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 the two numbers says anything about the country.
That is why the word is worth a closer look. A poll is a systematic questioning of a larger group of people who are all given 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.
- ESOMAR and WAPOR require eight disclosures with every published poll.
- British Polling Council members must publish full tables within two working days.
Create a survey for free
With empirio.ai you can create a modern online survey in minutes — with hosting in the EU.
- AI-built survey
- Adjust by drag & drop
- Real-time analysis
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 analysed statistically. The word carries a second, older meaning in British English, namely the act of voting itself, which is why polling day and polling station 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. Standardisation is not red tape, 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 dissertation that shows.
| Term | What it means | Example |
|---|---|---|
| Poll | opinion measured with few questions, often published | voting intention before a general election |
| Survey | the wider research instrument, often many questions | Labour Force Survey run by the ONS |
| Questionnaire | the tool carrying the individual questions | 15 questions across three pages |
| Data collection | the umbrella term for gathering evidence | surveys, observation or experiments |
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, whilst a survey can run to a hundred items and measure behaviour 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 club outing and asking about voting intention before a general 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.
| Type | Question it answers | Typical example |
|---|---|---|
| Opinion poll | What does the public think about an issue? | voting intention before a general election |
| Market research survey | How do product, price or advertising land? | testing new packaging |
| Academic survey | Does a hypothesis survive testing? | questionnaire in a dissertation |
| Employee survey | How satisfied are the staff? | annual staff survey at work |
| Customer survey | How satisfied are customers and why? | short feedback after a purchase |
| Private poll | What suits the group best? | picking a date for a club outing |
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 analyse. 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 fieldwork period, then analyse 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!
- Settle the research question. Write down in one sentence what you want to know afterwards and what you will do with the answer.
- 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.
- 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.
- Distribute and cap the fieldwork. Fix in advance how long the poll runs. An open-ended field period invites you to keep collecting until the hoped-for result appears.
- Analyse and interpret. Work out the figures, always state how many people answered and say plainly who the result applies to.
Create a survey for free
With empirio.ai you can create a modern online survey in minutes — with hosting in the EU.
- AI-built survey
- Adjust by drag & drop
- Real-time analysis
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 organisation 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 24 August 2026). For voting intention 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 Kingdom
In the United Kingdom the polling industry regulates disclosure itself, through a body set up for exactly that purpose. Members of the British Polling Council agree to its Rules of Disclosure, under which every published report has to name the client commissioning the survey, the dates of interviewing, the method used, the universe represented, the percentages the conclusions rest on, and the sample size with its geographical coverage (accessed 24 August 2026). Members also state in their write-ups that they are members of the BPC and abide by its rules, which gives you a quick first check.
Two BPC rules go beyond the international standard and are worth knowing. First, the full computer tables must appear on the research organisation's own website within two working days of publication, so a poll whose tables are nowhere to be found after two days is missing a required element. Second, voting intention polls must carry the Council's uncertainty statement, which puts the historical record plainly: there is a 9 in 10 chance that a party's true support lies within 4 points of the published estimate, and a 2 in 3 chance that it lies within 2 points.
Common mistakes in polls
The mistakes that cost the most are not made during analysis but long before, and they cannot be repaired afterwards. 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 canteen 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 canteen?” A pilot with five people finds phrasing like this every time.
Polling mistake 3: percentages without a base
“62 per cent are in favour” sounds solid and is not, as long as the number of respondents is missing. With eight answers, 62 per cent means five people, and a single different answer would have moved the figure by 12 percentage points. For your own number of cases you can calculate the margin of error. 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 marks, 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 organisation, the respondents, the fieldwork 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 standardisation? Survey: definition and types
- Not sure whether your study is qualitative or quantitative? Telling qualitative and quantitative apart
- Need the methodological groundwork for your dissertation? 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 EU hosting plus a data processing agreement are included in the free plan, which keeps UK GDPR paperwork short.