The summary.
FOLLOW A METHOD
Identify, theorise, test, plan, implement, verify, document.
Random changes create new faults and obscure the original. Change one thing at a time, and reverse it if it does not help.
THE MOST USEFUL QUESTION IS WHAT CHANGED
Most faults follow a change. But people conceal actions they fear being blamed for — ask neutrally, because that is usually the key information.
FIVE QUESTIONS THAT HALVE THE PROBLEM SPACE
One user or many? One device or many? One application or all? One location or everywhere? Constant or intermittent?
TEST WITH A DIFFERENT USER ACCOUNT
It distinguishes a profile or permission problem from an application problem, and saves hours.
FOR NETWORK FAULTS, TEST AN EXTERNAL ADDRESS BY NUMBER
It separates connectivity from name resolution in one step.
SOFTWARE FAULTS IMITATE HARDWARE FAULTS CONVINCINGLY
Check whether the fault persists with minimal software before concluding hardware.
And copy the data off a failing drive before any further diagnosis.
