Judging how long it takes.
WHAT DRIVES EFFORT
Scope of change Complexity Quality of what is delivered Environment availability Number of configurations
WHY DELIVERY QUALITY MATTERS
Testing a defective build takes far longer, because of reporting and retesting cycles.
WHAT TO ESTIMATE SEPARATELY
Test design Execution Defect reporting and retesting Regression
WHAT TO ADD
Contingency for retesting, which is always needed.
HOW MUCH
A substantial proportion, based on historical defect rates.
WHAT TO BASE ESTIMATES ON
Previous similar work, where records exist.
WHY RECORDS MATTER
Estimates without history are guesses.
WHAT TO TRACK
Actual effort against estimate, to improve future ones.
WHAT TO STATE WITH ANY ESTIMATE
The assumptions it rests on.
WHAT THE COMMON ASSUMPTIONS ARE
The environment is available The build is testable Requirements are stable Defect volume is typical
WHAT TO DO WHEN ONE FAILS
Revise the estimate, and say so immediately.
WHAT TO RESIST
Estimates set by the date rather than the work.
WHAT TO OFFER INSTEAD
What can be tested in the time available, and what the residual risk is.
