Thursday, September 03, 2009

We All Love Surveys

| |

Except when they have nothing to do with our interests and seem just plain boring. However, since qualitative data (from focus groups) only work well to give insight to why people do what they do, you need an effective method to define your audience. Surveys, baby, are your best bet.

Before scheduling and writing, you must keep in mind that timing is important, as is knowing what you're trying to accomplish. "To understand if our user base has changed since last year, and if so, how; and to understand which features they find most attractive." That's a good start.

When writing a survey, you need to focus on two things: be descriptive (to profile the audience) and be explanatory (to discover their beliefs). There are three sub-categories following that; Characteristic (demographic, technological), Behavioral (web use, usage, etc.), and Attitudal (satisfaction, preference). Make sure these questions are relevant, specific, and clear, as people have a tendency to always give you answers - whether reliable or not.

The types of questions you can ask vary. A close-ended (single-answer multiple choice) questions must be exhaustively selective. Checklists give a measure of reader information desires. Likert scales (Very interesting to very uninteresting) can make answers more relevant to more people.

Fielding the survey is the hardest part. You'll need to think about choosing a proper sample size, but also considering the possibility of haphazard responses, sampling bias (hitting outside your target mark), and a flurry of other biases (timing, expectation, and invitation) that may cause less than accurate responses.

When presenting the survey, you need to emphasize importance, hold privacy ridiculously high, and always present a desirable reward. People don't work for nothing, you know. Pace the survey well by throwing in thematically relevant questions in the middle, then end it with either an open-ended (free response) question or a demographic survey.

Report it. Already? Yeah. Tabulate (count) and get back into statistics class mode because that's what you shall be doing with any apparent bell curves or measurement errors. Draw conclusions without falling into the confusion of correlation and causation. Remember, practice makes perfect - it's just that this time, you've got more studs on your belt to get started.


Post a Comment

Popular Posts