Test Estimation
How long will it take? As a Lead, you are responsible for predicting the future. Test estimation is a science of ranges, assumptions, and defensible numbers.
1 The Hook — The "Liar" Problem
A Project Manager asks you: "How long to test this?" You say "Two days." They say "Can you do it in one?" You say "Okay." Congratulations, you just lost all credibility.
If you change your estimate without changing the scope, you've just admitted your first number was a guess (or a lie). Professional estimation is about showing your work so your numbers can't be bargained away.
2 The Rule — The PERT Formula
Use Three-Point Estimation (PERT) to calculate a realistic weighted average.
Estimate = (O + 4M + P) / 6
- Optimistic (O): Best case, no bugs, environment is perfect.
- Most Likely (M): Realistic, average number of bugs.
- Pessimistic (P): Worst case, critical blockers, environment down.
3 The Analogy — The House Renovation
Buffer vs. Reality.
When a builder estimates a kitchen renovation, they give you a price. But as soon as they rip out the old cupboards, they find **Leaking Pipes** (Bugs) or **Rotten Wood** (Bad Requirements). A pro builder includes a **Contingency Fund** (Buffer) because they know hidden problems always exist. Testing is the same. You aren't just estimating the "Happy Path" (The Kitchen Cabinets); you are estimating the "Unhappy Path" (The Leaking Pipes).
4 Watch Me Do It — NZ E-commerce Sprint
Scenario: Estimating testing for a new "Express Shipping" option.
-
Break it down (WBS):
Design Tests: 2h
Environment Setup: 1h
Execution: 4h
Defect Retesting: 2h -
Apply Three-Point (for Execution):
Optimistic: 2h (Easy)
Most Likely: 4h (Standard)
Pessimistic: 10h (Payment gateway fails) - Calculate: (2 + 16 + 10) / 6 = 4.6 hours.
- The Result: "It will take 9.6 hours. This assumes the Staging environment is available by 9am Monday."
5 Decision Tool — Which Technique?
🧠 Expert Judgement (Quick)
- Use when: You've done this 100 times before.
- Pro: Very fast.
- Con: Hard to defend if challenged.
📊 Metrics-Based (Data-driven)
- Use when: You have historical data (e.g. "Last time took X hours").
- Pro: Very hard to argue with.
- Con: Requires record-keeping.
6 Common Mistakes
🚫 Accepting the PM's estimate
I used to think: I'll try my best to finish in the 4 hours they gave me.
Actually: If the work takes 8 hours, you'll either fail or work overtime. Both are bad. Your job is to say: "Here is why it takes 8 hours. Which of these tests should we skip to make it 4?"
🚫 Forgetting Defect Retesting
I used to think: I'll estimate based on running the tests once.
Actually: Tests find bugs. You have to wait for a fix, re-install the app, and run the test again. You must factor in 20-30% of your time just for defect management.
7 Now You Try — Calculate the PERT
Scenario: You're estimating a Regression Run.
- Optimistic (O): 4 hours
- Most Likely (M): 6 hours
- Pessimistic (P): 14 hours
What is the PERT estimate?
Math: (4 + (4x6) + 14) / 6 = (4 + 24 + 14) / 6 = 42 / 6 = 7.
8 Self-Check
Q1. What is "Expert Judgement"?
It is an estimation technique based on the experience of senior team members. It is often used in Agile (Planning Poker) to reach a consensus estimate.
Q2. Why should you always state your "Assumptions"?
Because if an assumption is proven wrong (e.g. "The server was not ready"), your estimate is no longer valid. Stating assumptions protects you when things out of your control go wrong.
9 Interview Prep
"What do you do when your estimate is rejected as being 'too high'?"
Answer: "I show them my Work Breakdown Structure (WBS). I say: 'This estimate covers Design, Execution, and Retesting. If we need to finish in half the time, we must reduce the scope. Which features are lowest risk that we can skip?' I never just lower the number without lowering the work."
10 Next Step
You've estimated the work. Now, how do we measure if we're actually doing it well? Next: Test Metrics.