I want to argue here for a new FPA manual v3.0. The proposed changes are:
- Abolish the detailed analysis
- Rename the high-level (global) analysis to FPA analysis
- Introduce an indicative analysis, based on the function model, in addition to the current indicative analysis, based on the data model
The reasons for abolishing the detailed analysis (and renaming the high-level analysis) are as follows:
For a project calculation purposes a detail count is irrelevant
To create project calculations there are two main variables, the product size (function points) and the Product Delivery Rate (hours per function point). Product Delivery Rates are very inaccurate, usually 10% or more. The difference in inaccuracy between a high-level analysis and a detailed analysis is 1%. Because mostly the product size will vary in time, this inaccuracy difference is negligible.
Detailed analysis is (often) not possible
A detailed analysis requires insight into entities and attributes. At a global design that is not available. Many features in the (current) detailed designs also contain insufficient detail information for a detail count.
Detailed analysis increases complexity and reduces accountability
There is a balance between complexity (more details) and accountability (overview). A detailed analysis increases the complexity and greatly reduces the accountability (understanding from the end users). A detailed analysis costs much more (extra) effort, while it delivers less value.
The FPA Manual is needlessly complex and therefore FPA is less used
By removing the detailed analysis from the FPA Manual this manual will become less complex and therefore more accessible.
The reasons for the introduction of an indicative analysis, based on a global function model are:
At the start of a project there is more often a function model than a data model. An indicative analysis, based on a mathematical function model would help. The principles for such a model are:
- Estimate and / or count all input and output functions
- Calculate the product size based on all functions * (xxx function points)
- Determine the inaccuracy of the estimate product size
I would also like to know if there are other suggestions for improvements for a new FPA manual version 3.0.
About the author
Marten Eisma is Information Architect at CGI in the Netherlands and is a member of the Nesma Basis of Estimate committee.