Marketing surveys can help organizations understand customer attitudes, awareness, satisfaction, preferences, behaviors, needs, purchase intent, brand perceptions, and other information that may not be available through transactional or behavioral data alone. The quality of the results, however, depends heavily on how the survey is designed.

The AAMA guide to designing a marketing survey provides a practical framework for creating surveys that are focused, understandable, and useful for decision-making. It is designed for marketing and advertising professionals, agencies, brands, students, academics, researchers, entrepreneurs, and organizations that need to collect structured information from customers, prospects, members, or other audiences.

A good survey should make it easy for respondents to understand what is being asked and easy for researchers to interpret the answers. Every question should have a clear purpose.

Start With the Research Objective

Before writing a single survey question, define what the organization needs to learn.

A survey should not begin with:

“What questions should we ask?”

It should begin with:

“What decision are we trying to make, and what information do we need in order to make it?”

Possible research objectives include:

• Measure brand awareness
• Understand customer satisfaction
• Identify purchase barriers
• Compare product preferences
• Measure advertising recall
• Understand media behavior
• Evaluate customer experience
• Assess interest in a new product
• Identify audience segments
• Measure consideration or purchase intent

A clear objective makes it easier to determine which questions belong in the survey and which do not.

Define the Target Population

The target population is the group the research is intended to represent.

Depending on the project, this might include:

• Current customers
• Former customers
• Prospective customers
• Category buyers
• Website visitors
• Event attendees
• Members
• Business decision-makers
• Residents of a particular market
• Users of a particular product

The target population should be defined before recruitment begins.

A survey of current customers cannot automatically answer questions about people who have never purchased. Likewise, a survey of highly engaged newsletter subscribers may not represent the broader market.

Define Who Qualifies to Participate

Eligibility criteria determine who should be included in the study.

Screening questions may be used to confirm characteristics such as:

• Age
• Geography
• Product usage
• Purchase history
• Professional role
• Industry
• Decision-making authority
• Category involvement
• Customer status

Screening should be limited to criteria that are genuinely necessary for the research.

Do not collect sensitive or unnecessary information simply because the survey platform allows it.

Determine the Survey Mode

Decide how respondents will complete the survey.

Common modes include:

• Online surveys
• Email surveys
• Website surveys
• Mobile surveys
• Telephone surveys
• In-person surveys
• Mail surveys
• Intercept surveys
• Panel surveys

The mode can influence who responds, how questions should be written, how long the survey can reasonably be, and what types of questions are practical.

An online survey, for example, may allow images, randomized questions, and skip logic, while a telephone survey requires wording that works when heard rather than read.

Keep the Survey Focused

Survey length should reflect the importance of the research and the level of respondent commitment.

A long survey does not automatically produce better research. It can increase abandonment, reduce attention, and encourage respondents to rush through later questions.

Before adding a question, ask:

• Does this question support the research objective?
• Will the answer influence a decision?
• Is this information available elsewhere?
• Do we need this information at this level of detail?
• Is the burden on the respondent justified?

If the answer is no, remove the question.

Write Questions in Plain Language

Questions should be easy to understand on the first reading.

Avoid:

• Unnecessary jargon
• Long sentences
• Complex grammar
• Vague wording
• Technical language without explanation
• Terms that different respondents may interpret differently

Instead of:

“How would you characterize your propensity to utilize subscription-based meal-preparation services?”

ask:

“How likely are you to use a meal-kit subscription service in the next six months?”

Clear language improves both the respondent experience and the quality of the data.

Ask One Thing at a Time

Avoid double-barreled questions that ask respondents to evaluate two or more things at once.

For example:

“How satisfied are you with our product quality and customer service?”

A respondent may be very satisfied with the product and dissatisfied with customer service, making the answer difficult to interpret.

Instead, ask separate questions:

“How satisfied are you with the quality of the product?”

“How satisfied are you with the customer service you received?”

Each question should measure one idea.

Avoid Leading Questions

A leading question suggests the answer the researcher expects or prefers.

For example:

“How much did our convenient new checkout process improve your shopping experience?”

This assumes that the process was convenient and that it improved the experience.

A more neutral version would be:

“How would you rate your experience with the new checkout process?”

Neutral wording reduces the risk of influencing the response.

Avoid Loaded Questions

Loaded questions contain assumptions that may not apply to every respondent.

For example:

“Why do you prefer our new packaging?”

This assumes that the respondent prefers the new packaging.

A better sequence might be:

“Which packaging do you prefer?”

Then, only for respondents who prefer the new version:

“What influenced your preference?”

Do not force respondents to accept an assumption in order to answer the question.

Avoid Absolutes

Words such as “always,” “never,” “every,” and “none” can create unnecessarily rigid questions unless the research truly requires an absolute answer.

For example:

“Do you always compare prices before making a purchase?”

may produce less useful information than:

“How often do you compare prices before making a purchase?”

The second version allows respondents to describe behavior more accurately.

Define the Time Period

Questions about behavior should usually specify a clear period.

Instead of:

“How often do you purchase coffee?”

consider:

“During the past 30 days, how many times have you purchased coffee from a coffee shop?”

A defined period reduces ambiguity and makes responses easier to compare.

The appropriate period depends on how frequently the behavior normally occurs.

Be Careful With Memory

Respondents may have difficulty accurately recalling behavior over long periods.

Questions such as:

“How many advertisements did you see last year?”

are unlikely to produce reliable answers.

When possible:

• Use shorter recall periods
• Ask about recent events
• Use ranges rather than exact counts
• Compare self-reported behavior with available behavioral data

Research design should recognize the limits of human memory.

Choose the Right Question Type

Different questions require different response formats.

Common survey question types include:

• Multiple choice
• Single-select
• Multi-select
• Rating scales
• Ranking
• Numeric entry
• Open-ended questions
• Matrix questions
• Yes or no
• Semantic differential scales

Choose the format that best matches the information being collected.

Do not use a complicated format when a simpler question would produce the same information.

Single-Select Questions

Single-select questions allow respondents to choose one answer.

They are appropriate when the response categories are mutually exclusive.

For example:

Which of the following best describes your current customer status?

• Current customer
• Former customer
• Prospective customer
• Never considered the product

Include an appropriate “Other” or “Not sure” option when the listed choices may not cover every reasonable response.

Multi-Select Questions

Multi-select questions allow respondents to select more than one answer.

For example:

Which of the following marketing channels did you use during the past 12 months? Select all that apply.

• Paid search
• Social media
• Email
• Direct mail
• Television
• Radio
• Events
• Other

Clearly indicate when respondents may select multiple answers.

Rating Scales

Rating scales measure the strength, quality, frequency, likelihood, satisfaction, or agreement associated with a response.

Examples may include:

• Very dissatisfied to very satisfied
• Very unlikely to very likely
• Strongly disagree to strongly agree
• Never to very often
• Poor to excellent

Use labels that clearly describe the meaning of the scale points.

Keep Scale Direction Consistent

When several rating questions appear together, keep the direction of the scale consistent whenever possible.

For example, if one question uses:

1 = Very Dissatisfied
5 = Very Satisfied

do not suddenly use:

1 = Very Likely
5 = Very Unlikely

without a strong reason.

Inconsistent scale direction increases the likelihood of respondent error.

Use Balanced Scales

A balanced scale provides comparable positive and negative options.

For example:

• Very dissatisfied
• Somewhat dissatisfied
• Neither satisfied nor dissatisfied
• Somewhat satisfied
• Very satisfied

Avoid scales that offer several positive options and only one negative option unless the research objective genuinely requires that structure.

Decide Whether a Neutral Option Is Appropriate

Some questions benefit from a neutral midpoint, while others require respondents to make a choice.

A neutral option can be useful when respondents may genuinely have no positive or negative view.

However, forcing respondents into a positive or negative category when they are genuinely neutral can distort the results.

The choice should reflect the research objective rather than a desire to produce stronger-looking findings.

Include “Not Applicable” When Needed

Respondents should not be forced to evaluate something they have never experienced.

For example:

“How satisfied are you with our customer support?”

should include an option such as:

“I have not used customer support”

if that situation is possible.

Otherwise, respondents may choose arbitrary answers simply to continue.

Use Ranking Carefully

Ranking questions ask respondents to place several options in order.

For example:

“Rank the following product features from most important to least important.”

Ranking can be useful when relative priority matters, but it becomes difficult when too many items are included.

If respondents must rank 15 or 20 options, the results may reflect fatigue more than meaningful preference.

Consider limiting the list or using another method.

Use Open-Ended Questions Strategically

Open-ended questions allow respondents to answer in their own words.

They are useful for discovering:

• Unexpected concerns
• Audience language
• Reasons behind ratings
• Unlisted needs
• Product feedback
• Customer experiences
• Suggestions

Examples include:

“What is the main reason you chose this product?”

“What could we do to improve your experience?”

Open-ended questions can produce rich information but require more effort from respondents and more analysis from researchers.

Use them where depth is valuable rather than after every closed-ended question.

Avoid Requiring Open-Ended Answers Without Reason

Making every open-text field mandatory can frustrate respondents and produce low-quality responses such as “N/A,” “none,” or random characters.

Require an open-ended response only when the information is essential.

Optional comment fields often produce better responses because participants can contribute when they genuinely have something to say.

Use Matrix Questions Sparingly

Matrix questions place several items under the same response scale.

For example:

How satisfied are you with the following?

Item	Very Dissatisfied	Dissatisfied	Neutral	Satisfied	Very Satisfied
Product Quality	○	○	○	○	○
Delivery	○	○	○	○	○
Customer Service	○	○	○	○	○

Matrices can save space but become difficult to complete on mobile devices and can encourage respondents to select the same answer repeatedly without careful consideration.

Use shorter matrices when possible.

Randomize When Order Could Influence Responses

The order of answer choices can influence what respondents select.

When there is no meaningful reason for one option to appear before another, consider randomizing the order.

Randomization may be useful for:

• Brand lists
• Product features
• Reasons for purchase
• Advertising concepts
• Preference options

Do not randomize answer choices that have a natural order, such as age ranges, frequency scales, or satisfaction scales.

Use Skip Logic

Skip logic directs respondents to different questions based on previous answers.

For example:

“Have you purchased from us during the past six months?”

If yes, the respondent may receive questions about the purchase experience.

If no, those questions can be skipped.

Good skip logic makes the survey shorter and more relevant.

Poorly configured logic can create confusing or impossible survey paths, so test every branch before launch.

Put Screening Questions Early

Eligibility questions should generally appear near the beginning of the survey.

There is little value in asking someone to complete several minutes of questions before discovering that they do not qualify for the study.

Keep screening concise and avoid revealing the exact qualification criteria when doing so could encourage people to manipulate their answers.

Begin With Easy Questions

The first substantive questions should be easy to understand and relevant to the survey topic.

Avoid beginning with:

• Sensitive questions
• Long open-ended questions
• Complex ranking tasks
• Detailed demographic questions
• Difficult recall questions

Starting with straightforward questions helps respondents become comfortable with the survey.

Group Related Questions

Organize questions into logical sections.

For example:

1. Product Usage
2. Customer Experience
3. Brand Perceptions
4. Purchase Intent
5. Demographics

A logical sequence reduces cognitive effort and makes the survey easier to follow.

Use Transitions Between Sections

When the topic changes, provide a short transition.

For example:

“The next few questions are about your experience using the product.”

This gives respondents context and prevents abrupt shifts between unrelated subjects.

Put Sensitive Questions Later

Questions involving income, age, health, political beliefs, personal identity, or other potentially sensitive subjects should generally appear only when they are necessary for the research.

When they are required:

• Explain why the information is needed when appropriate
• Provide reasonable response ranges
• Include a prefer-not-to-answer option when appropriate
• Place them later in the survey
• Follow applicable privacy and research requirements

Do not collect sensitive information without a clear reason.

Ask Demographic Questions Only When Useful

Demographic questions are often included automatically even when they provide little value.

Before asking about age, income, education, household structure, gender, occupation, or other characteristics, determine whether the information will actually be used in the analysis.

If the answer will not influence interpretation or decision-making, consider removing the question.

Every additional question adds burden.

Avoid Collecting Personally Identifiable Information Without Need

Surveys should not collect names, email addresses, phone numbers, account numbers, or other identifying information unless it is necessary for the research process.

If contact information is being collected for an incentive, follow-up, or recruitment purpose, consider whether it can be stored separately from survey responses.

Researchers should use appropriate privacy, consent, security, and data-retention practices.

Write Clear Instructions

Do not assume respondents will automatically understand how to complete every question.

Instructions might include:

• Select one
• Select all that apply
• Rank your top three
• Please enter a number
• Think about your most recent purchase
• Answer based on the past 30 days

Instructions should be short and placed where respondents need them.

Avoid Excessive Required Questions

Required questions can help prevent missing data, but forcing responses can also lead people to abandon the survey or provide inaccurate answers.

Make a question required when the information is essential to the research or survey logic.

Allow respondents to skip questions when a forced answer would be inappropriate.

Design for Mobile Devices

Many respondents will complete online surveys on a phone.

Mobile-friendly surveys should avoid:

• Extremely wide matrices
• Tiny response controls
• Long paragraphs
• Excessive scrolling within individual questions
• Complicated drag-and-drop activities
• Large numbers of answer choices on one screen

Test the survey on actual mobile devices before launch.

Keep the Visual Design Simple

The survey interface should support comprehension rather than distract from the questions.

Use:

• Readable typography
• Clear spacing
• Strong contrast
• Consistent buttons
• Simple progress indicators
• Clear section breaks

Avoid decorative elements that make the survey look promotional or influence responses.

Be Careful With Branding

Branding can be appropriate when respondents already know which organization is conducting the research.

However, prominent branding may influence responses when the research is intended to measure awareness, preference, perception, or reactions without revealing the sponsor.

Decide whether sponsor identification should appear based on the research objective and ethical requirements.

Use Progress Indicators Carefully

A progress indicator can help respondents understand how much of the survey remains.

It is most useful when the survey has a predictable structure.

Complex skip logic can make progress estimates inaccurate, so make sure the indicator does not tell respondents they are 90% complete when substantial sections remain.

Estimate Survey Length Realistically

If respondents are told that a survey takes five minutes, it should generally take about five minutes.

Underestimating survey length can frustrate participants and increase abandonment.

Test completion time with several people before launch and account for respondents who read more slowly.

Consider Incentives

Incentives can improve participation in some studies, particularly when the audience is difficult to reach or the survey requires significant time.

Incentives might include:

• Gift cards
• Discounts
• Prize drawings
• Account credits
• Donations
• Research-panel compensation

The incentive should not be so large that it creates inappropriate pressure to participate.

Researchers should also consider whether incentives could attract respondents who are more interested in the reward than in providing thoughtful responses.

Pretest the Survey

Always test the survey before full launch.

A pretest can reveal:

• Confusing wording
• Missing answer choices
• Broken skip logic
• Repetitive questions
• Technical problems
• Mobile issues
• Excessive length
• Ambiguous instructions
• Unexpected interpretations

People involved in writing the survey may overlook problems because they already understand what each question is intended to mean.

Test with people who were not part of the drafting process whenever possible.

Conduct Cognitive Testing When Appropriate

Cognitive testing asks participants to explain how they interpret and answer survey questions.

This can reveal whether respondents understand a question in the way the researcher intended.

For example, if asked:

“How often do you shop online?”

different respondents may interpret “shop” as browsing, purchasing, or both.

Cognitive testing can identify these ambiguities before data collection begins.

Determine the Sample Size

The appropriate sample size depends on the research objective, population, sampling method, desired precision, expected response rate, subgroup analysis, and available resources.

A larger sample can improve precision under appropriate sampling conditions, but it does not correct poor recruitment or biased sampling.

A carefully selected sample of 500 people may provide more useful information than thousands of responses from people who do not represent the target population.

Consider Response Rate

Response rate measures the percentage of eligible people invited to participate who complete or respond to the survey.

Formula: Response Rate = Completed or Valid Responses ÷ Eligible Invitations × 100

A low response rate does not automatically invalidate a survey, but it raises questions about whether respondents differ systematically from people who did not participate.

Researchers should consider potential nonresponse bias when interpreting the findings.

Monitor Data Quality

Survey platforms can collect responses from inattentive, automated, duplicate, or otherwise low-quality participants.

Depending on the study, quality checks may include:

• Completion time
• Duplicate responses
• Inconsistent answers
• Straight-lining
• Nonsensical open-text responses
• Failed attention checks
• Geographic inconsistencies
• Screening inconsistencies

Quality controls should be designed carefully so legitimate respondents are not removed simply because their answers are unusual.

Clean the Data Before Analysis

Before analyzing results, review the dataset for problems.

Possible issues include:

• Incomplete responses
• Duplicates
• Invalid responses
• Coding errors
• Missing values
• Incorrect skip patterns
• Outliers
• Inconsistent formatting

Document any exclusions or adjustments made during data cleaning.

Weighting

Survey weighting adjusts the influence of different responses so the final sample more closely reflects known characteristics of the target population.

For example, weighting may be used when one demographic group is overrepresented and another is underrepresented.

Weighting can improve some estimates, but it cannot fully repair a fundamentally poor sample. Researchers should document when and how weighting has been applied.

Analyze Beyond the Overall Average

Overall results may conceal important differences among audience groups.

Where sample size and research design support it, researchers may compare findings by:

• Customer status
• Geography
• Age group
• Product usage
• Purchase frequency
• Industry
• Business size
• Awareness
• Customer value
• Other strategically relevant segments

Subgroup analysis should be driven by the research objective rather than searching for interesting differences after the fact.

Avoid Overinterpreting Small Differences

Two percentages being different does not automatically mean the difference is meaningful.

For example, 47% versus 49% may simply reflect normal sampling variation.

Researchers should consider sample size, confidence intervals, statistical significance where appropriate, effect size, and practical importance before treating small differences as meaningful.

Separate Statistical Significance From Practical Significance

A difference can be statistically significant while being too small to matter in practice. A large study may detect very small differences that have little business importance.

The reverse can also happen. A difference may appear practically important but fail to reach statistical significance because the sample is too small.

Good analysis considers both statistical evidence and business relevance.

Report the Question Wording

When presenting survey results, preserve the exact wording of important questions or make the questionnaire available when appropriate.

A finding such as:

“72% of customers are satisfied”

is easier to evaluate when readers can see how satisfaction was actually measured.

Question wording provides important context.

Report the Sample

Survey reports should explain who participated.

Useful information may include:

• Target population
• Sample size
• Recruitment method
• Geographic scope
• Field dates
• Eligibility criteria
• Response rate
• Weighting
• Relevant participant characteristics

Readers need this context to understand what population the findings can reasonably represent.

Report Limitations

Every survey has limitations.

Potential limitations may include:

• Convenience sampling
• Small sample size
• Low response rate
• Self-selection
• Self-reported behavior
• Limited geography
• Recall problems
• Missing groups
• Measurement error

These limitations should be acknowledged rather than hidden.

Clear limitations make the findings easier to interpret responsibly.

Do Not Treat Survey Responses as Behavior

What people say they do and what they actually do may differ.

A respondent may report that price is their primary purchase consideration while their observed behavior suggests convenience plays a larger role.

Surveys are especially useful for understanding attitudes, perceptions, stated preferences, experiences, and reported behavior. Where possible, combine survey findings with sales, analytics, CRM, observational, or experimental data.

Avoid Turning Survey Results Into Causal Claims

A survey may show that satisfied customers are also more likely to repurchase, but that relationship does not necessarily prove that satisfaction caused the repurchase behavior.

Survey research can identify associations and attitudes. Causal claims generally require stronger designs, such as controlled experiments or carefully constructed longitudinal research.

Use Open-Ended Responses to Add Context

Open-ended responses can help explain why quantitative patterns exist.

For example, if satisfaction declines, customer comments may reveal recurring concerns about delivery times, product quality, pricing, or customer service.

Qualitative responses should not be treated as population estimates, but they can provide valuable context and identify issues worth investigating further.

Build the Survey Around the Decision

The finished survey should remain connected to the decision that motivated the research.

Before launch, review every question and ask:

“What will we do differently depending on how people answer this?”

Not every question needs to trigger an immediate action, but the survey should produce information that helps reduce uncertainty.

Collecting data because it might be interesting later can make surveys longer without making them more useful.

Recommended Marketing Survey Structure

A practical survey may follow this sequence:

1. Introduction
2. Consent or Required Research Information
3. Screening Questions
4. General Behavior or Experience
5. Primary Research Questions
6. Brand, Product, or Campaign Evaluation
7. Purchase or Behavioral Intent
8. Open-Ended Feedback
9. Relevant Demographics or Classification Questions
10. Closing Message

Not every survey requires every section. The structure should reflect the specific research objective.

Pre-Launch Survey Checklist

Before launching a marketing survey, confirm:

• The research objective is clear
• The target population is defined
• Eligibility criteria are appropriate
• Every question supports the research objective
• Questions use clear language
• Double-barreled questions have been removed
• Leading and loaded wording has been removed
• Time periods are clearly defined
• Response options are complete and balanced
• Scales are consistent
• “Not applicable” options are included where necessary
• Skip logic works
• Sensitive questions are justified
• Unnecessary identifying information has been removed
• The survey works on mobile devices
• Estimated completion time is accurate
• The survey has been pretested
• Data-quality procedures are defined
• The analysis plan is established

A short review before launch can prevent problems that cannot be corrected after responses have already been collected.

Good Survey Design Makes the Research Easier to Trust

Survey software makes it easy to distribute questions, but technology does not determine whether those questions produce useful evidence.

Good survey design requires discipline. Researchers must decide what they actually need to know, ask questions neutrally, respect respondents’ time, select an appropriate sample, and report limitations clearly.

When those decisions are made carefully, surveys can provide valuable evidence about how audiences think, feel, evaluate, and describe their experiences.

Related AAMA Resources

Continue developing your research with the Marketing Research Methods Guide, A/B Testing Guide for Marketers, Marketing Metrics & KPI Reference, Campaign Measurement Framework, Marketing & Advertising Glossary, and Common Marketing Formulas. These resources provide additional guidance for selecting research methods, evaluating evidence, designing measurement systems, and interpreting marketing results.

The AAMA Resource Library will continue expanding with survey worksheets, research templates, measurement tools, and professional resources designed to support responsible and useful marketing research.