Evidence.
WHAT TO MEASURE
Task completion: do people succeed?
Time taken Errors made Where people abandon What support is asked about
WHAT NOT TO MEASURE ALONE
Opinions about appearance.
People are poor judges of what helps them.
WHAT TO OBSERVE
Actual attempts, by real people.
Five people attempting a task reveals most usability problems.
WHAT TO COMPARE
Before and after a change One version against another, where volume allows
WHAT TO WATCH FOR
Hesitation Backtracking Wrong first choices Reading the same thing twice
THOSE ARE FAILURES, EVEN WHEN THE TASK COMPLETES
WHAT ANALYTICS SHOWS
Where people leave, and how far they get.
WHAT IT DOES NOT SHOW
Why.
WHAT TO COMBINE
Figures to find where, observation to find why.
WHAT TO ACT ON
The largest failure point first.
