In this paper we are challenging six traditional notions about assessment that are unhelpful when designing 'assessment for learning'-programmes for competency-based education. We are arguing for the following: Reductionism is not the only way to assure rigour in high-stakes assessment; holistic judgements can be equally rigorous. Combining results of assessment parts only because they are of the same format (like different stations in an OSCE) is often not defensible; instead there must be a logically justifiable combination. Numbers describe the quality of the assessment. Therefore, manipulating the numbers is usually not the best way to improve its quality. Not every assessment moment needs to be a decision moment, disconnecting both makes combining summative and formative functions of assessment easier. Standardisation is not the only route to equity. Especially with diverse student groups tailoring is more equitable than standardisation. The most important element to standardise is the quality of the process and not the process itself. Finally, most assessment is too much focussed on detecting deficiencies and not on valuing individual student differences. In competency-based education-especially with a focus on learner orientation-this 'deficiency-model' is not as well aligned as a 'differences-model'.