Example 1: TE 20, σ 2, deadline 22
z = (22 − 20) ÷ 2 = 1. Φ(1) ≈ 84.1%. These are the mini-calculator defaults. A deadline of 20 (z = 0) is about 50%.
Turn a PERT mean and standard deviation into a probability of hitting a deadline, and the duration that corresponds to 80, 90 or 95% confidence.
The chance of finishing by a deadline D is the normal CDF of z = (D − TE) ÷ σ. If TE is 20 days, σ is 2 and D is 22, z = 1 and the chance is about 84%. The model assumes a normal path total and independent tasks. It is a planning aid, not a guarantee.
z = (deadline − TE) ÷ σ, then Φ(z).
Once you have a path expected time TE and a path standard deviation σ, a deadline becomes a z-score: how many standard deviations the deadline sits above or below the mean. The normal curve then gives a probability. That is the same output as the probability section on the PERT calculator.
You need TE first; see expected time. A simulation range, which does not assume normality, lives on the project duration simulator.
z = (D − TE) ÷ σ
P(finish by D) = Φ(z)
If σ = 0, the probability is 1 when D ≥ TE and 0 otherwise. Duration needed for 80/90/95% one-sided confidence is TE + zσ with z ≈ 0.842, 1.282 and 1.645.
Each example uses numbers you can load into the calculator above; the results match the tool exactly.
z = (22 − 20) ÷ 2 = 1. Φ(1) ≈ 84.1%. These are the mini-calculator defaults. A deadline of 20 (z = 0) is about 50%.
For TE = 20 and σ = 2, the one-sided 90% duration is 20 + 1.282 × 2 ≈ 22.6. Commit to 20 and you are late about half the time in the model; commit to 22.6 and you are late about 10% of the time, if the assumptions hold.
Path totals are treated as normal. Tasks are treated as independent. Only one path is modelled. Shared vendors, weather and scope changes violate those. State them when you quote 84%.
Several near-critical paths, or skewed three-point estimates, are why the simulator exists. Compare P50 and P90 there with this z-score before you pick a date.
A 50% date is the expected time. Many teams cannot live with a coin-flip deadline, so they quote a one-sided duration: TE + 0.84σ for about 80%, TE + 1.28σ for about 90%, TE + 1.65σ for about 95% (normal table, one tail). Those are planning dates, not promises that history will match the table. If the organisation already uses P80 from Monte Carlo, do not mix a PERT 90% from this page with that P80 and call them the same thing. State the method. If you only have one path and three-point estimates, this z-score is the honest shortcut. If you have several near-critical paths, the chance of hitting the date is lower than any single-path z suggests, which is again why the simulator exists.
Write the deadline, TE, σ, z and the resulting probability in one sentence, plus the assumption list. Example: “On the current critical path TE is 40 days, σ is 4, a 45-day deadline is z = 1.25, about 89% in this model, assuming independent tasks and one path.” Then point to the PERT calculator so a reviewer can change P on the long-tailed task. Do not report “we are 89% sure” as if it were a measured hit-rate from past projects.
z = (deadline − expected time) ÷ σ, then the normal CDF of z. That is PERT’s usual shortcut for one path.
Planning often uses 80% or 90% one-sided durations, not the 50% mean.
O, M and P were equal. The probability is 100% if the deadline is at or after TE, otherwise 0%.
One path. Use the critical path, or simulate, if several chains could set the date.
This is an estimate before work starts. SPI is a forecast from earned value once work is underway.
The PERT calculator accepts a table and a deadline.
Written by Umesh Manral. Formulas checked against the site's test suite on 11 November 2026. Last updated 10 November 2026. Formulas and tests: methodology. Report a calculation error.