Quota sampling

Imagine the situation: a research company needs to survey 1,000 city residents within a week. It has no database of all residents, so a "clean" random sample is impossible. If you simply survey whoever agrees first, you get a skew: more active, mobile, loyal people. To get closer to the real structure of the population, the company sets quotas in advance — by gender, age, district. The interviewer does not just "catch" people on the street: they recruit the required number of respondents into each of the predefined cells. This is how quota sampling arises in practice.

This approach is widely used in commercial and social research when there is no time or resources for a strict probability scheme, but relying on a fully convenience sample is risky. Quotas make it possible to keep the key audience parameters under control even under constraints.

Definition and essence of quota sampling

Quota sampling is a way of recruiting respondents in which the researcher sets quotas (norms) in advance for important audience characteristics — gender, age, type of locality, customer group, and so on — and then recruits people until these quotas are met. Within each quota the selection of respondents is usually non-probabilistic: the interviewer decides who exactly will be included in the sample.

By design, the quota scheme resembles a stratified sample, but with one important difference: in ideal stratification the selection within each stratum is random and controlled, whereas in quota sampling it is a convenience selection. That is why quota sampling belongs to non-probabilistic methods, even if from the outside it looks "very representative".

When quota sampling is useful

There is no complete list of the population. In city-wide and national surveys it is rarely possible to obtain a complete register of all residents or customers. Quotas at least bring the structure of the sample closer to known statistics: censuses, official statistical data, CRM systems.

Tight deadlines and budget. A full probability design requires more time and resources: you need to prepare a sampling frame, random selection, contact control. The quota scheme is simpler to implement and faster in the field stage, which matters in "needed by Monday" projects.

You need a reasonable compromise between quality and cost. In its pure form a convenience sample is too vulnerable to criticism, while strict stratification may turn out to be unattainable. Quotas allow you to take a step away from a chaotic recruitment of respondents toward a controlled audience structure.

How quotas are set

Choosing the variables. First of all, quotas include characteristics that strongly affect the metric under study: gender, age, region, company size, product type. For social surveys it is helpful to rely on guidance from materials about sociological research and official statistics.

Defining the target shares. The sources can be government statistics, CRM data, previous studies, the results of monitoring projects (monitoring research). For example: "women — 55%, men — 45%; age 18–24 — 15%, 25–44 — 50%, 45+ — 35%".

Transferring quotas into an operational scheme. For each interviewer or recruitment channel a table is defined: how many respondents with particular characteristics need to be recruited. As the work progresses, quotas "close": when, for example, enough men aged 18–24 have been recruited, the interviewer stops including new respondents from this cell in the sample.

Quota sampling in online surveys

On the internet, quotas are most often implemented not through manual selection by an interviewer, but through a combination of filters, triggers, and panels.

Screening questions and display logic. In a survey builder you can start the questionnaire with a short screening block and, if a respondent does not match the quota, end the survey early (for example, via a "Thank you, you are not part of the target group" screen). Branching logic and screening questions, described in the glossary of questionnaire elements, provide additional help.

Working with respondent panels. Panel services let you set audience parameters: gender, age, region, interests. Inside the panel the quotas for these parameters are controlled automatically, and you do not need to manually keep track of closing the cells.

Mixed schemes. Often a researcher combines several channels: some quotas are filled through a panel, some through their own communications with customers, some through media placements. It is important that the total structure of the sample matches the set quotas, even if it differs slightly across channels.

How to implement this in SurveyNinja

In real projects quotas are most often implemented not by one "magic button", but by a combination of several settings. In this sense SurveyNinja provides a builder from which you can assemble a working scheme for a specific case.

1. Screening and filters at the start of the questionnaire. In the first 2–3 questions you place the key quota criteria: gender, age, role in the company, region. Based on the answers you turn on branching logic: suitable respondents are shown the main questionnaire, unsuitable ones — a short closing screen. This reduces noise and helps keep the sample structure more accurate.

2. Limits on the number of responses. Through the collection limit settings ("Configuring limits" in SurveyNinja help) you can fix the maximum number of completed questionnaires for the project. In simple cases this is enough: you know that you are recruiting, for example, 500 respondents from the site's audience, and after that the collection stops automatically.

3. Splitting the flow into several links. When distributing the survey across different channels (email, social media, partner databases), it is convenient to use separate links and hidden variables for each source. Then in the report it is easy to control how many questionnaires came from each "branch", and if necessary slow down the channels that have already closed their quotas.

4. Pairing with a respondent panel. If a more rigorous quota scheme is needed, SurveyNinja can be used as a "showcase" for the questionnaire, while recruitment of respondents is delegated to a partner panel. In this case you set the target shares by gender, age, and geography on the panel side, and in SurveyNinja you simply receive the already filtered responses into a single database.

This approach does not turn an online survey into an ideal academic study, but it lets you build quotas into the usual workflow with the builder without making it more complicated for a marketer or product analyst.

Pros and cons of quota sampling

Advantages. Quotas help avoid crude skews (for example, when a survey "about the city" suddenly turns out to be mostly young men from the center) and make the sample visually similar to the population being studied. At the same time, such a design is cheaper and simpler than strict probability schemes.

Limitations. Formally, quota sampling remains non-probabilistic: within the quotas respondents are selected by convenience, not at random. This means that classic estimates of the margin of error and confidence intervals work only approximately. In addition, quotas are usually set for several characteristics, while skews are possible for the rest.

The risk of "over-control". If you try to account for too many parameters, the quota table becomes complex and practically unworkable. Interviewers start "filling" cells with any available respondents just to meet the plan, and this worsens data quality.

Common mistakes

Quotas on only one characteristic. A frequent situation: the researcher controls only gender and age, ignoring, for example, the type of locality or customer status. As a result, the sample may turn out "correct" on the basic parameters, but heavily skewed on other important characteristics.

Relying on outdated data. If the structure of the population is taken from an old report or a dubious source, the quotas will only reinforce an incorrect picture of reality. For city-wide and national surveys it is worth relying on fresh data from official statistics and relevant research.

No control over quota fulfillment. Recruiting respondents "to the max" without regular reconciliation with the quota table often leads to some cells being heavily overfilled by the end of the field stage, while others are almost empty. Fixing this after the fact can be difficult and expensive.

Mixing with "snowball" sampling. Sometimes the quotas are supplemented with a recommendation "let the respondent invite similar acquaintances" — essentially this is already a different scheme, closer to Snowball Sampling. It is important to understand that such steps take the project even further away from probability approaches and complicate the interpretation of the data.

Practical recommendations

Choose 2–4 key characteristics for quotas. It is better to control a few truly important parameters well than to try to account for everything at once and end up losing control. Start with gender, age, and a broad regional breakdown, and then add characteristics as needed.

Spell out the selection rules within quotas. Even if the selection is non-probabilistic, set simple and clear criteria for interviewers: where to look for respondents, what limits there are on the number of interviews in one place, how to avoid an "overload" of active users. This reduces the risk of strong skews within the cells.

Document the methodology in the report. In the "Methodology" section, honestly describe which quotas were used and exactly how the selection of respondents took place. This will help readers of the report correctly assess the reliability of the results and not treat the quota sample as a full-fledged probability one.

Combine quota sampling with subsequent weighting. If after the field stage you see residual skews, you can apply response weighting methods, which the term Weighted Survey describes. This will not turn the quota sample into a perfect one, but it will help align it a little better with the structure of the population.

Quota sampling is a working compromise that lets you build the basic logic of representativeness into real constraints on money, time, and access to the audience. It is important to remember that it is precisely a compromise: it makes survey results more robust than a chaotic convenience sample, but it does not provide the same guarantees as a strict probability design.