Apprentices' performance after vocational educational training (VET) is commonly attributed to more or less effective training. This implies the assumption that learning is significantly affected by vocational instruction (instructional sensitivity). However, the question has not been investigated yet if VETs are effective, i.e., that they foster competence development and if we are able to track those developments with adequate assessments. We propose the modeling of instructional sensitivity via Differential Item Functioning (DIF) and apply this method to 534 apprentices. Results indicate that over the span of initial VET apprentices significantly improve their performance on a competence-based assessment. Moreover, we identify items which are particularly sensitive to instruction and therefore seem to be especially valuable for future development of vocational assessments.