The study consists on the analysis of the accuracy and correspondent computational time of "no-interactive" coupling of the numerical model, when applied to practical engineering studies. Two different models were used in the first part of the domain, where no breaking occurs: COULWAVE and CANAL. For the second part of the domain, where wave-structure interaction phenomena dominates, COBRAS-UC was used forced by the data obtained in the first part of the domain. Results from the coupling technique were compared with a set of experimental results obtained for the wave propagation along a constant slope bottom (1:20) that ends on a sea-wall structure, as well as with the results obtained with COBRAS-UC applied to the whole domain. The main conclusions are that the coupling section position is not relevant in this case, depending primarily on the area where each phenomenon dominates. The magnitudes to be transferred between models should be, at least, η(t) and u(t). The information of v(t) seems to be not so important. The computational time reduces to approximately 50% for coupling CANAL and COBRAS-UC and for 37% when coupling COULWAVE and COBRAS-UC. Coupling models should be choosing taking a balance between computational time and results accuracy.