The current market situation is characterized by a significant interest in speaker recognition functionalities in telecommunication systems (e.g. phone banking). This paper presents a field-test assessment of a speaker recognition algorithm, in a realistic context. Such field tests are particularly useful because the requirements for those real-world systems can be significantly different from those focused on by the research laboratories. Therefore, the results presented in this paper are divided into two groups. The quantitative ones describe the performance achieved in terms of Equal Error Rates (EER) as a function of the field-test conditions and different limitations on the enrollment and test duration. On the other hand, we discuss some innovative qualitative outcomes which are mainly based on non-technical but subjective impressions reported by the participants.