aged 60+.

Interlocking is what pushes the minimum up. Age bands alone, with no gender cross, have a smallest share of 16.0% (163,147 of 1,018,315 thousand), so the same floor needs 30 ÷ 0.16021 = 187.3, rounded up to 188 interviews. Adding gender as an interlocking quota takes it to 388, about double.

Our recommendation: interlock only the variables you will compare across cells. Run the rest as independent quotas. The cost in interviews is visible before fieldwork, so check it first.

For age and gender in India, equal cells cost very little: ten equal cells give a design effect of 1.026, so 400 interviews behave like about 390. The cost is small because the cell shares are close to each other.

Our working: Kish's approximation is deff = 1 + CV²(w), the relative variance of the weights. With cell weights wᵢ = pᵢ ÷ sᵢ it reduces to Σ pᵢ² ÷ sᵢ (our derivation). With sᵢ = 10% for every cell and the pᵢ from the table, Σ pᵢ² ÷ sᵢ = 1.026. Equal age bands alone give 1.025. A 50/50 gender split against 50.7% men gives 1.000.

Equal cells also reach the floor sooner. Ten cells × 30 = 300 interviews, against 388 for proportional cells. Effective sample size at 300 is 300 ÷ 1.026 = about 292.

The cost grows when a group is far from its share. Using the same formula:

• Using the all-ages 36/64 urban and rural split as a stand-in for adults (illustrative), a 50/50 sample gives deff 1.078, so 400 behaves like about 371.
• An illustrative case: a group that is 10% of the population but 33.3% of the sample gives 0.1² ÷ 0.333 + 0.9² ÷ 0.667 = 1.245, so 400 behaves like about 321. The shares in this second case are invented to show the arithmetic.

Two limits apply. The formula prices only the weighting, and it assumes random selection within each cell (PracTools). A quota sample has no random selection, so treat 1.026 as a minimum cost. Weighting a quota sample back to population totals is a long-standing practice. So set cell sizes by what you need to read in each cell, and match the population only where a headline total matters.

A quota sample cannot carry a margin of error the way a random sample can, because recruiters choose who fills each cell. The American Association for Public Opinion Research (AAPOR) task force on non-probability sampling notes that when interviewers choose within quotas, their choices generally lead to selection bias.

A quota only reduces that bias if the variables you quota on are the ones that drive the outcome you are measuring. The same report quotes Neyman on a 1929 Italian purposive sample, a method akin to quota sampling: its averages on seven control variables agreed with the country, but agreement on characters not used as controls was often poor. It also records that in the United States, Gallup's quota poll called the 1936 election correctly, while the three leading pollsters, all using quota sampling, called 1948 wrongly. A 1949 review named quota sampling as one potential source of error among many.

There is one narrow exception: a 1966 method by Sudman shows sampling-error estimates can apply to quota samples drawn under his specific procedure and assumptions. Few briefs meet that standard.

What to report instead: the quotas, the count in each cell, how people were found and screened, and the effective sample size if you weighted. Skip the ± figure. Percentages and margins belong to quantitative designs; see qualitative vs quantitative research for choosing between them.

In qualitative research, quotas set who must be in the room, not what share of the population they represent. Write one row in the screener for each slot, then fill each row.

An illustrative design: 16 in-depth interviews in two cities, crossing two genders, two age bands and the two cities, gives 2 × 2 × 2 = 8 cells of two people each. For a focus group, quotas become group composition: who sits together, and how many of each type.

Three habits help:

1. Interlock only what you must compare, such as city or gender. Make everything else a minimum count.
2. Recruit a reserve for the hardest cells, because people drop out. See over-recruitment.
3. Put each quota into the screener as a pass or fail test. Our research screener template shows how.

For how many interviews or groups a study needs, see sample size in qualitative research.

We take your quota grid as part of the brief. Send us the research objective, method, target audience, geography, sample size, timelines and the grid. We review it, assess feasibility, and send a tailored proposal and quote. Our market research brief template is a one-page way to write it.

What we can do on a brief like this:

• Recruit by city and target by profile. We recruit by city and target by household and demographic profile.
• Match on profile details. Our participant profile records gender, date of birth, state, city and PIN code, household income band, education, the chief wage earner's education, which household items the household owns, and research languages.
• Check identity before invitation. We check participant identity before anyone is invited.
• Run sessions in languages other than English. You can also supply your own moderators.
• Run qualitative formats. In-depth interviews, focus groups, diary studies, online bulletin boards, ethnography, shop-alongs, product tests, and usability and app testing.

Our panel registration currently accepts adults aged 18 to 55. Cells for ages 56 and over, which include 60+ and part of 45–59 in this grid, are ones to raise with us at the brief stage, where we assess feasibility.

Participants are invited only when their profile matches a study's requirements. Registration is an expression of interest and does not guarantee an invitation, and rewards depend on each study's terms. People who want to take part can register as a participant.

We do not use a fixed rate card. Cost depends on method, audience, sample requirements, geography and recruitment complexity. We offer a free pilot. Email your grid to support@qualfacto.io, Monday to Friday, 10:00 to 18:00 IST. We aim to reply within one business day.