PERT estimate calculator

How the PERT formula works

PERT (Program Evaluation and Review Technique) turns three guesses into one estimate with a measure of uncertainty. You give an optimistic duration O (everything goes right), a most likely duration M and a pessimistic duration P (things go wrong, short of disaster). The expected duration is a weighted average that counts the most likely value four times: E = (O + 4M + P) ÷ 6. The standard deviation is the width of the range divided by six: SD = (P − O) ÷ 6. Treating the result as roughly normal, about 68% of outcomes fall within E ± 1 SD and about 95% within E ± 2 SD.

Example: O = 4 hours, M = 6, P = 14. E = (4 + 24 + 14) ÷ 6 = 7 hours. SD = 10 ÷ 6 = 1.67 hours, so the 68% range is 5.3–8.7 hours and the 95% range 3.7–10.3 hours. The plain average (4 + 6 + 14) ÷ 3 would give 8 hours, because it gives the long pessimistic tail more weight. The table in the result converts the same numbers into one-sided targets: 90% of the time the task should finish within E + 1.28 SD, here 9.1 hours.

Getting useful three-point estimates

The method is only as good as the three inputs. The most frequent mistake is a pessimistic value that is too close to the most likely one; real tasks overrun far more often than they finish early, so P is usually much further from M than O is. Ask "what would make this take twice as long?" before settling on P. Estimates are also better when the person doing the work gives them and when tasks are small enough to picture, ideally under a few days each.

For a whole project, add the expected values of the tasks on the critical path, but combine the uncertainty by adding the variances (SD²) and taking the square root, not by adding the standard deviations. The normal approximation is a simplification: it works best for sums of many tasks and is rougher for a single one, which is why the 95% upper value, rather than E, is the safer number to quote when a deadline has consequences.