I spent a substantial amount of time designing the first survey for the WoW Learning project. I was initially unsure what data would be useful, but I knew I wanted enough data to make statements about particular groups of people: men were more social, most women created tank characters initially, millennials were using WoW more for learning, etc. That resulted in the three-part design of the survey: in-game demographic data, the essay question about play motivations, and real-world demographic details. Privacy was important and encouraging people to complete was also important. It was reasoned that in-game demographics were details people would not be as sensitive about, so they were asked for first and real-world demographic details left until the end. It was also recognized that many people would not complete the essay section. Putting the in-game demographic details first meant that certain types of data could be collected that could also be used for other purposes, e.g. answering whether women initially choose healer characters. After use by a small test audience, the survey was modified to include sample answers or explanations of how answers should be calculated.
Where possible and sensible, permissible option lists were used to help reduce the need for data standardization after the survey. For example, I know there are only so many WoW character classes and roles. I know that players can only belong to one of several types of realms. In the case where I wanted time estimates from players, providing a list of ranges means everyone’s has the same degree of accuracy and is expressed in the same units. This was not deemed necessary for year of birth, country of residence, and nationality; they were left as free text. The birth year worked out fine, but some normalization had to be done on the countries and nationalities, e.g. English and Scottish changed to British and Belgium changed to Belgian, etc.
Read more to download the survey as a PDF and see options for specific questions.