How can you avoid bias in your questionnaires?
November 30, 2023
A well-constructed questionnaire is one that will enable you to gather the most authentic, reliable and realistic responses from your respondents. It is therefore extremely important that you eliminate any risk of biasing your respondents' answers.
💡 A bias is an element that systematically disturbs the authenticity of the interaction (direct or indirect) between the interviewee and the person seeking information.
Here are a few simple and effective ways to avoid bias in your surveys:
- Avoid yes/no questions
- Don't ask leading questions
- Pay attention to the order of questions and answers
- Be as anonymous as possible
- Clarify your language
- Eliminate assumptions
- Sometimes prefer open-ended questions
Avoid yes/no questions
Try to avoid yes/no questions. Respondents tend to want to say yes - to appear positive and ambitious. This is known as acquiescence bias.
Turn your questions around. For example, for a concept test, instead of "Do you find it essential to have beautiful packaging when you buy cookies?", ask instead: "Which of the following do you find essential when you buy cookies?".
Don't ask leading questions
Beware of leading questions, which encourage respondents to answer the way you want them to. This can lead to confirmation bias. You can remedy this by avoiding unnecessary adjectives in your questionnaire. For example, don't say "What do you think of [the brand's] delicious chocolate?". Instead, say "What do you think of [the brand's] chocolate?
And don't forget that answers to scale questions can also be suggestive. Make sure your answers to scale questions include as many positive options as negative ones, and certainly not just positive ones.
Pay attention to order
The way you order your questions and answers has an effect on how people react. This is called order bias.
In the questions
Order bias can occur when a question you've asked - or an item you've presented - has an impact on how your respondents perceive the rest of the survey. For example, if you present a series of ads, all subsequent ads will probably be compared to the previous ones in the respondent's mind, even subconsciously. Here's what you can do to avoid order bias in questions:
- Don't combine the answers to the questions you're about to ask with those that come before. Think about the order in which you ask the questions, to ensure that you move from the most general at the beginning to the most specific towards the end of the survey.
- Whenever possible, ask questions at random. The Episto platform integrates this function, allowing you to mix the order of questions in your questionnaire.
In the answers
The answers given can also be influenced by the order in which you present them. People are more likely to choose the answer(s) at the top of the list. If this happens, you won't have an accurate representation of your audience's opinions. Episto's programming platform will allow you to randomize response options. It's a good idea to do this for all answers to all questions, except when using scaled answers.
Be as anonymous as possible
Don't give away your (brand) identity! If you do, you risk causing a sponsorship bias, i.e. respondents will feel obliged to give positive answers about the brand asking the questions. It also prevents respondents from letting their perceptions of your brand cloud their answers.
You generally want to know what consumers think of your brand, but sometimes you want to know what people think independently of your image. For example, if you're conducting research into new product development, you want to know what consumers think without your brand's image - whether positive or negative - influencing the answers given.
During recruitment campaigns on social networks, Episto is systematically careful not to quote or mention the brand to avoid this sponsorship bias.
Clarify your language
Avoid confusing respondents or leaving room for interpretation in answers by objectifying your language. For example, instead of using the terms "recently", "often" or "rarely", specify the time period as "daily", "weekly" or "once a year". Make sure you know how your ideal customer speaks, which will help you adapt your language.
There are 6 levels of French language proficiency, from A1 (beginner level) to C2 (perfectly correct expression on all subjects). Clear language corresponds to level B1, which is understood and mastered by 95% of the population. You should therefore use this level of language in your questionnaires to ensure that all your respondents understand you. Don't forget to limit the use of anglicisms, and to explain any technical terms you may need to use in your questionnaire.
Eliminate assumptions
Avoid making assumptions that force respondents to answer falsely. You can easily remedy this by asking yourself whether or not you should include the options "None", "Does not apply" and "Other" in your answer list. For example, the question "What type of laptop do you own?" assumes that the respondent owns a laptop, so you need to ensure that they can answer truthfully by adding the "None" or "Does not apply" options.
Instead of "What did you like about your last vacation?", rephrase the question in two separate parts: "What did you think of your last vacation?" "Very pleasant", "Quite pleasant", etc. Don't forget to add an option for the answer "I didn't take a vacation [in the past year]", then follow this up with an open-ended question like "Why do you think that?". With Episto, it's even possible to customize the title of a question according to the answer given to the previous question.
Sometimes prefer open-ended questions
Favour open-ended questions if you're not sure you can give an exhaustive list of answers. In this case, to avoid any bias, you're better off analyzing written answers rather than forcing answers into an exclusive choice.
For example, when you're offering brand names from a defined sector, it's not complicated to provide respondents with a simple, exhaustive list of brand names. On the other hand, when it comes to reasons why respondents wear make-up or go on vacation, some answers may be unexpected - but no less interesting!
Thanks to the intelligent features of the Episto platform, you'll find it easy to analyze these open-ended responses and take unexpected ones into account.
Would you like to design a questionnaire that's adapted to social network research and that really engages your respondents? Read this article to discover Episto's tips.