Some of you may have already noticed that my article about software cost estimation has not been developed further for some time. One argument, as always, is the time available for this. There are always enough other tasks – professional, private, health-related, CMS-changing. But that’s not all.
The series of articles is based on a larger training course I gave, in which I initially pointed out the problems of estimation, the difference between effort and time forecasts, and the multitude of parameters that influence this. It continues with formal procedures such as Function Point Analysis and COCOMO II. But FPA leads to the expected amount of code based on what is to be developed. And COCOMO II converts this number into the expected effort. This is influenced by the nature of the project and a variety of parameters. It is a pure joy to tweak these parameters. A bit of staff shortage here, a bit of time pressure there, a cherry on top due to a lack of skills – delicious.
Well, what about today? The function points remain; they are the requirements with their cornerstones. And with COCOMO II, many of the parameters remain. But what about the glue SLOC, in today’s age of AI? And the broad parameters are also influenced more or less intensively by the use of AI. In my opinion, the coming path will be from analysis via FP-AI to an evaluation of the resulting abstract weight in its parameters to a range of time as the framework for estimation. Perhaps with probabilities, but that’s it.
You’ll find my new approach here, I’ll let you know.
