Worst case scenario:
Ethnomethodology fails because you can’t dive into the users’ culture for a year and even if you could, first you’d have to acquire all of their special training, education, experience, etc.
Ethnomethodological phenomenology fails because you might ask the user the wrong kinds of “why” questions, or ask them the wrong way, or be misheard or misinterpreted, or create a Heissenberg Effect or a Hawthorne Effect in the process of asking.
Structuralism fails because, you probably won’t share all of the codes necessary to understand the user, let alone do a proper analysis of their needs.
What then?
Interaction Culture 
Permalink
Add the “post” prefix to any of the above and try again? 🙂
Permalink
That is just the kind of sneaky thing he’d do, isn’t it.
Permalink
Sneaky and snarky. It can’t be beat.
Permalink
Adding the “post” prefix appears to have been Verbeek’s solution, as we saw in Roedl’s postphenomenology entry.
But seriously, these questions are why we don’t have recipes and why professional and critical judgment are central to our success. Every method is in the end philosophically indefensible. Jonathan Culler has a great phrase for this: “methodological fictions,” which refers to the little lies we engage in just to be able to do research, so we don’t become Hamlet.
So the real question is not “How do we eliminate these problems?” but rather, “How do we manage them?”. The answer to that is we legitimate our judgment with a combination of established theory and explicit evidence (the big one being “data”) that demonstrates we are reasonable to think and act the way we do, as evaluated via review structures (such as peer review). So even if we wouldn’t care to prove to Plato or God that we know we are Right, we can demonstrate that we are operating with due diligence and externally verifiable intellectual integrity.