To reduce cheating in written tests and exams, assessors often randomly vary the order of questions across students. However, little is known about the potential unintended side effects of question order. This paper examines whether randomizing students to start with an easier or harder question makes a difference to overall assessment performance in incentivized testing situations under time pressure. Using data from more than 8,000 online tests and exams administered in econometrics and statistics courses at two of Germany's largest universities, we find no evidence that the difficulty of the first question(s) has an effect on overall assessment performance. Our findings are good news for people designing (online) assessments, because randomizing the order of questions can be used as an effective tool to mitigate cheating, but does not affect students' overall performance.