Estimate Reconciliation and Scope Validation
You are the estimating manager for a refinery turnaround. The project scope includes replacing 200 valves, 500 meters of piping, and refurbishing a distillation column. The estimate was prepared by two estimators: Estimator X used a bottom‑up method and arrived at $4.2M; Estimator Y used a parametric method based on historical data and arrived at $3.6M. The difference is $600k (14%). The project manager wants to use the lower estimate to stay within budget. You have the following data: the parametric model has a correlation coefficient (R²) of 0.85 and was based on 10 similar projects; the bottom‑up estimate includes detailed takeoffs but may have missed some scaffolding and insulation work. Historical data show that similar turnarounds have an average cost overrun of 8% with a standard deviation of 5%.
1. Analyze the discrepancy between the two estimates. Which estimate is more reliable and why? What additional information would you seek?
Read the worked answer
The bottom‑up estimate is generally more reliable for a defined scope because it is based on specific quantities and unit costs. However, it may be incomplete if scaffolding and insulation were omitted. The parametric estimate has a good R² (0.85) but is based on only 10 projects and may not capture the specific complexities of this turnaround. The 14% difference is significant but within the range of estimate uncertainty for this stage. I would seek: a detailed comparison of the scope included in each estimate, verification of the quantities and unit costs in the bottom‑up estimate, an assessment of whether the parametric model’s database includes similar turnarounds, and a check for missing items like scaffolding and insulation. I would also consider using a range estimate (e.g., $3.6M to $4.2M) and applying a contingency based on historical overruns.
2. Given the historical overrun data, what contingency would you recommend for each estimate? How would you present the final estimate to the project manager?
Read the worked answer
Historical average overrun is 8% with a standard deviation of 5%. For a 50% confidence level (mean), contingency = 8% of the base estimate. For a higher confidence level, say 80% (z≈0.84), contingency = 8% + 0.84*5% = 12.2%. For the bottom‑up estimate of $4.2M, an 8% contingency is $336k, total $4.536M. For the parametric estimate of $3.6M, 8% is $288k, total $3.888M. However, the parametric estimate may be less accurate due to its limitations. I would present both estimates with their contingencies and explain the risks. I would recommend using the bottom‑up estimate as the base and adding contingency for missing scope, then comparing to the parametric estimate as a sanity check. The final estimate should be a range, e.g., $4.2M to $4.5M including contingency, and I would advise the project manager against using the lower estimate without addressing the missing scope.