More and more I find myself leaning toward a sub version of the "Agile programming methodology", though this kind of thinking works better for somewhat smaller web-based programs than traditional large software releases like SQL Server. As a small independent developer I too have experienced users not really knowing what they want, but if I use a modular "Agile" approach I can mock up a more or less working version of a new feature and put it quickly into production to see if this is really what the users want and if it is I can (hopefully) spend the time to improve the code later to eliminate bugs etc... When you don't know if a feature will really meet the needs of your users because they don't really know what they want themselves, it just makes a lot more sense to focus on speed over quality, since spending a lot of time getting a feature perfect when in the end it won't really meet the business need of the user just makes no sense at all.