Skip to content

guide

Audience survey questions: what to ask and how

The order, wording and answer options decide what a survey can prove. Most of the damage is done before the first interview.

Takwimu editorial team, Editorial team9 min read

Audience survey questions should ask what people did before what they think, one thing at a time, in the order a person remembers a day: media use yesterday first, then programmes, then opinions, then who they are.

Key points:

  • Ask about behaviour before opinion, and about all stations before your own.
  • Put one idea in each question and keep opinion scales to four or five choices.
  • Wording can move a result by tens of points, so fix it and keep it across waves.
  • Use closed questions to count and open questions to find reasons.
  • Pretest in every language you will field.

What should an audience survey ask?

An audience survey needs five blocks of questions, in this order. Leave one out and you lose the use of the others.

Block What it gives you Example
Screening The right people Age, county, works in media
Use yesterday Reach, share, dayparts Did you listen to radio yesterday?
Habit Weekly reach, frequency On how many of the last seven days?
Opinion Appreciation, reasons How would you rate this programme?
Profile Who the audience is Education, work, household

What order should the questions go in?

Move from the general to the specific, and from fact to feeling. Pew Research Center warns that earlier questions can unintentionally provide context for the questions that follow.

For a station, the rule is plain. Do not name your station until the respondent has named theirs. If the first five questions are about your breakfast show, your station will be front of mind for every answer after that.

Put age, sex and county at the start only if you need them to screen. Other personal questions go at the end, once trust exists.

How do you word a question so it gives one answer?

Ask only one question at a time. Pew calls a question that asks about two concepts double-barrelled, and says such questions are hard to answer and often hard to interpret.

This is the commonest fault in station questionnaires:

  • Weak: "How satisfied are you with the news and music on this station?"
  • Stronger: "How satisfied are you with the news on this station?" Then the same for music.

Name the period too. "Do you listen to radio?" has no time in it. "Did you listen to any radio yesterday, on any device?" has one.

How much can wording move an answer?

Wording can move a result by more than any real change in audience. In a January 2003 survey, Pew found that 68% favoured military action in Iraq, but only 43% did when the question added "even if it meant that U.S. forces might suffer thousands of casualties".

Your questions carry the same risk on a smaller scale. "Which station do you listen to?" and "which stations did you listen to yesterday?" measure different things: the first gets loyalty and the second gets behaviour.

So freeze wording once it works. A rewrite between waves can look like a rise or a collapse.

Why do people agree with statements?

People tend to agree with what is put to them. Pew calls this acquiescence bias, reports that it is stronger among less educated and less informed respondents, and recommends a choice between alternative statements.

  • Weak: "The presenter talks too much. Do you agree?"
  • Stronger: "Would you say the presenter talks too much, too little or about the right amount?"

The stronger form has no easy yes in it. It gives you a number that can fall as well as rise.

How does wanting to look good bend answers?

People adjust answers to look good. Pew notes that respondents understate some behaviour and overstate things such as church attendance and charitable giving.

The mode matters here too. Holbrook, Green and Krosnick found telephone respondents more likely than face-to-face respondents to present themselves in socially desirable ways.

In media this shows up as more news and fewer late-night talk shows than people consume. Soften the question: "Some people like these programmes and some do not." Ask for behaviour on a specific day, and keep your interviewers neutral.

How many answer options should you give?

Keep opinion scales short. Pew advises four or perhaps five choices at most, with more allowed for facts such as religious affiliation.

Order matters as well. Pew reports that telephone respondents more often choose items heard later in a list, and that people completing a survey themselves tend to choose items near the top. Rotate lists of stations, and keep ordered scales in order.

Always offer "don't know" or "can't remember" on recall questions. A forced guess is worse than a gap.

Should you use open or closed questions?

Use closed questions to count and open ones to learn. A list changes what people say. In a 2008 Pew poll, 58% chose the economy when it was offered, but only 35% volunteered it in the open version, and fewer than one in ten gave an answer outside the five read out.

That gap is the subject of our page on prompted recall. For a station, the practical rule is to ask open first and prompt second.

Open questions earn their cost on reasons. "What made you switch away?" gives words a closed list would never contain, and those words can be coded into why audiences leave.

What questions measure appreciation?

Ask about a named programme, and ask only people who heard it. A score from someone who has never heard the programme is noise.

Use a simple scale and the same wording in each wave, then add one open question on what they would change. Our page on the appreciation index shows how to turn this into a figure you can track.

What should a readership survey ask?

A readership survey uses the same logic with a shorter day. Ask whether the person read any newspaper yesterday, which title, in print or online, and whether the copy was their own or shared.

Shared copies are why readership and newspaper circulation count different things. Add questions on sections read if you sell advertising by section.

Print needs a large enough sample. The Communications Authority's report for Q4 of 2025/26 put seven-day newspaper readership at 17%, and 84% said they rarely or never read a newspaper. Reader survey questions about one title will reach few people in a general sample. See newspaper readership in Kenya.

What should a viewership survey ask?

A viewership survey must ask where viewing happened, because much of it is shared. The Communications Authority reports television as the preferred medium in social settings such as bars, restaurants and hotels.

Ask whether the person watched television yesterday, on what screen, where and with whom, then which channels and at what times. Keep channel lists to what a person can receive in that county. Shorter lists get cleaner answers. The pattern is covered in TV viewership in Kenya.

What target audience questions do advertisers want?

Advertisers want to know who your audience is in their terms. Age, sex, county, urban or rural, work and education are the base. Everything else is optional.

Add category questions only if a buyer needs them, such as who decides on groceries or who owns a motorbike. Each extra item costs interview time and adds personal data, so include it only when it serves a stated decision. Our guide to the audience profile shows what to report.

How do you translate a questionnaire?

Translate the meaning and test the result. Have one person translate into Kiswahili or the local language, and a second translate back into English without seeing the original, then compare.

Check spoken forms. Stations have nicknames, and listeners may not use the name on the licence. Time words differ too, because a question about "morning" means little if the blocks are read out in different ways.

How do you test questions before fieldwork?

Pew calls pretesting an essential step, especially for questions used for the first time. Sit down with ten people from your population and ask them to explain each question in their own words.

Then run a small pilot in the field and time it. Cut anything that nobody understands or that gives the same answer from everyone. Every minute you remove helps the data, as interviews that run long lose care towards the end. Shorter is safer.

Where TI fits

A questionnaire should be judged by the decision it lets you make. Takwimu Intelligence (TI) carries two: a 43-item cross-media day-after recall questionnaire and a 7-item programme panel on appreciation and what people would change. Open answers are coded against two frames, why an audience leaves and what holds an audience.

To see it on your own station or title, book a walk-through.

Takwimu Intelligence (TI) is audience measurement software for Kenyan media.

Sources

  1. 1

    Writing Survey Questions · Pew Research CenterOpen and closed questions, wording, order, acquiescence and pretesting. Opened 5 November 2026.

  2. 2

    Telephone versus Face-to-Face Interviewing of National Probability Samples with Long Questionnaires · Holbrook, Green and Krosnick, Public Opinion Quarterly 67(1), 2003Satisficing and social desirability by mode across three national experiments. Opened 5 November 2026.

  3. 3

    Audience Measurement and Industry Trends Report, August 2026 · Communications Authority of Kenya, August 2026Seven-day access, daily use, place of consumption, radio genres and languages, and the evening advertising window for Q4 2025/26. Opened 5 November 2026.

More from the library