Estimating a task with one number hides the uncertainty. The PERT method asks for three: an optimistic, a most likely and a pessimistic duration. It combines them into an expected duration and a standard deviation, then gives the probability of meeting a deadline. This calculator does the arithmetic for a single task, or for a whole path if you enter its totals.
How the calculation works
- Expected duration = (optimistic + 4 × most likely + pessimistic) ÷ 6
- Standard deviation = (pessimistic − optimistic) ÷ 6
- 95% range = expected ± 2 × standard deviation
- Probability of meeting the deadline = normal distribution at (deadline − expected) ÷ standard deviation
Example
The optimistic estimate is 10 days, the most likely 15 and the pessimistic 30. The expected duration is 16.7 days with a standard deviation of 3.3 days. A 95% range runs from 10.0 to 23.3 days. With a 20-day deadline, the probability of finishing in time is 84.1%.
Why it matters
A deadline with a probability attached is a better basis for decisions than a single date. PERT makes teams think about what could go wrong, and shows how much a long pessimistic tail adds to the expected duration. For a chain of independent tasks, the expected durations add up, as do the variances.
Good to know
The method treats the beta distribution as roughly normal, which is only an approximation, and assumes independent tasks. Estimates are often over-optimistic, so ask for the pessimistic case first and compare the result with past projects. A single critical path hides parallel paths that can also run late, and in a real project the delays of tasks are often correlated, which makes the total less predictable than the independent model suggests. For a project with many tasks, apply the method to each and then add the expected durations along the critical path. This tool is for education.
Related tools
The earned value calculator tracks cost and schedule performance, the sprint velocity calculator forecasts a backlog, and the normal distribution calculators give probabilities for other thresholds.