Page 138 -

7 - PROJECT COST MANAGEMENT

7.2.2.3 Parametric Estimating

Parametric estimation tools for software projects typically include an estimation algorithm with adjustment
factors for specific cost drivers. Most estimation tools for software projects use some measure of product size as the
primary input variable, such as estimated number of function points or number of use cases. Parametric estimation
tools can be calibrated for the specific software development organization, infrastructure tools, complexity of the
software to be developed, and experience or ability of the team, or a provided calibration that most closely matches
the characteristics of the project being estimated can be used.

Calibration of a parametric estimation tool using local, historical data is preferred to the use of provided
calibrations because local data will include factors unique to the local organization and the software produced by
the organization; also, since the methods and tools used on projects are frequently changed for newer technology. 7

7.2.2.4 Bottom-Up Estimating

Bottom-up estimation is often used to estimate effort and cost of software projects. Estimates are made for
individual software components and rolled up. The cost of integration and testing of the software components need
to be added when bottom-up estimates are made only for the effort needed to develop the software components.
Additional costs for project management, quality assurance, configuration management, and other project cost
factors should also be included.

7.2.2.5 Three-Point Estimating

Estimating software size or effort can be based on expert judgment and the three-point PERT algorithm.
Using this approach, experts estimate the size or effort for individual software components as small, (e.g., 20%
probable), medium (50% probable), or large (e.g., 80% probable)—the percentages for small and large depend
on parameters in the PERT algorithm. The PERT algorithm is used to estimate the mean and standard deviation of
size or effort for each software component and the mean and standard deviation for the collection of components;
probability distributions for effort or size can be calculated from the mean and standard deviation. Also, the mean
and standard deviation of a size estimate can be used as inputs to a parametric estimation algorithm to compute
a probability distribution for effort. Other forms of statistical estimation, including Monte Carlo simulation, can be
used to overcome some of the undesired effects of three-point estimation, such as nodal bias or merge bias that
occurs when applying three-point estimation to an activity network.

7.2.2.6 Reserve Analysis

Estimates can be made to establish cost and schedule reserves to be included in the project estimate and the
project budget. Past projects can be analyzed to support the amount of reserve that should be included for a new
project by determining the difference between the known effort (cost) at the start of the previous projects and the
amount of effort (cost) that was eventually required to complete the project.

©2013 Project Management Institute. Software Extension to the PMBOK Guide Fifth Edition 129
®