The algebra-based introductory statistics course is the most popular undergraduate course in statistics. While there is a general consensus for the content of the curriculum, the recent Guidelines for Assessment and Instruction in Statistics Education (GAISE) have challenged the pedagogy of this course. Additionally, some arguments have been made that the curriculum should focus on a randomization approach to statistical inference instead of using asymptotic tests. We developed a preliminary version of a randomization based curriculum which we then implemented with 240 students in eight sections of introductory statistics in fall 2009. The Comprehensive Assessment of Outcomes in Statistics (CAOS) assessment test was administered to these students and showed that students learned significantly more about statistical inference using the new curriculum, with comparable learning on most other questions. The assessment results demonstrate that refining content, improving pedagogy and rethinking the consensus curriculum can significantly improve student learning. We will continue to refine both content and pedagogy resulting in improved student learning gains on CAOS items and other assessment measures. (Contains 1 figure and 6 tables.)