PERT Estimation: Statistical Confidence for Project Planning

When you need numbers you can defend.

---

What Is PERT?

PERT (Program Evaluation and Review Technique) uses three-point estimation to account for uncertainty. It's more rigorous than T-shirts or story points--use it when precision matters and you need to communicate confidence levels to stakeholders.

Unlike other techniques that embrace abstraction, PERT produces actual time estimates with statistical backing. This makes it ideal for external commitments, resource planning, and high-stakes projects.

---

The Three Estimates

For each task, provide three estimates:

Estimate	Symbol	Description
Optimistic	O	Best case--everything goes right
Most Likely	M	Realistic expectation based on experience
Pessimistic	P	Worst case--complications arise

The key is intellectual honesty:

• Optimistic isn't wishful thinking--it's genuinely achievable if nothing goes wrong
• Most Likely is what you'd bet money on
• Pessimistic includes realistic problems, not catastrophes (server catches fire ≠ pessimistic estimate)

---

The PERT Formula

Expected Duration = (O + 4M + P) / 6

Standard Deviation = (P - O) / 6

The formula weights the most likely estimate heavily (4x) while accounting for risk through optimistic and pessimistic bounds.

Why These Weights?

The PERT distribution approximates a beta distribution. The 4x weight on M reflects that most outcomes cluster around the most likely scenario, while O and P capture tail risks.

---

A Complete Example

Let's estimate a database migration:

The Three Estimates:

Scenario	Days	Reasoning
Optimistic	3	Schema is simple, existing tests pass, no data issues
Most Likely	5	Typical migration with some debugging, minor fixes needed
Pessimistic	10	Data inconsistencies found, rollback needed, multiple attempts

Calculations:

Expected = (3 + 4×5 + 10) / 6
Expected = (3 + 20 + 10) / 6
Expected = 33 / 6
Expected = 5.5 days

Standard Deviation = (10 - 3) / 6
Standard Deviation = 7 / 6
Standard Deviation = 1.17 days

Interpretation:

• Expected duration: 5.5 days
• 68% confidence: 4.3 to 6.7 days (±1 SD)
• 95% confidence: 3.2 to 7.8 days (±2 SD)

You can tell stakeholders: "We expect 5.5 days, with high confidence it'll be under 8 days."

---

When to Use PERT

PERT is more work than other techniques. Use it when the extra precision matters:

Critical Path Estimation

When one task blocks many others, you need confidence intervals, not rough guesses.

External Commitments

Client deadlines, launch dates, and contract obligations warrant statistical backing.

Resource Planning

Staffing decisions based on vague estimates lead to either overwork or underutilization.

Risk Assessment

PERT naturally identifies high-variance tasks. If P - O is large, that task needs attention.

Budget Justification

"5.5 days ± 1.2 days based on PERT analysis" is more defensible than "about a week."

---

Aggregating PERT Estimates

For a project with multiple tasks, sum the expected values. Standard deviations combine using the square root of the sum of squares:

Project Expected = E₁ + E₂ + E₃ + ...

Project SD = √(SD₁² + SD₂² + SD₃² + ...)

Multi-Task Example

Task	E (days)	SD (days)
Database migration	5.5	1.17
API updates	4.0	0.67
Frontend changes	3.0	0.83

Project Expected = 5.5 + 4.0 + 3.0 = 12.5 days

Project SD = √(1.17² + 0.67² + 0.83²)
Project SD = √(1.37 + 0.45 + 0.69)
Project SD = √2.51
Project SD = 1.58 days

Result: 12.5 days expected, with 95% confidence between 9.3 and 15.7 days.

---

PERT vs Story Points

Aspect	PERT	Story Points
Output	Time estimates	Relative complexity
Precision	High	Medium
Effort to create	High	Low
Works for external stakeholders	Yes	Requires translation
Tracks velocity	No	Yes
Acknowledges uncertainty	Explicitly	Implicitly

Use story points for sprint planning; use PERT for commitments with consequences.

---

Common PERT Mistakes

1. Optimistic ≠ Most Likely

If your optimistic and most likely estimates are the same, you're not accounting for what could go right. Push the optimistic estimate lower.

2. Pessimistic ≠ Catastrophe

A pessimistic estimate includes realistic problems--key engineer gets sick, requirements change, unexpected complexity. It doesn't include earthquakes.

3. Forgetting Dependencies

PERT aggregation assumes tasks are independent. If Task B depends on Task A, you can't simply add standard deviations.

4. Over-Precision

"5.47 days" implies false precision. Round appropriately--"5.5 days" is plenty specific.

5. Using PERT for Everything

PERT's overhead isn't justified for low-stakes work. Use simpler techniques for routine tasks.

---

Tips for Better Three-Point Estimates

Think Scenario, Not Number

Instead of "What's my optimistic estimate?" ask "What scenario would lead to the fastest completion?" Scenarios are easier to reason about.

Use Historical Data

If you've done similar work before, use those actuals to calibrate your estimates. "Last migration took 6 days, so M = 5-6 days."

Include the Right People

The person doing the work should estimate. If that's not possible, include them in the discussion.

Document Assumptions

Write down what you assumed for each estimate. "Optimistic assumes no data cleanup needed" helps future you understand the estimate.

Review and Calibrate

After completing a task, compare actuals to estimates. Were you consistently optimistic? Pessimistic? Adjust your future estimates accordingly.

---

PERT in Practice: A Real Project

Let's estimate a feature: user authentication with OAuth providers.

Task Breakdown

Task	O	M	P	E	SD
OAuth library research	0.5	1	2	1.08	0.25
Google OAuth integration	1	2	4	2.17	0.50
GitHub OAuth integration	0.5	1	2	1.08	0.25
Database schema changes	0.5	1	3	1.25	0.42
Session management	1	2	4	2.17	0.50
Frontend components	1	2	3	2.00	0.33
Testing & bug fixes	1	2	5	2.33	0.67
Total	--	--	--	12.08	1.15

Interpretation

• Expected: 12 days
• 68% confidence: 10.9 - 13.2 days
• 95% confidence: 9.8 - 14.4 days

Communication: "We'll deliver OAuth in 12 days, with high confidence we'll be done within 2.5 weeks."

Risk Identification

Notice testing has the largest SD (0.67). This suggests it's the highest-risk task. Consider:

• Starting testing earlier in parallel
• Allocating extra buffer
• Breaking down the testing estimate further

---

Confidence Levels Explained

Probability	Range
50%	E (expected value)
68%	E ± 1 SD
95%	E ± 2 SD
99.7%	E ± 3 SD

For external commitments, 95% confidence is typically appropriate. For internal planning, 68% may suffice.

---

Try It Now

Ready to calculate PERT estimates? Our free tool handles the math:

PERT Calculator - Three-point estimation with statistical output

---

Next up: Planning Poker - team-based estimation that surfaces hidden assumptions.

---

Last updated: January 2026