This confirms my belief that <i>nothing</i> compares to real-life usability testing.<p>Not enough people do it, and those that do are always surprised at the results. You can't second-guess users, they are unpredictable. Well not entirely, but the factors to consider are too numerous and complex to predict reliably.<p>I now rarely accept arguments for change on the basis of "usability" unless it has actually been visibly demonstrated by real users, rather than hypothetical textbook ones. Kudos to Skritter.