PERT Estimating: Three-Point Estimation, Standard Deviation & Confidence Intervals
Program Evaluation and Review Technique (PERT) is a statistical tool used in project management to analyze and represent the uncertainty inherent in task duration estimates. Unlike deterministic estimating (which produces a single-point estimate), PERT uses three estimates per activity to model the probability distribution of completion times. On the PMP exam, PERT and three-point estimating appear in 2–4 questions, typically testing your ability to calculate the weighted average (expected duration), standard deviation, variance, and confidence intervals for a project's total duration.
PERT was developed by the US Navy in the 1950s for the Polaris missile program — a project involving over 3,000 contractors and 200,000 activities. The technique gave program managers a way to predict completion dates with statistical confidence, even when individual activity durations were highly uncertain. Today, PERT remains a core technique in the PMBOK's Estimate Activity Durations process and is a staple of PMP exam math.
The Three Estimates: O, M, and P
Every PERT calculation starts with three estimates for each activity:
- O (Optimistic): The minimum possible duration, assuming everything goes perfectly. This should have less than a 1% chance of being attained. It is the absolute best-case scenario.
- M (Most Likely): The duration that would occur most frequently if the activity were repeated many times under the same conditions. This is the modal estimate — not an average.
- P (Pessimistic): The maximum possible duration, assuming everything goes wrong (except major catastrophes). This should also have less than a 1% chance of being exceeded.
The PMP exam will give you these three values directly and ask you to calculate the expected duration (PERT weighted average), the standard deviation, and/or the variance. Sometimes, instead of asking for the duration of a single activity, the exam asks you to sum PERT estimates across a path (or the entire project) and calculate confidence intervals.
The PMP exam always provides the three values explicitly. You never need to derive them from prose. Look for phrases like "Given the following three-point estimate: Optimistic = 10 days, Most Likely = 15 days, Pessimistic = 30 days." The formulas are straightforward once you have these numbers.
PERT Beta Distribution (Weighted Average)
Formula: Expected Duration (tₑ) = (O + 4M + P) ÷ 6
This is the classic PERT formula. It gives more weight to the Most Likely estimate (4×) and produces a beta distribution that is smooth and bell-shaped when O and P are symmetric around M. The formula is derived from the beta distribution, which is the theoretical foundation of PERT.
Worked example: A software development task has O = 8 days, M = 14 days, P = 26 days.
tₑ = (8 + 4×14 + 26) ÷ 6 = (8 + 56 + 26) ÷ 6 = 90 ÷ 6 = 15 days.
The expected duration is 15 days — slightly higher than the Most Likely (14 days) because the Pessimistic (26) is farther from M than the Optimistic (8) is. This right-skew is common in real projects where bad outcomes are more extreme than good outcomes.
Triangular Distribution
Formula: Expected Duration (Triangular) = (O + M + P) ÷ 3
The triangular distribution is a simpler alternative to PERT. It gives equal weight to all three estimates. While PERT (beta) is the more theoretically rigorous approach, the PMBOK recognizes both methods. The PMP exam primarily tests the PERT beta distribution, but you should know the triangular formula exists and when you might use it.
Worked example (same task): (8 + 14 + 26) ÷ 3 = 48 ÷ 3 = 16 days.
| Characteristic | PERT (Beta) Distribution | Triangular Distribution |
|---|---|---|
| Formula | (O + 4M + P) ÷ 6 | (O + M + P) ÷ 3 |
| Weight on Most Likely | 4× (heavily weighted) | Equal to O and P |
| When to use | Standard approach on PMP exam | Small datasets or limited history |
| Shape | Smooth, bell-curve-like (beta) | Linear, triangular shape |
| Exam prevalence | Primary (90% of questions) | Secondary (10%, usually mentioned by name) |
Unless the question explicitly says "triangular distribution," use the PERT formula (O + 4M + P) ÷ 6. If a question says "three-point estimating" without specifying, PERT is the default. The triangular formula is tested less frequently, but you should know it — the PMBOK includes both and the exam can ask about either.
Standard Deviation and Variance
Once you have the expected duration, the next step is to measure uncertainty. Standard deviation and variance quantify how spread out the possible durations are. The wider the gap between Optimistic and Pessimistic, the higher the uncertainty.
Standard Deviation (σ) for a Single Activity
Formula: σ = (P − O) ÷ 6
The standard deviation represents the average amount by which the actual duration could deviate from the expected duration. For PERT, the divisor is 6 because the range (P − O) is assumed to span approximately 6 standard deviations (3 on each side) for a beta distribution.
Worked example: O = 8, P = 26. σ = (26 − 8) ÷ 6 = 18 ÷ 6 = 3 days.
This means the actual duration of this task is expected to fall within ±3 days of the expected 15-day estimate approximately 68% of the time.
Variance (σ²) for a Single Activity
Formula: σ² = [(P − O) ÷ 6]²
Variance is the square of the standard deviation. It is used when summing across multiple activities because variances add (standard deviations do not — you cannot simply add standard deviations together).
Worked example: σ² = (3)² = 9 days².
Standard Deviation for the Project (Multiple Activities on a Path)
Formula: σproject = √(σ₁² + σ₂² + σ₃² + ... + σₙ²)
To find the total project standard deviation, you sum the variances of all activities on the critical path, then take the square root. This is a critical distinction: you cannot add standard deviations directly. The central limit theorem governs this — the sum of independent random variables has a variance equal to the sum of their variances.
Worked example: Three activities on the critical path have standard deviations of 2, 3, and 1.5 days.
Sum of variances = 2² + 3² + 1.5² = 4 + 9 + 2.25 = 15.25
σproject = √15.25 ≈ 3.91 days.
| Metric | Single Activity Formula | Project (Path) Formula |
|---|---|---|
| Expected Duration | (O + 4M + P) ÷ 6 | Sum of activity tₑ |
| Standard Deviation (σ) | (P − O) ÷ 6 | √(Sum of σ²) |
| Variance (σ²) | [(P − O) ÷ 6]² | Sum of activity σ² |
This is one of the most common PERT mistakes on the exam. When calculating project-level standard deviation from multiple activities, you must: (1) square each σ to get σ², (2) sum all σ² values, (3) take the square root of the sum. Do NOT simply add the individual standard deviations together — that gives an incorrect result.
Confidence Intervals: The 68-95-99.7 Rule
Once you know the project's expected duration and its standard deviation, you can calculate confidence intervals — the range of durations within which the actual project completion is likely to fall. The PMP exam uses the normal distribution's empirical rule (68-95-99.7) for this purpose.
The Empirical Rule (Normal Distribution)
- 68% confidence: Project duration = Mean ± 1σ (one standard deviation)
- 95% confidence: Project duration = Mean ± 2σ (two standard deviations)
- 99.7% confidence: Project duration = Mean ± 3σ (three standard deviations)
Worked example: A project has an expected duration (from PERT) of 120 days and a project standard deviation of 8 days.
- 68% confidence: 120 ± 8 = 112 to 128 days
- 95% confidence: 120 ± 16 = 104 to 136 days
- 99.7% confidence: 120 ± 24 = 96 to 144 days
The PMP exam often asks: "What is the probability that the project will finish by day 128?" Since 128 is 1σ above the mean, and the probability of being within ±1σ is 68%, the probability of being below 128 is 50% + (68%/2) = 84%. The probability of being above 128 is 16%. These one-tail probabilities are tested frequently.
| Confidence Level | Range (Mean ± kσ) | Probability of Exceeding Upper Bound |
|---|---|---|
| 68% (1σ) | Mean ± 1σ | 16% |
| 95% (2σ) | Mean ± 2σ | 2.5% |
| 99.7% (3σ) | Mean ± 3σ | 0.15% |
One-Tail vs. Two-Tail Questions
A two-tail question asks about a range (e.g., "probability of finishing between 112 and 128 days"). A one-tail question asks about a single bound (e.g., "probability of finishing by day 128"). For one-tail, remember that 50% of outcomes fall below the mean. Then add the half-interval probability. For 1σ above the mean: 50% + 34% = 84%. For 1σ below the mean: 50% − 34% = 16%.
1σ = 68% of data, 2σ = 95%, 3σ = 99.7%. For one-tail questions, it's always 50% ± (half the interval percentage). Above the mean: 50% + (interval%/2). Below the mean: 50% − (interval%/2).
When PERT Is Used on the PMP Exam
PERT questions on the PMP exam fall into several categories. Understanding the pattern helps you prepare:
- Direct Calculation (40%): You are given O, M, and P for a single activity and asked to compute the expected duration (tₑ) or standard deviation. These are the most straightforward PERT questions — just plug and chug.
- Multi-Activity Path (25%): You are given three-point estimates for several activities on the critical path and asked to calculate the total project duration (sum of tₑ) and the project standard deviation (root of sum of variances).
- Confidence Interval (20%): After calculating the expected duration and standard deviation, you are asked the probability of completing by a specific date. This tests your understanding of the 68-95-99.7 rule and one-tail probabilities.
- PERT vs. Triangular (10%): The exam contrasts the two methods. You may be asked to compute both and explain why they differ, or to select the appropriate method for a given scenario.
- PERT + CPM Integration (5%): A scenario combining network diagram analysis with PERT estimates. You identify the critical path using PERT expected durations, then calculate the project's probability of completing by a deadline.
Combining PERT with CPM
This is an advanced topic that appears in harder PMP questions. The approach is to replace deterministic activity durations with PERT expected durations, run the forward/backward pass on the network, identify the critical path using these adjusted durations, and then calculate confidence intervals for that path.
Worked example — PERT + CPM integration:
Three activities make up the critical path of a network (identified using PERT expected durations):
| Activity | O | M | P | tₑ | σ | σ² |
|---|---|---|---|---|---|---|
| A | 5 | 8 | 17 | (5+32+17)÷6 = 9 | (17−5)÷6 = 2 | 4 |
| B | 10 | 12 | 20 | (10+48+20)÷6 = 13 | (20−10)÷6 = 1.67 | 2.79 |
| C | 3 | 5 | 7 | (3+20+7)÷6 = 5 | (7−3)÷6 = 0.67 | 0.45 |
Project expected duration = 9 + 13 + 5 = 27 days.
Project variance = 4 + 2.79 + 0.45 = 7.24.
Project standard deviation = √7.24 ≈ 2.69 days.
Now the exam might ask: "What is the probability of completing this project within 30 days?"
30 days is 30 − 27 = 3 days above the mean. Z = 3 ÷ 2.69 = 1.12 standard deviations. Since 1.12σ falls between 1σ (68% → 84% one-tail) and 2σ (95% → 97.5% one-tail), the probability is approximately 87% (interpolating between 84% and 97.5%). Most PMP questions stick to exact σ values (1, 2, or 3) to avoid interpolation. If 30 were exactly 2σ above (27 + 2×2.69 = 32.38), it would not match, but if the question asked about day 30 (which is approximately 1.12σ), they would expect recognition that it is between 1σ and 2σ. In practice, the exam keeps numbers clean: deadlines at exactly ±1σ, ±2σ, or ±3σ from the mean.
Common PERT Mistakes and How to Avoid Them
- Confusing the PERT and triangular formulas: PERT divides by 6; triangular divides by 3. The weights are (1, 4, 1) for PERT and (1, 1, 1) for triangular. If you are unsure, look for the phrase "beta distribution" or "PERT" — that signals the weighted formula.
- Adding standard deviations instead of variances: For the project-level σ, always square each σ, sum, then square root. Never add σ values directly.
- Misapplying the empirical rule: Remember that 68% is the two-tail probability (within ±1σ). For one-tail (above or below), it is 50% ± 34%. For ±2σ, the two-tail is 95%, so one-tail above +2σ is 2.5%.
- Forgetting to identify the correct path: When combining PERT with CPM, always run the forward/backward pass using PERT expected durations to confirm which path is critical. The critical path can change when you replace deterministic estimates with PERT values.
- Ignoring that the distribution approximates normal at the project level: The central limit theorem ensures that the sum of many activity durations (each following a beta distribution) approximates a normal distribution. This is why the 68-95-99.7 rule applies at the project level even though individual activities follow a beta distribution.
Full PERT Worked Example — Step by Step
Scenario: Your project has four activities on the critical path. Their three-point estimates (O, M, P) in days are:
- Activity 1: (4, 6, 14)
- Activity 2: (8, 10, 18)
- Activity 3: (3, 5, 7)
- Activity 4: (6, 9, 24)
Step 1 — Calculate each activity's expected duration:
Activity 1: (4 + 24 + 14) ÷ 6 = 42 ÷ 6 = 7.0 days
Activity 2: (8 + 40 + 18) ÷ 6 = 66 ÷ 6 = 11.0 days
Activity 3: (3 + 20 + 7) ÷ 6 = 30 ÷ 6 = 5.0 days
Activity 4: (6 + 36 + 24) ÷ 6 = 66 ÷ 6 = 11.0 days
Step 2 — Calculate each activity's standard deviation:
Activity 1: (14 − 4) ÷ 6 = 10 ÷ 6 = 1.67 days
Activity 2: (18 − 8) ÷ 6 = 10 ÷ 6 = 1.67 days
Activity 3: (7 − 3) ÷ 6 = 4 ÷ 6 = 0.67 days
Activity 4: (24 − 6) ÷ 6 = 18 ÷ 6 = 3.0 days
Step 3 — Calculate each activity's variance:
Activity 1: (1.67)² = 2.79
Activity 2: (1.67)² = 2.79
Activity 3: (0.67)² = 0.45
Activity 4: (3.0)² = 9.0
Step 4 — Calculate project totals:
Project expected duration = 7.0 + 11.0 + 5.0 + 11.0 = 34.0 days
Project variance = 2.79 + 2.79 + 0.45 + 9.0 = 15.03
Project standard deviation = √15.03 ≈ 3.88 days
Step 5 — Confidence intervals:
68%: 34.0 ± 3.88 = 30.12 to 37.88 days
95%: 34.0 ± 7.76 = 26.24 to 41.76 days
99.7%: 34.0 ± 11.64 = 22.36 to 45.64 days
Step 6 — Probability analysis:
What is the probability of completing by day 38? 38 − 34 = 4 days above the mean. 4 ÷ 3.88 = 1.03σ. This is approximately 1σ, so the probability is approximately 84% (50% + 34% = 84%).
What is the probability of completing by day 42? 42 − 34 = 8. 8 ÷ 3.88 = 2.06σ. Approximately 2σ, so approximately 97.5% (50% + 47.5% = 97.5%).
PERT estimating gives you the statistical foundation to answer these questions with confidence. On the PMP exam, master these calculations and you will handle any PERT question that comes your way.
← Back to PMP Formulas Hub | Next: Risk Management Formulas →
📚 Sources & References
- 🔗 PMI Official PMP Certification — Project Management Institute
- 🔗 PMBOK Guide — Seventh Edition — PMI Standards
- 🔗 PMP Exam Content Outline (ECO) — Official exam blueprint