PERT Calculator

One estimate is a hope; three estimates, honestly gathered, are the beginning of a plan. PERT — the Program Evaluation and Review Technique, born in the US Navy’s Polaris programme — prices a task from its optimistic, most likely and pessimistic durations, weighting the middle four times the extremes. This calculator returns the expected duration and its standard deviation, the two numbers every honest schedule risk conversation is built from.

Why the Most Likely Counts Four Times

The weighting (O + 4M + P)/6 is not arbitrary numerology; it approximates the mean of a beta distribution stretched between O and P with its mode at M — the shape project durations actually take (a hard floor at the optimistic end, a long tail of trouble at the pessimistic end, a hump of likelihood in between). The plain average of the three estimates would grant the extremes equal citizenship and systematically flatter tasks whose risk is skewed; the 4-weight anchors the estimate near the realistic case while letting the spread speak through the second output. That standard deviation — (P − O)/6 — treats the optimistic-to-pessimistic range as roughly six sigmas of the underlying distribution, converting the estimators' gut-feel range into the schedule's first honest measure of uncertainty.

Worked Example: Four, Seven, Sixteen

The defaults are a task estimated at 4 units of time if blessed, 7 if normal, 16 if cursed. Expected duration = (4 + 28 + 16)/6 = 8.00 — pulled a full unit above the "most likely" 7 by the pessimistic tail, which is PERT's signature correction to human optimism. Standard deviation = (16 − 4)/6 = 2.00. The reading the calculator prints: roughly 68% confidence of landing between 6 and 10, roughly 95% between 4 and 12, and a promise made at 8 is a coin flip — the manager who quotes 10 (E + 1σ) is quoting a date with about an 84% pedigree, though neither figure should be mistaken for a guarantee, as the limitations section insists.

From Tasks to Projects: Adding Up the Chain

PERT's real jurisdiction is the network: a project is paths of tasks, the critical path is the longest, and the project's expected duration is the sum of the expected durations along it — expected values add, which is the arithmetic that makes the whole method tractable. Variances also add (for independent tasks), so the project's standard deviation is the square root of the summed task variances along the critical path — narrower, relatively, than any single task's spread, because independent troubles partially cancel. Two professional warnings ride along: the critical path itself can change when tasks vary (merge bias — parallel near-critical paths make the project later than its nominal path suggests), and task estimates from the same estimator are rarely independent in their errors. The calculator prices the task; the network sums belong to the project plan, built from these outputs task by task.

Honest Estimating: Feeding the Three Numbers

The formula is only as honest as its inputs, and the elicitation craft matters more than the arithmetic. Optimistic means "everything relevant goes right" — not the fastest ever seen, not a fantasy; pessimistic means "everything that reasonably can go wrong, does" — explicitly excluding acts of God, or the tail swallows the estimate; most likely means the mode of realistic repetitions, the duration you'd bet the median on. Anchor against reference classes where they exist (the last ten similar tasks out-argue any single estimator's imagination), estimate before discussing deadlines (a date heard first bends all three numbers toward it), and record who estimated what — PERT's Navy origins valued the audit trail as much as the average. Garbage in the three boxes produces a six-decimal expected duration of garbage, which this page will compute with perfect, misleading precision.

A Second Rehearsal, In Money

Run the machinery once on cost, the domain where its correction matters most. A work package estimated at $8k optimistic, $10k most likely, $26k pessimistic — the classic shape of software and construction estimates, with their fat right tails. PERT: E = (8 + 40 + 26)/6 = $12.33k, a 23% uplift on the "most likely" figure that naive budgets would have carried; SD = (26 − 8)/6 = $3k. The uplift is not pessimism but symmetry-breaking: the estimator's own three numbers already admitted the downside was three times the upside's distance, and the formula merely refuses to forget what the estimator knew. Budgets built from most-likely figures alone are, by this arithmetic, plans to overrun by a predictable margin.

PERT and CPM: The Sibling Methods

PERT is forever introduced alongside its twin, and the contrast clarifies both. CPM (Critical Path Method, DuPont, 1957) assumes durations are knowable and prices the path: one estimate per task, the longest chain rules the schedule, and management attention goes to crashing that chain. PERT (Navy, 1958) assumes durations are uncertain and prices the uncertainty: three estimates per task, expected values along the path, variances for the risk conversation. A year apart, an industry apart — chemical plants had repeatable tasks; a first-ever missile programme had none — the twins encode their birthplaces. Modern practice merges them (CPM networks with PERT-weighted tasks, Monte Carlo simulation over both), and this calculator supplies the PERT ingredient: the per-task mean and variance that any of those more elaborate engines consume as their raw material.

A Whole Path, Summed in Miniature

The network arithmetic, rehearsed on three tasks so the project section's claims are checkable. Tasks on one path: the worked example (E 8.00, SD 2.00), a second (E 5.00, SD 1.00), a third (E 3.50, SD 0.50). The path's expected duration adds the means: 8 + 5 + 3.5 = 16.50. Its variance adds the variances: 4 + 1 + 0.25 = 5.25, so the path SD is √5.25 ≈ 2.29 — noticeably less than the naive sum of the SDs (3.50), because independent wobbles partly cancel: the mathematics of why big projects are, relatively, more predictable than their riskiest single task. A promised date at 16.5 + 2.29 ≈ 18.8 carries the one-sigma pedigree of about eighty-four percent, computed exactly as the single-task note in the calculator above describes.

Related Tools

More from the Other Calculators hub: the DPMO Calculator — project quality in the same six-sigma dialect — and the Engagement Rate Calculator for a different department's favourite ratio.

Frequently Asked Questions

What is the PERT formula?

Expected duration E = (O + 4M + P) ÷ 6, with standard deviation (P − O) ÷ 6 — two formulas, four arithmetic operations, and the whole of the classical method between them. The weights approximate a beta distribution's mean from its endpoints and mode — the 1958 Polaris programme's solution to pricing tasks nobody had ever performed, where single-point estimates had proved worthless and ranges were all honest engineers would swear to. This calculator is that formula with its confidence reading attached.

How is PERT different from a simple three-point average?

The simple average (O + M + P)/3 weights the extremes as heavily as the realistic case: on the worked example it returns 9.00 against PERT's 8.00, dragged upward by granting the 16 full citizenship. Which is "right" depends on the true distribution's shape; decades of project data favour the beta-like weighting PERT encodes — most likely durations should dominate, extremes should inform but not govern.

What date should I actually promise?

Read the standard deviation as the price list of confidence: E is about a 50% date, E + 1σ about 84%, E + 2σ about 97.7% (reading the spread as roughly normal). Organisations differ in appetite — bid work is priced leaner than internal commitments — but the discipline is to choose the confidence knowingly, from these numbers, rather than promising the expected value and calling the resulting coin flip a plan.

Can PERT handle cost estimates too?

Yes — the three-point method prices any uncertain quantity: cost, effort, defect counts, durations. E and SD come out in the input's units (dollars, person-days), and chains of cost estimates sum exactly as duration chains do. What changes by domain is the elicitation: cost pessimism (overruns) behaves differently from schedule pessimism, and reference-class data matters even more, since cost optimism is the best-documented bias in all of project management.

Why is the standard deviation (P − O)/6?

Because the method models the range as spanning about six standard deviations (±3σ) of the underlying beta distribution — a deliberate, slightly conservative convention from the original PERT papers. Its practical virtue is that the whole uncertainty estimate hangs on the two numbers estimators can actually introspect (the believable extremes), rather than asking them for a variance they cannot feel. Modern practice sometimes fits fuller distributions; the (P−O)/6 figure remains the field's common currency.

What are PERT's known weaknesses?

Merge bias at parallel paths (addressed above), correlated estimation errors, the beta assumption itself on tasks whose real distributions are bimodal (a task that either takes a day or a month, nothing between — its "most likely" is a coin's edge), and the eternal garbage-in problem. Monte Carlo simulation over the network is the modern successor where the stakes genuinely justify it; PERT remains the method of first resort because its inputs are three numbers a human can defend in a meeting, on a whiteboard, in front of the people who must live with the resulting date.