What a label costs

Somewhere in the standard case study there is a sentence that goes: users struggle to find the information they need. It sits under a heading that says Research, so the reader takes it as a finding. Often it is a finding. Just as often it is a reasonable inference from three interviews, or a restatement of the brief, or something the designer believed before the study began and did not stop believing afterwards.

Nobody is lying. The format simply does not have a place to put the difference.

So I started labelling. Every claim in a case study on this site is marked as observed, interpreted, a hypothesis, or an unvalidated concept. Observed claims carry a source. Hypotheses are written down before the fieldwork that might disprove them, which is a slightly uncomfortable thing to publish.

It is worth paying for two reasons. The first is that it makes the strong claims mean something. If four claims on a page are marked observed and the rest are marked as interpretation, the four observed ones land harder than they would in a document where everything is asserted flatly.

The second reason is more selfish. Labelling forces the question at the moment of writing, when it is still cheap to answer. You sit down to type 'users want' and you have to decide, immediately, whether you watched that happen or concluded it. About a third of the time the honest answer is the second one, and knowing which third is most of what research practice actually is.

The failure mode I am watching for is that labels become decoration. A badge next to every sentence, applied loosely, meaning nothing. The discipline only holds if downgrading is possible: if a claim can move from observed to interpretation when the evidence turns out to be thinner than remembered, and if a concept that fails testing gets deleted rather than quietly relabelled.