The question behind the interface
Raw comment counts can overrepresent vocal customers, one channel or one unusually busy week. A topic mentioned often may be less disruptive than a rare failure that blocks an essential task. I would examine how the material was collected and what decision it can reasonably inform.
A proposed direction
Group feedback by the task affected, retain source context and distinguish frequency from severity. Turn unclear themes into research questions.
Evidence to decision
Theme: people ask where to change access. Check the journey before assuming the label is the problem.
Review the task →- Keep source context
- Group by task
- Choose an investigation
Static design example. The depicted action is not an interactive product control.
How I would investigate it
Trace a small sample of feedback records to their original context, using appropriately redacted material. Ask the team what changed because of previous feedback and which uncertainty prevents action now.
Before recruiting, we would agree the decision the study must support, the people whose experience matters and a realistic task. The session plan would specify a safe starting state and avoid leading the participant towards the proposed solution. Relevant access needs, consent and handling of research material would be included in the scope.
The tradeoff to examine
A taxonomy makes information easier to organise but can flatten important differences. Keep exceptions and contradictory evidence visible. Do not treat a support sample as representative of all customers or present a synthetic comment as a participant quotation.
What would count as useful evidence?
Whether another analyst can follow the evidence trail, distinguish priority from popularity and identify the next decision or study.
Record the actions, misunderstandings and assistance required. Keep participant comments separate from interpretation and show important exceptions as well as patterns. A revised design would remain a proposal until it had been evaluated; task completion in a small qualitative study would not establish a population-wide conversion rate.