In this paper I report the development and use of STACK assessments at the University of Edinburgh during the academic years 2017–2018 and 2018–2019, as part of a department-wide project to improve online assessment. The primary goal is to create high quality questions, with random versions and full worked solutions, for our major first and second year mathematical courses, some with over 600 students. The paper reports our first steps in creating a framework for systematic review and quality improvement. I also report on my experiences in developing questions over the last eighteen months, providing examples and advice for anyone else working in this area.