On Tue, 18 Sep 2012 16:42:44 +0100, Nelson Marques wrote:
So, we agree that there's fragmentation on the user base; and it's somehow 'common sense' to believe that we need to know and understand our user base, so we can take the proper calls on situations like this, correct?
The big question is... "how do we do that ?"...
Traditionally this answer should be provided by Marketing, correct? So what can Marketing do for us? (or should we just jump in big and bad based on 'assumptions' ?)
The obvious answer to that question is "you survey the userbase". The / real/ question is "how do you get people who are in the userbase to answer the survey in enough numbers to have a statistically valid sample?" Because without a good statistical sampling, you're not dealing with useful data that helps make a good data-driven decision. That typically means incentives. The next big question also is in constructing a survey that asks valid questions without assuming or suggesting an answer. The single biggest mistake people make when constructing surveys is to ask questions in a way that implies or infers there's a "correct" answer or a "desired" answer. The second mistake that's made is to ask questions that are not clear - using double negatives, poor sentence construction, etc. That becomes more difficult if you publish the survey in multiple languages. The third mistake is asking open-ended questions. In general, you want questions that have answers that are discrete and/or scale based ("on a scale of 1-5, where 1 means [...]"). Fill-in-the-blank/essay style responses take a lot of time to tabulate, and require a lot more work to interpret - and an incorrect interpretation of the respondent's intention can lead to a wrong result (from the respondent's perspective). The fourth is in asking questions that don't ultimately help make decisions. I can't begin to describe the number of surveys I've seen where questions were asked that had no apparent practical value (and in a few cases, surveys I worked with respondent data where the question asked clearly had no practical value to what was being researched). Building a good survey isn't a matter of going up to SurveyMonkey and asking a few questions. You /have/ to start with the goals in mind, and then take some time to determine what questions actually help define a direction, and then make sure the wording doesn't slant the responses to a particular pre-determined response - and that's not always a conscious thing. Jim -- Jim Henderson Please keep on-topic replies on the list so everyone benefits -- To unsubscribe, e-mail: opensuse-project+unsubscribe@opensuse.org To contact the owner, email: opensuse-project+owner@opensuse.org