Three-point estimating: when PERT beats a single guess
A practical guide to three-point duration estimates, weighted averages and the limits of turning uncertain activities into a promised date.
By Dr. Hassan Khames EliwaLast updated

A single duration hides the conversation you need
Ask how long an activity will take and someone will give you a number. Ask what could make it faster or slower and you usually learn much more.
Three-point estimating captures that second conversation. Instead of recording only a preferred duration, you record an optimistic duration, a most likely duration and a pessimistic duration. PERT combines them into a weighted planning estimate.
This is useful when uncertainty matters to a decision: mobilizing specialist equipment, reserving an outage, sequencing inspections or committing to a client milestone. It is less useful when the inputs are invented merely to make an existing promise look analytical.
Define the activity before estimating it
Consider installing and testing a small equipment package. Does the duration include delivery, lifting permits, electrical termination and commissioning? Are weekends available? Is the crew dedicated?
Unless everyone estimates the same scope under the same working conditions, the three numbers are not comparable.
Write down:
- The activity's start and finish conditions.
- The quantity and required quality standard.
- Crew size, equipment and working calendar.
- Access, permits and predecessor assumptions.
- Which risks are included and which are modeled separately.
Distinguish effort from duration. A task requiring several people does not necessarily become proportionally shorter when more people join it. Congestion, specialist availability and fixed waiting periods can dominate the calendar.
Choose three defensible points
The optimistic estimate describes a favorable but credible outcome. The most likely estimate describes the outcome you consider most plausible under the stated conditions. The pessimistic estimate describes an unfavorable but credible outcome.
For a simplified PERT model, these are treated as bounds and a mode-like central input. In real estimating workshops, people sometimes use specific percentiles instead. Do not quietly substitute percentile judgments into a bounds-based formula and assume the statistical meaning remains unchanged.
Suppose an activity has:
| Input | Duration | Supporting assumption |
|---|---|---|
| Optimistic, a | 4 days | Clear access and uninterrupted installation |
| Most likely, m | 7 days | Normal coordination and testing |
| Pessimistic, b | 16 days | Credible rework and access interruption |
The long upper tail matters. There is more room for disruption to extend the activity than for excellent performance to shorten it.
Calculate the PERT estimate
The conventional PERT approximation is:
Expected duration = (a + 4m + b) / 6
For this activity, (4 + 4 × 7 + 16) / 6 = 8 days.
The expected duration exceeds the seven-day most likely estimate because the unfavorable tail is longer. An expected value is a model-based average, not the duration that must occur most often.
The common PERT spread approximation is:
Standard deviation = (b − a) / 6
Here, (16 − 4) / 6 = 2 days, giving a variance of 4 days squared.
Use the PERT calculator to check the weighted duration and spread. Keep the assumptions with the result; a number without its estimating basis is difficult to update responsibly.
These are conventional approximations, not proof that the activity follows a particular real-world probability distribution.
Why not take a simple average?
The unweighted average gives each of the three points equal influence. PERT deliberately gives the most likely estimate more weight.
That can be helpful when the most likely duration comes from credible production knowledge and the outer values represent less common conditions. It is less persuasive when the middle estimate is merely a manager's target.
A weighted formula cannot correct biased inputs. If the optimistic duration assumes perfect access while the most likely duration quietly assumes overtime that has not been approved, the calculation mixes incompatible scenarios.
The real advantage of three-point estimating is often the explicit discussion of uncertainty. The extra arithmetic is secondary.
Combine activities carefully
Suppose three activities must happen in sequence, without overlap. Their estimates are:
| Activity | Optimistic | Most likely | Pessimistic | PERT mean | PERT standard deviation |
|---|---|---|---|---|---|
| Installation | 4 days | 7 days | 16 days | 8.00 days | 2.00 days |
| Connections | 3 days | 5 days | 9 days | 5.33 days | 1.00 day |
| Testing | 2 days | 3 days | 8 days | 3.67 days | 1.00 day |
Using unrounded values, their expected durations total 17 days.
If durations are independent, their variances add: 4 + 1 + 1 = 6 days squared. The standard deviation of the total is the square root of 6, approximately 2.45 days.
Do not add the individual standard deviations. Do not calculate from rounded table values when the underlying values are available.
The independence assumption is often questionable. The same access restriction, weather pattern or engineering issue may affect several activities. Shared risks can increase the spread of the combined outcome.
A project is not always one fixed path
Adding means along a fixed sequence is straightforward. Forecasting a whole project network is harder.
Parallel paths may compete to become critical. If one path finishes early, another may determine completion. Resource constraints can also create relationships that the schedule logic does not fully express.
A simple PERT calculation for the currently critical path can therefore understate project completion risk. The GAO Schedule Assessment Guide discusses schedule risk analysis and the importance of a credible underlying schedule.
For consequential commitments, consider a properly built schedule risk model, including correlations, alternative paths and discrete risk events. Simulation is not automatically better: it still depends on scope, logic and realistic inputs.
Do not sell a rough range as a guarantee
For the single activity, the expected duration plus or minus two estimated standard deviations is 4 to 12 days.
That arithmetic does not establish a reliable confidence level. The familiar normal-distribution interpretation requires assumptions that may not hold for a skewed activity duration, and the conventional PERT spread itself is approximate.
The pessimistic input is 16 days, so credible modeled outcomes extend beyond that two-standard-deviation band.
The standard deviation calculator is useful for describing observed durations from comparable completed tasks. Its sample standard deviation is not the same calculation as PERT's range-based approximation. Historical data and expert estimates serve different purposes.
Use the estimate to make a better decision
A useful estimate should change something. Perhaps you arrange an earlier inspection, protect an access window or avoid booking downstream crews against the most optimistic outcome.
Separate the expected working duration from contingency and the contractual commitment. Otherwise, teams may hide allowances inside activity estimates and add another unexplained buffer above them.
Update the inputs when assumptions change. Securing the equipment delivery may narrow the range; discovering an unresolved design interface may widen it.
Once execution begins, compare installed progress and incurred cost using the earned value calculator. PERT helps frame uncertainty before and during planning; earned value helps assess performance against the approved plan. Neither replaces field evidence.
A strong estimate is not the one with the most decimal places. It is the one whose assumptions survive questioning and whose uncertainty is visible before someone commits to a date.
Calculators used in this guide
Frequently asked questions
Is the PERT expected duration the most likely finish time?
No. It is a weighted mean from the three inputs. The most likely input represents a different concept, and the two can differ substantially when the range is asymmetric.
Can I use three-point estimating for costs?
Yes, if all three inputs describe the same cost scope and price basis. Be careful with shared risks and correlations when combining multiple cost estimates.
What if all three estimates are equal?
The formula returns that common value and a zero spread. This describes the inputs, not proof that the real activity has no uncertainty.
Can I add PERT durations for parallel activities?
Not to obtain project elapsed time. Parallel and sequential activities must be evaluated through their schedule relationships, calendars and resource constraints.
Does PERT provide a guaranteed completion date?
No. It provides a planning estimate based on assumptions. A credible commitment also needs schedule logic, risk analysis, resource checks and a deliberate contingency policy.
Sources and standards
About the author
Dr. Hassan Khames Eliwa holds a PhD in Construction & Built Environment and has more than 20 years of international project controls experience. He is Director of Eliwa Project Controls Limited in Auckland and founder of PMMilestone.org and PMMilestone.com.
Buy me a coffee