5 Tips for Writing Survey Questions that Don't Yield Statistical Garbage
Thursday is guest post day here at Duct Tape Marketing and today’s guest is from Josh Pigford – Enjoy!
While the end result of a survey might make the conclusions look cut and dry, there are many ways that data can be manipulated or misrepresented to change the truth. However, the sloppiest (and probably most common) method of fumbling the truth is when data is simply misunderstood.
Even if your survey questions sound great, if they’re not credible they won’t produce valid results. In case you don’t know the difference between discriminant and regression analyses (really, who does?), this post will cover five tips to make sure that your sweet survey doesn’t turn into a statistical bummer.
1. Think First, Ask Second
Think broadly about why you’re creating a survey. What are you really trying to figure out? When you have a clear idea, make a list detailing the kind of information that you’re setting out to look for. Now you can begin to write your questions, always keeping in mind that they must match your original informational targets.
If you realize after collecting all your responses that your questions are actually asking something rather different from what you originally intended, then your data will also be telling something rather different from what you intended. You’ll have to settle for either “different,” or a big lie.
2. Ratings vs. Rankings
Depending on the kind of information you’re looking for, you can either ask your respondents to rate or rank several items in a list. However, it’s important to realize the difference between these two types of questions.
A ranking will only tell you which items are more or less preferred relative to each other, but you won’t actually know from a ranking if a respondent likes or dislikes any items. For this latter purpose, you must use a rating question.
3. The Multiple Choice Golden Rule
Most surveys depend heavily on multiple choice questions since prepping for standardized tests have wiped out this country’s ability to formulate an original answer. Or we’re too lazy. Either way, the possible responses to any multiple choice question must be mutually exclusive. This means that no two answers could equally serve as appropriate responses.
Not only do non-exclusive answers annoy people (you’ve made them think too hard!), but they’ll make accurately analyzing your data nearly impossible. If someone could choose one of two answers and feel good about either response, you won’t be able to determine the respondent’s actual preference.
4. Surveys Are Not Like Airplane Exit Rows
More “legroom” in your question doesn’t make it better. You wrote your survey with a specific purpose, so make sure your questions are direct without giving your respondents too much leeway in answering.
If you want to know how to make your company’s logo look more cutting edge, make sure you specify your desire for responses regarding the logo’s impact, not simply ask about the company in general.
5. Don’t Get Too Excited in One Question
Surveys are awesome, we know, but that’s no excuse for asking more than one question at a time. Each question you ask needs to be aimed at collecting one unique point of information, or else you’ll end up skewing your data by mixing results that should be separated.
It’s even possible to get so excited that in a fit of survey-exuberance you accidentally put two contradictory questions together. That’s not only pretty weird, but it will also void the results for that part of the survey.
Statistics and the wild field of data analysis include another laundry list of Dos and Don’ts, but hopefully these five tips can get you started on creating surveys that also produce credible results.
If you’re ever in doubt about how a question comes off, ask some friends to test it out. If you don’t have any friends, well then you’ve got bigger problems than that tricky survey question.
Josh Pigford is co-founder + CEO of PopSurvey, where they’re building online survey software to try and make the survey industry a little less coma-inducing. They’ve got a huge collection of survey templates to help you get started with the click of a button!