Skip to content
Tutorials Center
My learning

Topics

Lesson 19 of 41 ยท Scheduling and Critical Path

PERT and Three-Point Estimates

Learn PERT: optimistic, most likely and pessimistic estimates, the expected-time and variance formulas, the probability of finishing by a date and a full runnable project example.

The Program Evaluation and Review Technique (PERT) was created in the 1950s for the US Navy's Polaris missile program. Its key idea is simple and still important: durations are uncertain, so estimate them as a range, not a single number. CPM assumes each duration is exact; PERT builds on the same network but gives each activity three estimates and turns them into an expected duration and a measure of risk.

Three estimates

For each activity, ask the people who will do the work:

  • O (optimistic): the shortest time if everything goes right.
  • M (most likely): the time you expect under normal conditions.
  • P (pessimistic): the longest time if most things go wrong (but not a disaster).

Durations tend to be skewed: things can take much longer than planned but rarely much shorter. So P is usually further from M than O is.

The formulas

  • Expected duration: TE = (O + 4M + P) / 6. It is a weighted average that treats M as four times as important as the extremes (based on a beta distribution).
  • Standard deviation: SD = (P - O) / 6. A wide range means high uncertainty.
  • Variance: SD squared.

Example: O = 4, M = 6, P = 14 days. TE = (4 + 24 + 14) / 6 = 7 days (more than the most likely 6, because of the long tail), SD = (14 - 4) / 6 = 1.67.

A simpler triangular estimate uses (O + M + P) / 3. PERT weights the most likely value more.

Python PERT estimate for one activity

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Applying PERT to a network

  1. Gather O, M and P for every activity.
  2. Compute TE and variance for each.
  3. Use TE as the duration in the CPM calculation to find the critical path.
  4. Add the variances of the activities on the critical path (assuming activities are independent). The path's SD is the square root of that sum.
  5. The project duration is treated as roughly normally distributed with mean = sum of critical-path TEs and that SD.
Python PERT on the website project

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Probability of finishing by a date

If the project duration is roughly normal, you can convert a deadline into a Z score: Z = (deadline - mean) / SD. The normal table (or code) turns Z into a probability.

Rules of thumb for a normal distribution: mean plus or minus 1 SD covers about 68%, 2 SD about 95%, 3 SD about 99.7%. So "mean plus 2 SD" is a 97.5% completion date.

Python Probability of meeting a deadline

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Strengths and weaknesses

Strengths: acknowledges uncertainty; gives probabilities rather than a false single date; encourages discussion about risk; the ranges point to the riskiest activities.

Weaknesses:

  • Estimating three values per activity is time-consuming.
  • The beta/normal assumptions are approximations.
  • Only the critical path is counted, but near-critical paths can also delay the project, and the true completion time is often later than PERT says (the "merge bias"). Monte Carlo simulation fixes this; see Monte Carlo Schedules.
  • People may pad P to protect themselves or give O that is unrealistically low.

Tips for good three-point estimates

  • Write down the assumptions behind each value.
  • Make the estimator think of what could cause the pessimistic case.
  • Compare with similar completed work.
  • Ask for estimates separately before discussing, to avoid anchoring.
  • Re-estimate remaining work as the project proceeds.

PERT versus CPM

CPM PERT
Duration One value (deterministic) Three values (probabilistic)
Original use Repetitive construction work New research and development
Output Critical path, float Critical path, expected date, probability
Focus Time and cost trade-offs Time uncertainty

Modern tools combine both ideas.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. With O = 2, M = 5, P = 14, what is the PERT expected duration?

    (2 + 4 x 5 + 14) / 6 = 36 / 6 = 6.
  2. 2. How is the standard deviation of one activity estimated in PERT?

    The range between pessimistic and optimistic is treated as six standard deviations.
  3. 3. How do you get the standard deviation of the critical path?

    Variances of independent activities add; the SD is the square root of the sum.
  4. 4. A known weakness of PERT is that...

    Parallel near-critical paths can each delay the end; simulation handles this.

Summary

  • PERT uses O, M and P estimates: TE = (O + 4M + P) / 6 and SD = (P - O) / 6.
  • Add the variances along the critical path to get the project's uncertainty.
  • Use the normal distribution to get the probability of meeting a date.
  • PERT understates risk when several paths are close; simulation improves on it.

Next: Gantt Charts and Timelines