Risk calculators. See the number move.
Four working tools from our risk practice: contingency by project stage, the Standard Cost Category Beta Range Factor workbook, patent and latent contingency, and Latin Hypercube sampling. Change an input and the answer and graphs update.
Contingency by milestone
Generalized minimum contingency levels by project stage, applied to your base estimate.
Levels reflect historic risk under design-bid-build delivery. A project-specific risk analysis, such as our Beta Range Factor and Latin Hypercube method, should replace this rule of thumb as the design matures.
Beta Range Factor
Pick a category and the project's stage. The factors for requirements, design, market and construction pop up and add to the Beta Range Factor, then the cost distribution is drawn.
The published standard factors cover SCC 10 to 50. For SCC 60 to 80 the range factor is set case by case, so it is an input here.
Method: Beta Range Factor cost risk by Standard Cost Category, as described in FTA Oversight Procedure 40C. BRF = 1.05 + requirements + design + market + construction + post-construction. Cost = estimate × (1 + (BRF − 1) × Beta(1.5, 3.5)). Categories 90 and 100 are not modeled. Illustrative application, not an agency determination.
Name the risks, then price what you cannot name
Patent risks are identified and priced one by one. Latent risk is the allowance for what is still unidentified. Edit the register and a 2,500-trial Latin Hypercube simulation reruns.
| Patent risk | Prob. % | Min $M | Likely $M | Max $M |
|---|
Each patent risk fires with its probability and, when it does, draws a cost from a three-point Beta-PERT distribution. Illustrative; replace the sample register with your own.
Same trials, steadier answer
Plain random sampling leaves holes and clumps. Latin Hypercube puts exactly one point in every row and every column. Move the slider and compare coverage, then see how much steadier the P80 estimate becomes.
The stability test repeats each estimate 60 times on a skewed cost function and measures how much the P80 moves between repeats. Smaller is steadier. This is why we run Latin Hypercube rather than plain Monte Carlo.
Need the real number, defended?
These are starting points. A simulation-based risk analysis or a forensic delay analysis gives you a figure you can defend to a board, an auditor or a dispute board.