I work in epidemiology, specifically sending surveys to huge cohorts. This is what we've found through a lit review (sorry, don't have original sources but the

I work in epidemiology, specifically sending surveys to huge cohorts. This is what we've found through a lit review (sorry, don't have original sources but they're out there on PubMed). - incentives are huge. But know if you're asking someone to complete it for a social benefit (e.g. our health surveys) or for your profit. If you're doing a "feel good about helping others" thing, don't offer things that can offend. - make sure people know ahead of time what kind of questions and length of time to expect (properly set expectations) - try to "group" similar questions. This helps people's brains load different topics, so their responses are easier (fewer context shifts) - be aware that everything you do (colors, question wording, etc.) can significantly influence peoples' responses. You must be consistent with everyone who takes the survey (as much as you can -- you can't control the lighting in their rooms, as much as you'd like to). - Make sure to be neutral in your language. Don't introduce bias by writing leading questions. - try to use "standard instruments", i.e. other people's surveys that were validated and tested. Not only does this make your work easier, but you now have (the possibility of) a reference group to compare your results - don't expect very many responses (this is from my experience on all projects). People are busy. Your survey is a small part of their day - send the survey invite at least twice, spread by a couple of weeks (time permitting). Most people only complete their surveys if it's a convenient time to do so. Don't make them look back in their email -- send them a second invite (be careful beyond that, as you may annoy people). - thank them very much - have a detailed analysis plan before you create the survey. What are you trying to find out? Much like test driven development, knowing what you want to answer before you collect data is HUGE.