Most companies treat user understanding as a solved problem. It isn't. Hannah Shamji's four-level framework, surfaced in a recent Smashing Magazine piece by Vitaly Friedman, argues that what users say, think, feel, and do are four distinct data sources, and conflating them produces bad product decisions. The framework runs from Level 1 (stated opinions, least reliable) through Level 4 (underlying motivations, hardest to reach). Most teams never get past Level 2.
The article's most useful technical argument is about why direct questioning fails. Erika Hall's position is cited directly: asking a question is the worst way to get a true answer to it. A study on Dutch verbal probability terms backs this up with data, showing that words like 'possible' and 'likely' produce wildly inconsistent numerical interpretations across respondents. Friedman also drops a specific usability testing practice worth noting: he does not use speak-aloud protocol, because verbalizing a task obscures the emotions happening during it. Instead he watches mouse circles, scroll depth, and hesitation, then asks questions only after the user declares they are done or stuck.
The piece is worth reading in full for the operational detail on reaching Level 4, which requires repeat interviews, task walkthroughs, and a trust relationship built over time. Friedman also flags that the four levels frequently produce contradictory data, and links to his own mixed-method reconciliation approach. The Geoffrey Roberts Emotion Wheel gets a mention as a practical instrument for moving past binary good-or-bad sentiment coding in interview sessions.
[READ ORIGINAL →]