Abstract : Constraint Programming (CP) is a general technique for solving combinatorial optimization problems. Real world problems are quite complex and solving them requires to divide work into different parts. Mainly, there are: the abstraction of interesting and relevant sub-parts, the definition of benchmarks and design of a global model and the application of a particular search strategy. We propose to identify for each of these parts some common pitfalls and to discuss them. We will successively consider undivided model, rigid search, biased benchmarking and wrong abstraction.