The summary.
PEOPLE DESCRIBE SOLUTIONS, NOT PROBLEMS
And the solution they describe is frequently not the best one. Ask what they are trying to achieve, how they do it now, and what goes wrong.
Ask how they would know it was successful — almost nobody states the measure unprompted.
WATCH PEOPLE DO THE WORK
What people describe and what they do differ, consistently. Look for workarounds — they reveal where the current system fails.
ADJECTIVES ARE NOT REQUIREMENTS
Fast, easy, intuitive and robust are untestable. Ask of everything: how would we verify this?
And define your terms once — most requirement disputes are vocabulary disputes.
WHEN EVERYTHING IS DECLARED ESSENTIAL, ASK WHICH HALF
It forces a real ordering without asking anyone to abandon anything.
Deliver a thin path end to end first — it proves the approach and surfaces integration problems early.
SHOWING SOMETHING REVEALS WHAT REVIEWING DOCUMENTS DOES NOT
People recognise what is wrong far more reliably than they specify what is right.
REPEATED CHANGES IN ONE AREA MEAN THE REQUIREMENT WAS NEVER UNDERSTOOD
Return to the problem rather than continuing to adjust the solution.
