The article draws attention to the problem of measuring the effect that programming solutions have on the programming productivity and performance, in educational programming and the correctness-saving program improvements. The results of some experiments concerning these questions are discussed. The hypothesis that  functional models can provide a metric scale capable of separating features of programming languages and systems from features of programs and programming solutions is proposed. The results of a preliminary demonstrative experiment in studying the dependence of the program productivity upon the opted compiler and, on the other hand, upon the representation of the programming solution in the opted programming languages are described. Analysis of these results leads to a method that can reveal such dependencies. The long experience in sifting educational and contest programs revealed some unnoticed aspects of this problem.