There are several software products that from a program model or its source code, cal-culate its SIZE in function points. However, comparisons between the results produced by different tools for the same system, frequently have an unacceptable variation. These numbers, also often differ greatly from a manual count.

The answer to this variation is in how these tools calculate the number of function points. Some are based on files, screens, reports and other elements to derive a num-ber. Although there is often a DIRECT relationship between these objects and data functi-ons and transactions functions of Function Point Analysis (FPA), it must be remembe-red that the technique measures only the logical functions of the system. And these tools have difficulties in differentiating logic functions from physical functions. For EXAMPLE, not every file or table from a program file corresponds to an internal logical file or external interface file. Or even an elementary process can be implemented through multiple screens. To do the measurement in a correct way, the software should have enough intelligence to make this judgment. That is, this software would have to have the skill to read the program and interpret the USER´s requirements. However, there is no software with this artificial intelligence.

Other tools are based on the BACKFIRING technique, which is to derive the number of function points from the program number of lines of code, based on a previous relati-onship established between LOC and FP. However, this is a technique that has been widely criticized, and whose application is restricted.

There are software products to support the process of counting function points that automate a part of the process, but the decision and analysis of that should be consi-dered, remains as the responsibility of the human user who enters the data, and not of the software.

There are several software products that from a program model or its source code, cal-culate its size in function points. However, comparisons between the results produced by different tools for the same system, frequently have an unacceptable variation. These numbers, also often differ greatly from a manual count.

The answer to this variation is in how these tools calculate the number of function points. Some are based on files, screens, reports and other elements to derive a num-ber. Although there is often a direct relationship between these objects and data functi-ons and transactions functions of Function Point Analysis (FPA), it must be remembe-red that the technique measures only the logical functions of the system. And these tools have difficulties in differentiating logic functions from physical functions. For example, not every file or table from a program file corresponds to an internal logical file or external interface file. Or even an elementary process can be implemented through multiple screens. To do the measurement in a correct way, the software should have enough intelligence to make this judgment. That is, this software would have to have the skill to read the program and interpret the user´s requirements. However, there is no software with this artificial intelligence.

Other tools are based on the backfiring technique, which is to derive the number of function points from the program number of lines of code, based on a previous relati-onship established between LOC and FP. However, this is a technique that has been widely criticized, and whose application is restricted.

There are software products to support the process of counting function points that automate a part of the process, but the decision and analysis of that should be consi-dered, remains as the responsibility of the human user who enters the data, and not of the software.