Garri is an important food product made from cassava. In the past, selection for agronomic and yield traits was important to the breeders in the early breeding stages. Little effort was made to select end-user traits, leading to the loss of genotypes with good product qualities. To understand what opportunities might exist for selecting end-user traits in an early breeding stage, we estimated the heritability, the GXE effect, and predictability of cassava garri quality traits from a preliminary yield study of two years and two locations. This study provides a framework for predicting cassava garri quality through genomic prediction. In another experiment, we involved cassava farmers in evaluating cassava varieties using the triadic comparison of technologies (tricot) approach to enable farmers to select varieties suitable for their local conditions. We compared on-farm data with on-station data, and we used farmers-generated data to identify what drives farmers' decisions to replant varieties. We observed that on-farm yield correlated poorly with on-station yield. We identified promising varieties using farmers' rank data, and we identified some traits that influence farmers' decisions to replant.