The problematic of quality is intrinsic to all data analyst. Actuaries like other professions has been dealing with completeness, uncertainty or imprecision through the evolution of insurance. Nowadays, the insurance underwriting is being challenged and is shifting towards creative ways of pricing. The inclusion of external data from open data or specific data brings the dimension of data quality and the model impact further. Through an example of pricing using only the address, this presentation shows how data quality and the interpretability of the models modifies the selection of significant variables and how the problem of quality is a burden in an underwriting process.