Variable-rate (VR) K application may reduce within-field soil-test K (STK) and grain yield variability while maintaining or increasing crop yield. Replicated strip trials were established in eight Iowa fields to compare VR and uniform-rate (UR) K fertilization for corn-soybean rotations. Treatments were a control, VR based on STK from 0.1-0.3 ha grid soil sampling, and UR based on field-average STK. The K rates were those needed for the 2-year rotation applied before the first crop. Fertilizer was applied using calibrated custom applicator equipment to strips 18 m wide and 360 to 600 m long depending on the field. One trial was evaluated for one rotation cycle and seven trials were evaluated for two rotation cycles with treatments re-application before the third crop. Grain was harvested with yield monitors and DGPS. Yield response to K fertilizer was observed in several fields and years, but seldom there was a significant yield difference between VR and UR methods. High small-scale nutrient variability and use of K fertilizer recommendations for low-testing soils higher than needed to maximize yield may explain the lack of yield difference between fertilization methods. The total amount of fertilizer applied usually was smaller with VR, but the difference varied across fields according to STK levels at each field. Standard deviations (SD) for grain yield indicated inconsistent differences between K application methods. However, SD for STK was lower for VR because more K was applied to low-testing field areas and less or none to high-testing areas. We conclude that VR fertilization does not necessarily results in higher or less variable grain yields but reduces within-field STK variability and avoid excess fertilization of high-testing areas.