Myriad ingenious controlled environment (CE) screens have been proposed over past decades to accelerate genetic improvement for sub-optimal environments. With the recent development of robotic systems for greenhouse management and automated image-based data collection, interest in this effort has grown. Controlled environments are particularly well suited for screening novel transgenes in a confined setting, further intensifying the effort to identify such surrogate screens. History has demonstrated, however, that the vast majority of such screens have been unsuccessful in identifying useful genetic variation to improve crop yield in environments characterized by broad-reaching abiotic stresses such as drought and low nitrogen fertility. This results from fundamental differences in the factors that drive yield in the field at a commercial level relative to the traits that lead to successful performance in CE assays. Transgene screens, including surrogate screens, are often conducted using model species such as Arabidopsis; translation of these results to the crop species may not always be successful for a commercial level of tolerance. Nonetheless, there are examples of surrogate screens that have been successful in identifying meaningful variation. In addition to being firmly anchored in an understanding of crop physiology, another requirement for successful CE assays is rigorous statistical design and analysis. Spatial variation persists in CE screens, and the incorporation of these features in analytical models improves the ability to detect genetic differences. Finally, ongoing evaluations of the repeatability of surrogate assays are necessary to ensure the integrity of the data over time; diverse check entries can be used for this purpose.