A survey, often called a questionnaire, is a structured set of standardized questions administered to a defined sample of respondents in order to measure attitudes, behaviors, preferences or characteristics of a larger population. In market research, surveys are the main quantitative method of primary research: they turn individual answers into numbers that can be compared across segments, markets and time.
The value of a survey rests on two foundations: who is asked and how they are asked. A well-written questionnaire sent to the wrong sample produces confident but misleading results, and a perfect sample cannot rescue biased questions.
How a market research survey works
A survey project moves from an objective to a sampling plan, a questionnaire, fieldwork, data cleaning and analysis. Each stage introduces its own error, which is why professional researchers think in terms of total survey error rather than sampling error alone.
- Define the decision the survey must inform, for example pricing, positioning or product priorities.
- Specify the target population and the sampling frame that represents it.
- Draft, review and pretest the questionnaire on a small group.
- Field the survey and monitor completion rates and data quality.
- Clean responses, apply weighting if needed and analyze the results.
Pretesting is often skipped and often regretted. Cognitive interviews with five to ten respondents routinely reveal ambiguous wording, missing answer options and questions that people interpret differently from the author’s intent.
Survey modes and sampling
The survey mode is the channel through which questions reach respondents, and it shapes cost, speed, coverage and the type of bias that is likely. Online panels dominate commercial market research because they are fast and inexpensive, but they rely on opt-in samples that are not random.
| Mode | Speed and cost | Main strengths | Main risks |
|---|---|---|---|
| Online panel | Fast, low cost | Large samples, quotas, multimedia stimuli | Professional respondents, inattentive answers |
| Customer email survey | Fast, very low cost | Reaches known customers with purchase data | Low response, bias toward engaged customers |
| Telephone (CATI) | Slower, moderate to high cost | Interviewer can clarify and probe | Declining response, social desirability |
| Face-to-face (CAPI) | Slow, high cost | Long interviews, hard-to-reach groups | Interviewer effects, cost per complete |
| In-product or intercept | Immediate, low cost | Captures experience in the moment | Short questionnaires only, context bias |
Sample size determines precision, not accuracy. For a simple proportion near 50 percent at 95 percent confidence, about 400 completes give a margin of error near ±5 points and about 1,000 completes near ±3 points. These figures assume random sampling; with opt-in panels, the margin of error is only an approximation, and coverage and nonresponse bias can be larger than sampling error.
Question types and measurement scales
Survey questions are built from a small set of question types, each suited to a different measurement goal. The choice affects which statistics are valid in analysis.
| Question type | Example use | Analysis it supports |
|---|---|---|
| Single choice | Main brand used last month | Shares, cross-tabulation |
| Multiple choice | All channels where a product was seen | Reach, overlap |
| Likert scale | Agreement from 1 to 5 or 1 to 7 | Distributions, top-two-box scores |
| Numeric rating | Likelihood to recommend from 0 to 10 | NPS, means, segments |
| Ranking or MaxDiff | Most and least important features | Relative importance scores |
| Open-ended | Reason for a score | Coding, themes, text analysis |
Standardized metrics such as NPS and CSAT are themselves single survey questions with fixed scales. Their strength is comparability over time; their weakness is that a single number rarely explains why customers feel as they do, so researchers pair them with a follow-up open question.
Surveys versus other market research methods
A survey answers how many and how much, while qualitative methods answer why and how. Choosing between them depends on whether the team already knows which questions to ask.
| Method | Best for | Limitation |
|---|---|---|
| Survey | Measuring prevalence and differences between groups | Cannot explore issues the questionnaire omits |
| Focus group | Exploring language, perceptions, reactions | Small, non-representative, group influence |
| Customer interview | Deep understanding of individual decisions | Time-consuming, not statistically projectable |
| Conjoint analysis | Quantifying trade-offs between attributes and price | Complex design, needs larger samples |
| Secondary research | Using existing data and reports | May not match the exact question |
A common and effective sequence is qualitative first, quantitative second: interviews or focus groups uncover the vocabulary and options, and the survey then measures how widespread each one is.
Common survey biases and how to avoid them
Survey results are vulnerable to systematic bias that larger samples do not fix. The most frequent sources are listed below with practical remedies.
| Bias | How it arises | Remedy |
|---|---|---|
| Leading questions | Wording suggests a preferred answer | Neutral wording, balanced scales |
| Acquiescence | Respondents tend to agree | Mix positive and negative statements |
| Order effects | Earlier items influence later ones | Randomize answer and question order |
| Social desirability | Respondents give acceptable answers | Self-completion, indirect questions |
| Nonresponse bias | Those who answer differ from those who do not | Reminders, weighting, short questionnaires |
| Satisficing | Fatigue leads to careless answers | Shorter surveys, attention checks |
Length is a recurring trade-off. Completion quality usually declines noticeably once an online questionnaire exceeds about 15 to 20 minutes, so every question should earn its place by linking to a decision. Double-barreled questions, which ask about two things at once, are another classic error that makes answers impossible to interpret.
In plain words
A survey is a list of the same questions asked to many people, so their answers can be counted and compared. Instead of guessing what customers think, a company asks a carefully chosen group and uses the results to estimate what the wider market thinks.
It works a bit like tasting a spoonful of soup to judge the whole pot. The spoonful must be well stirred and representative, and the question must be fair, otherwise the conclusion about the whole pot will be wrong.
A real-world example
A regional coffee chain is considering a subscription plan and wants to know whether customers would pay a monthly fee. After six customer interviews reveal that convenience matters more than discounts, the insights team writes a ten-minute online survey and sends it to a panel sample of 800 coffee buyers in its three cities, with quotas for age and visit frequency.
The questionnaire includes a MaxDiff exercise on plan features and a price question tested at three monthly levels. Results show that mobile pre-ordering ranks above free drinks, and that interest drops sharply above the middle price level. The chain launches a pilot at that middle price with pre-ordering as the central benefit, and repeats a shorter follow-up survey three months later to compare satisfaction among subscribers and regular customers.