What Is a Poll? Definition, Types, and How to Judge One

Try a quick test. Two headlines: “Poll: majority backs the plan” and “Reader poll: majority backs the plan”. Both lean on the word poll, and only one of those numbers says anything about the country.

That is why the word deserves a closer look. A poll is a systematic questioning of a larger group of people who all receive the same questions in the same wording, so that their answers can be counted up and compared. Whether the result holds beyond the people who took part depends on how they were selected, not on how many there were. By the end of this article you will be able to size up any published poll in two minutes.

📌 The key points at a glance

• A poll asks many people the same questions in the same wording.
• The questionnaire is the tool, the poll is the whole project.
• Selection decides whether a result is representative, not sample size.
• AAPOR requires eleven disclosures at the moment results are released.
• Margin of error applies to probability samples only, says AAPOR.

What is a poll? The definition

A poll is a systematic questioning of a larger number of people about their opinion on a matter of general interest, whose answers are then counted and analyzed statistically. In American usage the word also covers the act of voting itself, which is why polling place and the polls refer to elections rather than to research.

Three features sit inside that definition, and all three have to be present. Take one away and you have a collection of opinions rather than a poll.

• More than one person answers, not a single individual.
• Wording and order of the questions stay the same for everyone.
• The answers are pooled and expressed as numbers.

The second feature is the one people underestimate most. Ask ten friends off the cuff and you will phrase the question slightly differently each time, which means the answers can no longer be added up. Standardization is not bureaucracy, it is the precondition that makes counting legitimate in the first place.

Poll, survey, and questionnaire: what each word means

Poll, survey, and questionnaire describe three different things: a poll usually measures opinion with a small number of questions, a survey is the wider research instrument, and the questionnaire is the tool that carries the individual questions. Everyday usage swaps them freely, and in a thesis that shows.

People often say poll and survey are interchangeable. In conversation that is fine, in research writing it is not: a poll is the narrower case, built around opinion, few questions, and quick publication, while a survey can run to a hundred items and measure behavior rather than views. The methodological ground both share is set out in the article on the survey as a research method.

The questionnaire is the tool. The poll is what you plan to do with it.

Which types of poll and survey exist

Purpose is what separates one type from another, and purpose then drives almost everything else: the audience, the length of the questionnaire, the demands placed on the sample, and how robust the result has to be. Picking a date for a team outing and asking about vote choice before a presidential election share a name and little else.

Six types cover nearly everything you will meet in practice. Effort and the demands on sample selection rise from the bottom of the table to the top.

A second question cuts across all six: how the answers are collected. Four routes are common, namely online, on paper, by telephone, and face to face. Online has become the standard route for most projects because it is quick, cheap, and produces answers that are ready to analyze. What that means methodologically is covered in the article on online surveys.

💡 Tip

Write down the purpose of your poll in a single sentence before you draft the first question. Anything that does not serve that sentence gets cut. This is the fastest way to take a questionnaire from 30 questions down to 15, and short questionnaires are finished far more often.

How a poll is run: the five steps

A poll runs in five steps: settle the research question, define the audience and the sample, build and test the questionnaire, distribute it and cap the field period, then analyze and interpret the results. The order is binding, because each step narrows the room the next one has.

Skipping step one is the expensive mistake. Start without a clear question and you will only notice at the analysis stage, by which point the data are already collected. Generally speaking, the more precisely you know what the closing sentence should say, the fewer questions the questionnaire needs. That half hour at the start saves you days later!

1. Settle the research question. Write down in one sentence what you want to know afterward and what you will do with the answer.
2. Define audience and sample. Decide who the statement should apply to and how you draw people from that group. The article on sampling goes through the options.
3. Build and test the questionnaire. Draft the questions, order them from easy to demanding, and have three to five people complete a trial run. The article on creating a questionnaire covers this step by step.
4. Distribute and cap the field period. Fix in advance how long the poll runs. An open-ended field period invites you to keep collecting until the hoped-for result appears.
5. Analyze and interpret. Work out the figures, always state how many people answered, and say plainly who the result applies to.

How to spot a credible poll

A credible poll discloses its own conditions: who ran it, who was interviewed, how many people that was, when and by what method they were questioned, how they were selected and weighted, and how the questions were worded. Without those details a percentage cannot be judged at all.

The international industry standard sets this out precisely. The ESOMAR and WAPOR guideline on published opinion polls lists eight items that must accompany every publication.

• Name of the research organization that carried out the poll
• The universe effectively represented, meaning who was interviewed
• Achieved sample size and geographical coverage
• Dates of fieldwork
• Sampling method, plus response rate for random probability samples
• Method of collection, such as face to face, telephone, or online panel
• Whether weighting was applied and to which universe
• Wording of the questions the published results rest on

The same ESOMAR and WAPOR guideline also settles a common argument in one line: a small-sample scientific survey is always better than a large-sample self-selecting one (accessed August 24, 2026). For vote choice it recommends a sample of no fewer than 1,000 respondents. Why size alone still does not do the job is explained in the article on representative surveys.

⚠️ Warning

Reader polls, click votes, and polls on social media are self-selecting: whoever wants to answers, and that tends to be the people with the strongest views. Numbers like these describe the participants and nobody else. As a mood check they are useful, as evidence about the population they are not.

What applies additionally in the United States

In the United States the reference point is a professional code of ethics rather than a statute. The AAPOR Code of Professional Ethics and Practices had its Standards for Disclosure last revised in April 2021 and lists eleven items that have to be disclosed at the moment results are released, among them the data collection strategy, the sponsor and any separate funding source, the exact question wording with its context, the population under study, the recruitment method, the modes and languages of collection, the field dates, sample sizes, weighting, and a statement of the limitations (accessed August 24, 2026).

Two points in the AAPOR Code are worth memorizing because they decide how a number should be read. First, researchers must state plainly whether the sample was drawn by probability or non-probability methods, and a margin of error may only be reported for probability samples. Seeing a margin of error on an opt-in online panel is therefore a red flag rather than a mark of quality. Second, AAPOR states that it can censure researchers and research companies that fail to disclose their methods, whether or not they are AAPOR members.

Common mistakes in polls

The mistakes that cost the most are not made during analysis but long before, and they cannot be repaired afterward. Once you know them, you start spotting them in other people's polls too.

All four share one trait: a number claims more than the data can carry. That is worse than having no number at all, because a wrong number gets passed on unchecked.

Polling mistake 1: the sample selects itself

Putting a poll link in your own Instagram story means questioning your followers and nobody else. The result then describes a group that differs sharply from the wider population in age, interests, and outlook. The fault is not the distribution channel but the sentence that later sits above the chart. Word the conclusion so that the group actually questioned appears in it.

Polling mistake 2: the question already contains the answer

“Do you agree that the cafeteria is too expensive?” is not a question but an invitation to agree. Leading questions shift results reliably in one direction, and they do so even when the respondent never notices. Neutrally worded, the same question reads: “How do you rate value for money in the cafeteria?” A pretest with five people finds phrasing like this every time.

Polling mistake 3: percentages without a base

“62 percent are in favor” sounds solid and is not, as long as the number of respondents is missing. With eight answers, 62 percent means five people, and a single different answer would have moved the figure by 12 percentage points. State the base alongside every percentage, and with small groups drop percentages altogether. Absolute numbers are more honest there.

Polling mistake 4: a correlation turns into a cause

If students who spend a lot of time in the library report better grades, that is a correlation and not yet an explanation. Perhaps the library helps, perhaps motivated people go there anyway, perhaps a third factor drives both. A poll records what occurs together, not what follows from what. Word results as a relationship unless your research design allows more.

Conclusion

A poll is quick to launch and just as quickly worthless when purpose, sample, and questions do not fit together. With other people's polls, check the organization, the respondents, the field dates, and the sampling method first. With your own, check whether your conclusion really only speaks about the people who answered. Do that and you will get usable results without needing large numbers.

Where to go next

• Want to understand the wider method and its levels of standardization? Survey: definition and types
• Not sure whether your study is qualitative or quantitative? Telling qualitative and quantitative apart
• Need the methodological groundwork for your thesis? Fundamentals of empirical research
• Want to know when a survey counts as representative? Creating a representative survey
• Ready to write the questionnaire? Creating a questionnaire

Want to run your own poll?

With empirio.ai, an online survey tool from Germany, you can set one up in a few steps. 50 responses a month are free, and the data stays on servers in the European Union, which matters as soon as respondents are based there.

Create a survey for free