Accuracy Range & Contingency from Parametric Tables
A beginner's guide to determining cost-estimate accuracy range and contingency from parametric tables — pre-built lookup tables, derived from rigorous risk models, that let you read a defensible contingency in seconds. Built on AACE International RP 119R-21.
What parametric tables are
The premise is elegant: rather than build a full risk model for every estimate, an organization (or AACE) does the rigorous analysis once across a representative range of projects, then distils the results into tables. A practitioner faced with a new estimate simply identifies its characteristics, finds the matching row, and reads off the accuracy range and contingency. It trades a little precision for enormous speed and consistency.
Reading the table
A parametric contingency table is typically organized by estimate class (the level of project definition you met in Cost Estimating) and often by project type. Here's an illustrative example tying class to accuracy range and a contingency to reach a target confidence:
| Estimate class | Definition | Accuracy range | Contingency (P50→P80) |
|---|---|---|---|
| Class 5 | 0–2% | −30% / +50% | ~35% |
| Class 4 | 1–15% | −20% / +30% | ~25% |
| Class 3 | 10–40% | −15% / +20% | ~15% |
| Class 2 | 30–70% | −10% / +15% | ~10% |
| Class 1 | 50–100% | −5% / +10% | ~5% |
For a Class 3 estimate (highlighted), the table says: expect accuracy of about −15% to +20%, and carry roughly 15% contingency to reach a comfortable confidence level. No modeling required — just identify the class and read across.
Accuracy range and contingency are linked
The deep point of this RP is that an estimate's accuracy range and its required contingency are two views of the same underlying uncertainty distribution. The accuracy range (e.g. −15%/+20%) describes how far the actual cost might land from the point estimate; the contingency is the amount you add to move from the base (around P50) up to your chosen confidence level on that same distribution.
Using parametric tables well
Use tables that were derived from models and project data relevant to your work — ideally your own organization's history. Identify the project's class and type honestly, read the baseline contingency, and document where you're using it as-is versus overriding for project-specific risks. Tables are excellent for early estimates, for consistency across a portfolio, and as an independent check on a bespoke model's output.
Nine things to remember
- Parametric tables give contingency from a few simple descriptors.
- They're pre-computed once from rigorous models, then looked up.
- They trade precision for speed, consistency, and auditability.
- Tables are usually organized by estimate class and project type.
- Example: a Class 3 estimate → ~15% contingency.
- Accuracy range and contingency are two views of one distribution.
- Definition drives both — the estimate-class idea once more.
- Tables describe the typical project — override for unusual risks.
- Great for early estimates, portfolios, and as a model cross-check.
Glossary
- Accuracy range
- How far actual may land from the estimate.
- Baseline
- The table value before project-specific overrides.
- Confidence level
- The P-value the table's contingency targets.
- Estimate class
- Level-of-definition category (Class 5→1).
- Override
- Documented departure for unusual risk.
- Parametric table
- Lookup of contingency by project descriptors.
- Point estimate
- The single base number before contingency.
- Standardization
- Consistent, repeatable treatment across estimates.