The summary.
WHAT A SMALL BUSINESS CAN REALISTICALLY DO
Talk to five customers, watch one person attempt the task, and read your support requests.
That is enough to avoid the largest mistakes. Formal methods are not required.
ASK ABOUT WHAT HAPPENED, NOT WHAT THEY WOULD DO
People are poor predictors of their own behaviour and will agree agreeably to anything hypothetical.
Ask about their last experience, specifically.
WATCHING BEATS ASKING
Give a realistic task, ask them to think aloud, and do not help.
Every intervention destroys the evidence. Five people reveals most problems.
SUPPORT REQUESTS ARE FREE RESEARCH
Every question about how to do something is a design failure that reached a person.
Fix the interface rather than improving the answer — an answer serves one person, a fix serves everyone.
A WORKAROUND IS AN UNSTATED REQUIREMENT
When people have invented their own way round something, that is the design telling you what it lacks.
BEFORE REDESIGNING ANYTHING
Measure the current position. Without a baseline you cannot tell whether you improved it.
WITH STAKEHOLDERS
Ask what problem an opinion solves, and for which user. That moves the discussion from taste to evidence.
