In this paper we propose a methodology for designing and administering a test of speaking in conditions that are of the highest importance for the testees. With examples of high stake assessments where speaking is not the only ability measured, we will demonstrate how a rigorous and scientific approach offers advantages to the test in terms of validity, reliability, sensitivity and equity. We will also examine how the adepts of this approach are an appropriate answer to the initial request and to the expectation of the final users. The benefits for the stakeholders will also reviewed.