In this paper a novel general test architecture for measuring speech intelligibility will be introduced. The entire approach is called CLuster-IDentification (CLID). It is a test framework, i.e., a battery of interdependent elementary variables that - in their definite organisational pattern - go to make up any (intelligibility) test. Only by putting these variables into a specific form a particular test is designed. Parameters that have to be itemized when, e. g., the intelligibility of a specific speech output system is aimed at to be assessed are, amongst others, structure of vocabulary, frequency of occurrence of test items, response form, etc. CLID-frame, consequently, is an environment in which factors that have an impact on the assessment procedure (and thus on the test results) are parametrized.