Before you start
Three-Point Estimate Calculator
Turn optimistic, most likely and pessimistic figures into a number you can plan to, for one task or a whole list, without the arithmetic that quietly makes every plan late.
Type the figures in, paste rows from a spreadsheet, or read a .xlsx or .csv inside your browser. Nothing is uploaded.
5 days
Expected, PERT
(O + 4M + P) / 6. Triangular reads it as 6 days.
6.5 days
80% confident, PERT
Triangular: 8 days. Four times in five, the task comes in at or under this.
33%
Chance of coming in by the most likely
The most likely figure is missed more often than it is met. Triangular: 20%.
PERT and triangular, side by side
- PERT
- Triangular
- Most likely
| Figure | PERT | Triangular |
|---|---|---|
| Expected (mean) | 5 days | 6 days |
| Standard deviationThe textbook shortcut (P - O) / 6 gives 1.67 days, narrower than the PERT curve itself. | 1.73 days | 2.16 days |
| P10 | 2.9 days | 3.4 days |
| P50, as likely over as under | 4.8 days | 5.7 days |
| P80 | 6.5 days | 8 days |
| P90 | 7.5 days | 9.2 days |
| Chance of coming in by the most likely | 33% | 20% |
What this assumed, and how
- PERT reads the three figures as a beta curve (shape 1.80, 4.20) with mean (O + 4M + P) / 6. The weight of four on the most likely figure is a convention, not a measurement.
- Triangular reads them as the straight-line shape through the three points, with mean (O + M + P) / 3. It gives the extremes more weight, so it reads longer whenever the pessimistic figure is further from the most likely than the optimistic one is.
- P80 is the figure the task comes in at or under four times in five, read from each curve directly. No simulation is needed for one task.
- The optimistic and pessimistic figures are treated as the true ends of the range. If yours are softer guesses, the real range is wider than either curve shows.
The sum of the most likely figures is a plan you will usually miss
Ask for a most likely figure and people give the one they would bet on. For almost any real task, the ways it can run long outnumber the ways it can run short: a review comes back with changes, a key person is out, a vendor is slow. So the range stretches further on the pessimistic side, and the task’s average sits above its most likely figure.
Add up a list of most likely figures and every task contributes that gap. On the sample list of twelve tasks, the most likely figures add up to 63 days and the simulated project comes in within 63 days about 7 times in 100. The plan was never wrong about any one task. It was wrong about the total.
The sum of the pessimistic figures is not a worst case either
The mirror-image mistake is adding every pessimistic figure and calling it the worst case. That total needs every task to hit its worst figure at once, which does not happen: some go badly, some go well, and they partly cancel. On the sample list the pessimistic figures add up to 134 days, and the longest of ten thousand simulated projects took under 100. A plan padded to the sum of pessimistic figures is money and time nobody needed.
What the simulation does
It draws a value for every task from that task’s own range, adds them up, and does it again ten thousand times. Then it counts: how often the list came in within the most likely total, what total four runs in five came in under (the P80), and so on. The random numbers are seeded, so the same list gives the same answer tomorrow and on somebody else’s computer.
PERT and triangular are two readings, not a right answer and a wrong one
PERT, the (O + 4M + P) / 6 formula, reads the three figures as a smooth curve that leans heavily on the most likely figure. The weight of four is a convention from the 1950s, not something anybody measured. Triangular reads them as a straight-line shape and takes the extremes more seriously, so it reads longer whenever the pessimistic figure is far out. Both are shown throughout. If the difference between them matters to your decision, that is worth knowing too.
Tasks that run late together
The textbook sum assumes each task’s luck is its own. In a real project, tasks share causes: the same overloaded team, the same slow approver, the same requirement nobody has pinned down. When one of those goes wrong, several tasks run late together, and the total’s range is wider than independent tasks would give. The setting for this is on the page, and the result always shows what the P80 would be under the other assumption, so the choice is visible rather than buried.
Where the uncertainty comes from
A long list usually has a few tasks carrying most of the uncertainty. The tool ranks every task by its share, which says where effort is best spent: splitting the widest task, or answering the question that makes it wide, narrows the total more than tightening everything else.
What this does not do
It adds durations end to end. Tasks that run side by side need a schedule, not a sum, and adding their durations overstates the time the work takes. It cannot tell you whether your three figures are any good: a range guessed in a hurry gives a precise-looking answer to the wrong question. And it does not choose the contingency for you. Whether to plan to the P50, the P80 or something else depends on what a late finish costs, which only you know.