Over-optimizing for the walk-up-and-use experience at the expense of a more realistic subset of usage is sometimes referred to as the "Pepsi Challenge" or "sip test" observational bias. You might even consider it a special case of the "streetlight effect" where people choose their research methods primarily based on how easy it is to collect that kind of data. (When I used to teach UX I had a whole lecture dedicated to alternate approaches to avoid falling into the trap of over-optimizing for novices, things like GOMS/KLM, log analysis, heuristic evaluations, diary studies, cognitive walkthough, surveys, etc.)