Three-Point Cost Estimate Calculator
Combine three-point construction cost estimates into an expected total and a quantified uncertainty screen. Choose conventional PERT or triangular item formulas, then inspect the independent variance sum, standard deviation and a central 90% or 95% normal approximation range.
By Dr. Hassan Khames EliwaLast updated
Result
- Independent variance totalSquared currency units; no covariance included.
- 1,388.89
- Standard deviation
- 37.27
- Normal range lower
- 226.96
- Normal range upper95% central normal approximation, not a guarantee; negative lower limits are not clipped.
- 373.04
Per-item estimates in a single consistent currency
| Item | Mean | Variance |
|---|---|---|
| 1 | 200.00 | 1,111.11 |
| 2 | 100.00 | 277.78 |
How it works
Give each estimate a defensible basis
Enter optimistic cost O, most likely cost M and pessimistic cost P for each item, with O ≤ M ≤ P. All values must be nonnegative and use the same currency and price date. Each row should represent a distinct cost exposure rather than duplicating a risk already included elsewhere. Inputs accept up to 200 rows and one trillion per amount, with no currency symbols, headings or thousands separators. The distribution choice applies to every row in the run.
Choose the moment formulas
Conventional PERT uses mean (O + 4M + P) ÷ 6 and variance ((P − O) ÷ 6)². This is the familiar three-point planning approximation, not a fitted probability model or a uniquely calibrated beta distribution. The triangular option uses mean (O + M + P) ÷ 3 and variance (O² + M² + P² − OM − OP − MP) ÷ 18. The implementation evaluates an equivalent difference-based triangular formula to reduce cancellation when the three costs are close together. Equal inputs give zero variance under either option.
Aggregate only under an explicit independence assumption
Expected total cost is the sum of item means. The independent variance total is the sum of item variances, and standard deviation is its square root. Variance has squared currency units, while standard deviation has ordinary currency units. Correlated steel prices, shared exchange rates or common weather exposure require covariance terms, which this tool does not request. Positive dependence can materially increase total uncertainty even when every individual three-point estimate is reasonable.
Interpret the normal range as an approximation
The selected central range is total mean ± z × standard deviation. The 90% option uses z = 1.6448536269514722; the 95% option uses z = 1.959963984540054. These are two-sided ranges, not one-sided contingency percentiles. A few skewed or dominant items may make a normal approximation poor. Limits can extend below zero or beyond the sum of input endpoints; they are deliberately not clipped because clipping would conceal that distribution mismatch. The range describes modeled cost uncertainty, not a statistical confidence interval for a mean estimated from observed samples.
Worked example
Scenario 1: conventional PERT at 95%
Enter rows 100,200,300 and 50,100,150, select PERT and 95%. The item means are 200.00 and 100.00, giving expected total cost 300.00. Item variances are 1,111.11 and 277.78, with the unrounded independent variance sum 1,388.89 when displayed. Standard deviation is 37.27. Applying the normal factor produces a lower range limit of 226.96 and upper limit of 373.04. Calculations retain unrounded moments; rounded table entries are not reused as inputs.
Scenario 2: triangular at 90%
Use one row 100,200,300, select triangular and 90%. Expected cost is 200.00, variance is 1,666.67 and standard deviation is 40.82. The central normal approximation runs from 132.85 to 267.15. A triangular distribution for this single item is not normal, so this displayed range is not the exact triangular percentile interval. It is useful for checking the selected approximation, not as evidence that the approximation fits a single-item estimate.
Expert tips
Common mistakes
- Do not add item standard deviations as though they were independent variances. Independent variances add; the square root is taken only after the variance sum.
- Do not call the upper endpoint a guaranteed budget cap. It depends on your inputs, distribution formulas, independence assumption and normal approximation.
- Do not mistake additional rows for additional information. Splitting one shared risk into many independent-looking items can artificially reduce the modeled relative uncertainty.
Frequently asked questions
Is the PERT variance an exact fitted beta variance?
No. This calculator uses the conventional range-over-six planning approximation. It does not infer beta shape parameters or calibrate a distribution from observed costs.
Why does triangular usually show more spread?
For symmetric inputs its variance is range squared divided by 24, compared with range squared divided by 36 under conventional PERT. Different skewness changes the triangular result.
Can I mix currencies?
No. Convert all costs to one currency and one price basis before entry. Exchange-rate uncertainty itself may need a shared risk model.
Does independence matter for the mean?
Expected item costs can still be added without independence. Independence matters here for adding variances without covariance terms and therefore for the reported standard deviation and range.
Is 95% the same as a P95 budget?
No. This is a central two-sided range. Under the normal approximation its upper endpoint is the 97.5th percentile, whereas the 90% range ends at the 95th percentile.
What if every item is certain?
Equal optimistic, most likely and pessimistic inputs produce zero variance. The standard deviation is zero and both range endpoints equal the total expected cost.
Related calculators
Related guides
Construction & ProjectCPI and SPI trends: read the direction, not just the numberRead period and cumulative CPI and SPI together, test the data behind changing trends, and turn earned-value signals into practical decisions.16 November 2026 · 8 min read
Construction & ProjectResource histograms and levelling: spot the peak earlyBuild a resource histogram, distinguish levelling from smoothing, and test whether a flatter labour plan is genuinely deliverable.16 November 2026 · 8 min read
Construction & ProjectCrew size, man-hours and productivity: labour arithmeticConvert scope and production rates into labour-hours and crew sizes, then test utilization, whole-shift bookings and labour cost.16 November 2026 · 8 min readSources 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