Extreme users are information-rich during exploration
Aliases: extreme user research · extreme-case sampling · critical case
What it is
Extreme-case sampling deliberately selects people near an edge of capability, frequency, environmental constraint, or outcome so mechanisms hidden in ordinary cases become visible. Heavy users expose accumulated interaction cost; infrequent users expose memory and learnability demands; resource-constrained users reveal assumptions about bandwidth, language, or assistive capability. “More valuable” means potentially more new information per exploratory case, not greater human importance than typical users.
Why it happens
System constraints often appear near thresholds. Short content fits while long content exposes truncation; repeated expert action amplifies small delays; low prior knowledge makes assumed concepts visible. Extreme cases increase contrast and help identify variables, boundaries, and failure chains. An edge experience may also combine several unusual causes, however, so it generates mechanism candidates and requirements rather than population prevalence.
Studying it
Define the relevant extreme dimension and selection threshold before recruitment rather than choosing merely interesting people. Maximum-variation sampling can compare both ends with intermediate cases, asking which difficulties change continuously and which are context-specific. Report why each case counts as extreme and whether its recruitment channel confounds the focal dimension with other attributes. Estimating generality then requires typical cases, stratified samples, or quantitative estimation.
Where it stops holding
Extreme users cannot estimate average experience, market prevalence, or default priority. An edge requirement may conflict with majority goals, so discovering it does not imply one universal solution. Extremity can also be a temporary state: novices learn and networks change. Low-frequency accessibility and safety needs should not be dismissed for failing to represent an average; their priority can derive from rights and consequence rather than prevalence.
Applying it
- Select both ends of a constraint such as lowest/highest experience, shortest/longest content, or stable/intermittent connectivity.
- Use extreme cases to expose assumptions and generate boundary requirements, never to report population rates.
- Separate general mechanisms, edge-specific conditions, and prevalence questions still requiring estimation.
- Test the resulting design with an intermediate case to detect edge optimization that damages the primary path; validate safety and access requirements by consequence separately.